release: v1.3.10

Co-authored-by: Jaaneek <jankiewiczmilosz@gmail.com>

.opencode/command/changelog.md

@@ -1,22 +1,19 @@
 ---
-model: opencode/kimi-k2.5
+model: opencode/gpt-5.4
 ---
 
 Create `UPCOMING_CHANGELOG.md` from the structured changelog input below.
 If `UPCOMING_CHANGELOG.md` already exists, ignore its current contents completely.
 Do not preserve, merge, or reuse text from the existing file.
 
-Any command arguments are passed directly to `bun script/changelog.ts`.
-Use `--from` / `-f` and `--to` / `-t` to preview a specific release range.
-
 The input already contains the exact commit range since the last non-draft release.
 The commits are already filtered to the release-relevant packages and grouped into
 the release sections. Do not fetch GitHub releases, PRs, or build your own commit list.
 The input may also include a `## Community Contributors Input` section.
 
 Before writing any entry you keep, inspect the real diff with
-`git show --stat --format='' <hash>` or `git show --format='' <hash>` so the
-summary reflects the actual user-facing change and not just the commit message.
+`git show --stat --format='' <hash>` or `git show --format='' <hash>` so you can
+understand the actual code changes and not just the commit message (they may be misleading).
 Do not use `git log` or author metadata when deciding attribution.
 
 Rules:
@@ -38,7 +35,12 @@ Rules:
 - Do not add, remove, rewrite, or reorder contributor names or commit titles in that block
 - Do not derive the thank-you section from the main summary bullets
 - Do not include the heading `## Community Contributors Input` in the final file
+- Focus on writing the least words to get your point across - users will skim read the changelog, so we should be precise
 
-## Changelog Input
+**Importantly, the changelog is for users (who are at least slightly technical), they may use the TUI, Desktop, SDK, Plugins and so forth. Be thorough in understanding flow on effects may not be immediately apparent. e.g. a package upgrade looks internal but may patch a bug. Or a refactor may also stabilise some race condition that fixes bugs for users. The PR title/body + commit message will give you the authors context, usually containing the outcome not just technical detail**
 
-!`bun script/changelog.ts $ARGUMENTS`
+<changelog_input>
+
+!`bun script/raw-changelog.ts $ARGUMENTS`
+
+</changelog_input>

bun.lock

@@ -26,7 +26,7 @@
     },
     "packages/app": {
       "name": "@opencode-ai/app",
-      "version": "1.3.7",
+      "version": "1.3.10",
       "dependencies": {
         "@kobalte/core": "catalog:",
         "@opencode-ai/sdk": "workspace:*",
@@ -79,7 +79,7 @@
     },
     "packages/console/app": {
       "name": "@opencode-ai/console-app",
-      "version": "1.3.7",
+      "version": "1.3.10",
       "dependencies": {
         "@cloudflare/vite-plugin": "1.15.2",
         "@ibm/plex": "6.4.1",
@@ -113,7 +113,7 @@
     },
     "packages/console/core": {
       "name": "@opencode-ai/console-core",
-      "version": "1.3.7",
+      "version": "1.3.10",
       "dependencies": {
         "@aws-sdk/client-sts": "3.782.0",
         "@jsx-email/render": "1.1.1",
@@ -140,7 +140,7 @@
     },
     "packages/console/function": {
       "name": "@opencode-ai/console-function",
-      "version": "1.3.7",
+      "version": "1.3.10",
       "dependencies": {
         "@ai-sdk/anthropic": "3.0.64",
         "@ai-sdk/openai": "3.0.48",
@@ -164,7 +164,7 @@
     },
     "packages/console/mail": {
       "name": "@opencode-ai/console-mail",
-      "version": "1.3.7",
+      "version": "1.3.10",
       "dependencies": {
         "@jsx-email/all": "2.2.3",
         "@jsx-email/cli": "1.4.3",
@@ -188,7 +188,7 @@
     },
     "packages/desktop": {
       "name": "@opencode-ai/desktop",
-      "version": "1.3.7",
+      "version": "1.3.10",
       "dependencies": {
         "@opencode-ai/app": "workspace:*",
         "@opencode-ai/ui": "workspace:*",
@@ -221,7 +221,7 @@
     },
     "packages/desktop-electron": {
       "name": "@opencode-ai/desktop-electron",
-      "version": "1.3.7",
+      "version": "1.3.10",
       "dependencies": {
         "@opencode-ai/app": "workspace:*",
         "@opencode-ai/ui": "workspace:*",
@@ -252,7 +252,7 @@
     },
     "packages/enterprise": {
       "name": "@opencode-ai/enterprise",
-      "version": "1.3.7",
+      "version": "1.3.10",
       "dependencies": {
         "@opencode-ai/ui": "workspace:*",
         "@opencode-ai/util": "workspace:*",
@@ -281,7 +281,7 @@
     },
     "packages/function": {
       "name": "@opencode-ai/function",
-      "version": "1.3.7",
+      "version": "1.3.10",
       "dependencies": {
         "@octokit/auth-app": "8.0.1",
         "@octokit/rest": "catalog:",
@@ -297,7 +297,7 @@
     },
     "packages/opencode": {
       "name": "opencode",
-      "version": "1.3.7",
+      "version": "1.3.10",
       "bin": {
         "opencode": "./bin/opencode",
       },
@@ -323,7 +323,7 @@
         "@ai-sdk/provider-utils": "4.0.21",
         "@ai-sdk/togetherai": "2.0.41",
         "@ai-sdk/vercel": "2.0.39",
-        "@ai-sdk/xai": "3.0.74",
+        "@ai-sdk/xai": "3.0.75",
         "@aws-sdk/credential-providers": "3.993.0",
         "@clack/prompts": "1.0.0-alpha.1",
         "@effect/platform-node": "catalog:",
@@ -423,7 +423,7 @@
     },
     "packages/plugin": {
       "name": "@opencode-ai/plugin",
-      "version": "1.3.7",
+      "version": "1.3.10",
       "dependencies": {
         "@opencode-ai/sdk": "workspace:*",
         "zod": "catalog:",
@@ -457,7 +457,7 @@
     },
     "packages/sdk/js": {
       "name": "@opencode-ai/sdk",
-      "version": "1.3.7",
+      "version": "1.3.10",
       "devDependencies": {
         "@hey-api/openapi-ts": "0.90.10",
         "@tsconfig/node22": "catalog:",
@@ -468,7 +468,7 @@
     },
     "packages/slack": {
       "name": "@opencode-ai/slack",
-      "version": "1.3.7",
+      "version": "1.3.10",
       "dependencies": {
         "@opencode-ai/sdk": "workspace:*",
         "@slack/bolt": "^3.17.1",
@@ -503,7 +503,7 @@
     },
     "packages/ui": {
       "name": "@opencode-ai/ui",
-      "version": "1.3.7",
+      "version": "1.3.10",
       "dependencies": {
         "@kobalte/core": "catalog:",
         "@opencode-ai/sdk": "workspace:*",
@@ -550,7 +550,7 @@
     },
     "packages/util": {
       "name": "@opencode-ai/util",
-      "version": "1.3.7",
+      "version": "1.3.10",
       "dependencies": {
         "zod": "catalog:",
       },
@@ -561,7 +561,7 @@
     },
     "packages/web": {
       "name": "@opencode-ai/web",
-      "version": "1.3.7",
+      "version": "1.3.10",
       "dependencies": {
         "@astrojs/cloudflare": "12.6.3",
         "@astrojs/markdown-remark": "6.3.1",
@@ -719,7 +719,7 @@
 
     "@ai-sdk/vercel": ["@ai-sdk/vercel@2.0.39", "", { "dependencies": { "@ai-sdk/openai-compatible": "2.0.37", "@ai-sdk/provider": "3.0.8", "@ai-sdk/provider-utils": "4.0.21" }, "peerDependencies": { "zod": "^3.25.76 || ^4.1.8" } }, "sha512-8eu3ljJpkCTP4ppcyYB+NcBrkcBoSOFthCSgk5VnjaxnDaOJFaxnPwfddM7wx3RwMk2CiK1O61Px/LlqNc7QkQ=="],
 
-    "@ai-sdk/xai": ["@ai-sdk/xai@3.0.74", "", { "dependencies": { "@ai-sdk/openai-compatible": "2.0.37", "@ai-sdk/provider": "3.0.8", "@ai-sdk/provider-utils": "4.0.21" }, "peerDependencies": { "zod": "^3.25.76 || ^4.1.8" } }, "sha512-HDDLsT+QrzE3c2QZLRV/HKAwMtXDb0PMDdk1PYUXLJ3r9Qv76zGKGyvJLX7Pu6c8TOHD1mwLrOVYrsTpC/eTMw=="],
+    "@ai-sdk/xai": ["@ai-sdk/xai@3.0.75", "", { "dependencies": { "@ai-sdk/openai-compatible": "2.0.37", "@ai-sdk/provider": "3.0.8", "@ai-sdk/provider-utils": "4.0.21" }, "peerDependencies": { "zod": "^3.25.76 || ^4.1.8" } }, "sha512-V8UKK4fNpI9cnrtsZBvUp9O9J6Y9fTKBRoSLyEaNGPirACewixmLDbXsSgAeownPVWiWpK34bFysd+XouI5Ywg=="],
 
     "@alloc/quick-lru": ["@alloc/quick-lru@5.2.0", "", {}, "sha512-UrcABB+4bUrFABwbluTIBErXwvbsU/V7TZWfmbgJfbkwiBuziS9gxdODUyuiecfdGQ85jglMW6juS3+z5TsKLw=="],
 
@@ -5289,6 +5289,8 @@
 
     "accepts/mime-types": ["mime-types@2.1.35", "", { "dependencies": { "mime-db": "1.52.0" } }, "sha512-ZDY+bPm5zTTF+YpCrAU9nK0UgICYPT0QtT1NZWFv4s++TNkcgVaT0g6+4R2uI4MjQjzysHB1zxuWL50hzaeXiw=="],
 
+    "ai-gateway-provider/@ai-sdk/xai": ["@ai-sdk/xai@3.0.74", "", { "dependencies": { "@ai-sdk/openai-compatible": "2.0.37", "@ai-sdk/provider": "3.0.8", "@ai-sdk/provider-utils": "4.0.21" }, "peerDependencies": { "zod": "^3.25.76 || ^4.1.8" } }, "sha512-HDDLsT+QrzE3c2QZLRV/HKAwMtXDb0PMDdk1PYUXLJ3r9Qv76zGKGyvJLX7Pu6c8TOHD1mwLrOVYrsTpC/eTMw=="],
+
     "ajv-keywords/ajv": ["ajv@6.14.0", "", { "dependencies": { "fast-deep-equal": "^3.1.1", "fast-json-stable-stringify": "^2.0.0", "json-schema-traverse": "^0.4.1", "uri-js": "^4.2.2" } }, "sha512-IWrosm/yrn43eiKqkfkHis7QioDleaXQHdDVPKg0FSwwd/DuvyX79TZnFOnYpB7dcsFAMmtFztZuXPDvSePkFw=="],
 
     "ansi-align/string-width": ["string-width@4.2.3", "", { "dependencies": { "emoji-regex": "^8.0.0", "is-fullwidth-code-point": "^3.0.0", "strip-ansi": "^6.0.1" } }, "sha512-wKyQRQpjJ0sIp62ErSZdGsjMJWsap5oRNihHhu6G7JVO/9jIB6UyevL+tXuOqrng8j/cxKTWyWUwvSTriiZz/g=="],

nix/hashes.json

@@ -1,8 +1,8 @@
 {
   "nodeModules": {
-    "x86_64-linux": "sha256-0JgEA54d1ZZ0IWUmxCJP2fnQ2cpmLO25G+hafUbHFLw=",
-    "aarch64-linux": "sha256-oB4Ptc+MH76MEw0DZodmCCz87qOmbzi26751ZM4DYyE=",
-    "aarch64-darwin": "sha256-712rb7B5gTRz1uTx4cJQSrmq9DoBUe+UxbvawYV4XTE=",
-    "x86_64-darwin": "sha256-GRCiEBDDEyVx1et04xqdIEQr3ykRMOBJoQy/xddSsCA="
+    "x86_64-linux": "sha256-5w+DwEvUrCly9LHZuTa1yTSD45X56cGJG8sds/N29mU=",
+    "aarch64-linux": "sha256-pLhyzajYinBlFyGWwPypyC8gHEU8S7fVXIs6aqgBmhg=",
+    "aarch64-darwin": "sha256-vN0sXYs7pLtpq7U9SorR2z6st/wMfHA3dybOnwIh1pU=",
+    "x86_64-darwin": "sha256-P8fgyBcZJmY5VbNxNer/EL4r/F28dNxaqheaqNZH488="
   }
 }

packages/app/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@opencode-ai/app",
-  "version": "1.3.7",
+  "version": "1.3.10",
   "description": "",
   "type": "module",
   "exports": {

packages/app/src/components/prompt-input.tsx

@@ -624,17 +624,18 @@ export const PromptInput: Component<PromptInputProps> = (props) => {
     if (!cmd) return
     promptProbe.select(cmd.id)
     closePopover()
+    const images = imageAttachments()
 
     if (cmd.type === "custom") {
       const text = `/${cmd.trigger} `
       setEditorText(text)
-      prompt.set([{ type: "text", content: text, start: 0, end: text.length }], text.length)
+      prompt.set([{ type: "text", content: text, start: 0, end: text.length }, ...images], text.length)
       focusEditorEnd()
       return
     }
 
     clearEditor()
-    prompt.set([{ type: "text", content: "", start: 0, end: 0 }], 0)
+    prompt.set([...DEFAULT_PROMPT, ...images], 0)
     command.trigger(cmd.id, "slash")
   }
 

packages/app/src/pages/session/composer/session-question-dock.tsx

@@ -11,6 +11,47 @@ import { useSDK } from "@/context/sdk"
 
 const cache = new Map<string, { tab: number; answers: QuestionAnswer[]; custom: string[]; customOn: boolean[] }>()
 
+function Mark(props: { multi: boolean; picked: boolean; onClick?: (event: MouseEvent) => void }) {
+  return (
+    <span data-slot="question-option-check" aria-hidden="true" onClick={props.onClick}>
+      <span data-slot="question-option-box" data-type={props.multi ? "checkbox" : "radio"} data-picked={props.picked}>
+        <Show when={props.multi} fallback={<span data-slot="question-option-radio-dot" />}>
+          <Icon name="check-small" size="small" />
+        </Show>
+      </span>
+    </span>
+  )
+}
+
+function Option(props: {
+  multi: boolean
+  picked: boolean
+  label: string
+  description?: string
+  disabled: boolean
+  onClick: VoidFunction
+}) {
+  return (
+    <button
+      type="button"
+      data-slot="question-option"
+      data-picked={props.picked}
+      role={props.multi ? "checkbox" : "radio"}
+      aria-checked={props.picked}
+      disabled={props.disabled}
+      onClick={props.onClick}
+    >
+      <Mark multi={props.multi} picked={props.picked} />
+      <span data-slot="question-option-main">
+        <span data-slot="option-label">{props.label}</span>
+        <Show when={props.description}>
+          <span data-slot="option-description">{props.description}</span>
+        </Show>
+      </span>
+    </button>
+  )
+}
+
 export const SessionQuestionDock: Component<{ request: QuestionRequest; onSubmit: () => void }> = (props) => {
   const sdk = useSDK()
   const language = useLanguage()
@@ -41,6 +82,9 @@ export const SessionQuestionDock: Component<{ request: QuestionRequest; onSubmit
     return language.t("session.question.progress", { current: n, total: total() })
   })
 
+  const customLabel = () => language.t("ui.messagePart.option.typeOwnAnswer")
+  const customPlaceholder = () => language.t("ui.question.custom.placeholder")
+
   const last = createMemo(() => store.tab >= total() - 1)
 
   const customUpdate = (value: string, selected: boolean = on()) => {
@@ -164,6 +208,13 @@ export const SessionQuestionDock: Component<{ request: QuestionRequest; onSubmit
 
   const submit = () => void reply(questions().map((_, i) => store.answers[i] ?? []))
 
+  const answered = (i: number) => {
+    if ((store.answers[i]?.length ?? 0) > 0) return true
+    return store.customOn[i] === true && (store.custom[i] ?? "").trim().length > 0
+  }
+
+  const picked = (answer: string) => store.answers[store.tab]?.includes(answer) ?? false
+
   const pick = (answer: string, custom: boolean = false) => {
     setStore("answers", store.tab, [answer])
     if (custom) setStore("custom", store.tab, answer)
@@ -230,6 +281,24 @@ export const SessionQuestionDock: Component<{ request: QuestionRequest; onSubmit
     customUpdate(input())
   }
 
+  const resizeInput = (el: HTMLTextAreaElement) => {
+    el.style.height = "0px"
+    el.style.height = `${el.scrollHeight}px`
+  }
+
+  const focusCustom = (el: HTMLTextAreaElement) => {
+    setTimeout(() => {
+      el.focus()
+      resizeInput(el)
+    }, 0)
+  }
+
+  const toggleCustomMark = (event: MouseEvent) => {
+    event.preventDefault()
+    event.stopPropagation()
+    customToggle()
+  }
+
   const next = () => {
     if (sending()) return
     if (store.editing) commitCustom()
@@ -270,10 +339,7 @@ export const SessionQuestionDock: Component<{ request: QuestionRequest; onSubmit
                   type="button"
                   data-slot="question-progress-segment"
                   data-active={i() === store.tab}
-                  data-answered={
-                    (store.answers[i()]?.length ?? 0) > 0 ||
-                    (store.customOn[i()] === true && (store.custom[i()] ?? "").trim().length > 0)
-                  }
+                  data-answered={answered(i())}
                   disabled={sending()}
                   onClick={() => jump(i())}
                   aria-label={`${language.t("ui.tool.questions")} ${i() + 1}`}
@@ -307,43 +373,23 @@ export const SessionQuestionDock: Component<{ request: QuestionRequest; onSubmit
       </Show>
       <div data-slot="question-options">
         <For each={options()}>
-          {(opt, i) => {
-            const picked = () => store.answers[store.tab]?.includes(opt.label) ?? false
-            return (
-              <button
-                data-slot="question-option"
-                data-picked={picked()}
-                role={multi() ? "checkbox" : "radio"}
-                aria-checked={picked()}
-                disabled={sending()}
-                onClick={() => selectOption(i())}
-              >
-                <span data-slot="question-option-check" aria-hidden="true">
-                  <span
-                    data-slot="question-option-box"
-                    data-type={multi() ? "checkbox" : "radio"}
-                    data-picked={picked()}
-                  >
-                    <Show when={multi()} fallback={<span data-slot="question-option-radio-dot" />}>
-                      <Icon name="check-small" size="small" />
-                    </Show>
-                  </span>
-                </span>
-                <span data-slot="question-option-main">
-                  <span data-slot="option-label">{opt.label}</span>
-                  <Show when={opt.description}>
-                    <span data-slot="option-description">{opt.description}</span>
-                  </Show>
-                </span>
-              </button>
-            )
-          }}
+          {(opt, i) => (
+            <Option
+              multi={multi()}
+              picked={picked(opt.label)}
+              label={opt.label}
+              description={opt.description}
+              disabled={sending()}
+              onClick={() => selectOption(i())}
+            />
+          )}
         </For>
 
         <Show
           when={store.editing}
           fallback={
             <button
+              type="button"
               data-slot="question-option"
               data-custom="true"
               data-picked={on()}
@@ -352,24 +398,10 @@ export const SessionQuestionDock: Component<{ request: QuestionRequest; onSubmit
               disabled={sending()}
               onClick={customOpen}
             >
-              <span
-                data-slot="question-option-check"
-                aria-hidden="true"
-                onClick={(e) => {
-                  e.preventDefault()
-                  e.stopPropagation()
-                  customToggle()
-                }}
-              >
-                <span data-slot="question-option-box" data-type={multi() ? "checkbox" : "radio"} data-picked={on()}>
-                  <Show when={multi()} fallback={<span data-slot="question-option-radio-dot" />}>
-                    <Icon name="check-small" size="small" />
-                  </Show>
-                </span>
-              </span>
+              <Mark multi={multi()} picked={on()} onClick={toggleCustomMark} />
               <span data-slot="question-option-main">
-                <span data-slot="option-label">{language.t("ui.messagePart.option.typeOwnAnswer")}</span>
-                <span data-slot="option-description">{input() || language.t("ui.question.custom.placeholder")}</span>
+                <span data-slot="option-label">{customLabel()}</span>
+                <span data-slot="option-description">{input() || customPlaceholder()}</span>
               </span>
             </button>
           }
@@ -394,33 +426,13 @@ export const SessionQuestionDock: Component<{ request: QuestionRequest; onSubmit
               commitCustom()
             }}
           >
-            <span
-              data-slot="question-option-check"
-              aria-hidden="true"
-              onClick={(e) => {
-                e.preventDefault()
-                e.stopPropagation()
-                customToggle()
-              }}
-            >
-              <span data-slot="question-option-box" data-type={multi() ? "checkbox" : "radio"} data-picked={on()}>
-                <Show when={multi()} fallback={<span data-slot="question-option-radio-dot" />}>
-                  <Icon name="check-small" size="small" />
-                </Show>
-              </span>
-            </span>
+            <Mark multi={multi()} picked={on()} onClick={toggleCustomMark} />
             <span data-slot="question-option-main">
-              <span data-slot="option-label">{language.t("ui.messagePart.option.typeOwnAnswer")}</span>
+              <span data-slot="option-label">{customLabel()}</span>
               <textarea
-                ref={(el) =>
-                  setTimeout(() => {
-                    el.focus()
-                    el.style.height = "0px"
-                    el.style.height = `${el.scrollHeight}px`
-                  }, 0)
-                }
+                ref={focusCustom}
                 data-slot="question-custom-input"
-                placeholder={language.t("ui.question.custom.placeholder")}
+                placeholder={customPlaceholder()}
                 value={input()}
                 rows={1}
                 disabled={sending()}
@@ -436,8 +448,7 @@ export const SessionQuestionDock: Component<{ request: QuestionRequest; onSubmit
                 }}
                 onInput={(e) => {
                   customUpdate(e.currentTarget.value)
-                  e.currentTarget.style.height = "0px"
-                  e.currentTarget.style.height = `${e.currentTarget.scrollHeight}px`
+                  resizeInput(e.currentTarget)
                 }}
               />
             </span>

packages/app/src/pages/session/file-tabs.tsx

@@ -52,6 +52,132 @@ function FileCommentMenu(props: {
   )
 }
 
+type ScrollPos = { x: number; y: number }
+
+function createScrollSync(input: { tab: () => string; view: ReturnType<typeof useSessionLayout>["view"] }) {
+  let scroll: HTMLDivElement | undefined
+  let scrollFrame: number | undefined
+  let restoreFrame: number | undefined
+  let pending: ScrollPos | undefined
+  let code: HTMLElement[] = []
+
+  const getCode = () => {
+    const el = scroll
+    if (!el) return []
+
+    const host = el.querySelector("diffs-container")
+    if (!(host instanceof HTMLElement)) return []
+
+    const root = host.shadowRoot
+    if (!root) return []
+
+    return Array.from(root.querySelectorAll("[data-code]")).filter(
+      (node): node is HTMLElement => node instanceof HTMLElement && node.clientWidth > 0,
+    )
+  }
+
+  const save = (next: ScrollPos) => {
+    pending = next
+    if (scrollFrame !== undefined) return
+
+    scrollFrame = requestAnimationFrame(() => {
+      scrollFrame = undefined
+
+      const out = pending
+      pending = undefined
+      if (!out) return
+
+      input.view().setScroll(input.tab(), out)
+    })
+  }
+
+  const onCodeScroll = (event: Event) => {
+    const el = scroll
+    if (!el) return
+
+    const target = event.currentTarget
+    if (!(target instanceof HTMLElement)) return
+
+    save({
+      x: target.scrollLeft,
+      y: el.scrollTop,
+    })
+  }
+
+  const sync = () => {
+    const next = getCode()
+    if (next.length === code.length && next.every((el, i) => el === code[i])) return
+
+    for (const item of code) {
+      item.removeEventListener("scroll", onCodeScroll)
+    }
+
+    code = next
+
+    for (const item of code) {
+      item.addEventListener("scroll", onCodeScroll)
+    }
+  }
+
+  const restore = () => {
+    const el = scroll
+    if (!el) return
+
+    const pos = input.view().scroll(input.tab())
+    if (!pos) return
+
+    sync()
+
+    if (code.length > 0) {
+      for (const item of code) {
+        if (item.scrollLeft !== pos.x) item.scrollLeft = pos.x
+      }
+    }
+
+    if (el.scrollTop !== pos.y) el.scrollTop = pos.y
+    if (code.length > 0) return
+    if (el.scrollLeft !== pos.x) el.scrollLeft = pos.x
+  }
+
+  const queueRestore = () => {
+    if (restoreFrame !== undefined) return
+
+    restoreFrame = requestAnimationFrame(() => {
+      restoreFrame = undefined
+      restore()
+    })
+  }
+
+  const handleScroll = (event: Event & { currentTarget: HTMLDivElement }) => {
+    if (code.length === 0) sync()
+
+    save({
+      x: code[0]?.scrollLeft ?? event.currentTarget.scrollLeft,
+      y: event.currentTarget.scrollTop,
+    })
+  }
+
+  const setViewport = (el: HTMLDivElement) => {
+    scroll = el
+    restore()
+  }
+
+  onCleanup(() => {
+    for (const item of code) {
+      item.removeEventListener("scroll", onCodeScroll)
+    }
+
+    if (scrollFrame !== undefined) cancelAnimationFrame(scrollFrame)
+    if (restoreFrame !== undefined) cancelAnimationFrame(restoreFrame)
+  })
+
+  return {
+    handleScroll,
+    queueRestore,
+    setViewport,
+  }
+}
+
 export function FileTabContent(props: { tab: string }) {
   const file = useFile()
   const comments = useComments()
@@ -65,11 +191,6 @@ export function FileTabContent(props: { tab: string }) {
     normalizeTab: (tab) => (tab.startsWith("file://") ? file.tab(tab) : tab),
   }).activeFileTab
 
-  let scroll: HTMLDivElement | undefined
-  let scrollFrame: number | undefined
-  let restoreFrame: number | undefined
-  let pending: { x: number; y: number } | undefined
-  let codeScroll: HTMLElement[] = []
   let find: FileSearchHandle | null = null
 
   const search = {
@@ -92,6 +213,10 @@ export function FileTabContent(props: { tab: string }) {
     if (file.ready()) return (file.selectedLines(p) as SelectedLineRange | undefined) ?? null
     return (getSessionHandoff(sessionKey())?.files[p] as SelectedLineRange | undefined) ?? null
   })
+  const scrollSync = createScrollSync({
+    tab: () => props.tab,
+    view,
+  })
 
   const selectionPreview = (source: string, selection: FileSelection) => {
     return previewSelectedLines(source, {
@@ -100,6 +225,12 @@ export function FileTabContent(props: { tab: string }) {
     })
   }
 
+  const buildPreview = (filePath: string, selection: FileSelection) => {
+    const source = filePath === path() ? contents() : file.get(filePath)?.content?.content
+    if (!source) return undefined
+    return selectionPreview(source, selection)
+  }
+
   const addCommentToContext = (input: {
     file: string
     selection: SelectedLineRange
@@ -108,14 +239,7 @@ export function FileTabContent(props: { tab: string }) {
     origin?: "review" | "file"
   }) => {
     const selection = selectionFromLines(input.selection)
-    const preview =
-      input.preview ??
-      (() => {
-        if (input.file === path()) return selectionPreview(contents(), selection)
-        const source = file.get(input.file)?.content?.content
-        if (!source) return undefined
-        return selectionPreview(source, selection)
-      })()
+    const preview = input.preview ?? buildPreview(input.file, selection)
 
     const saved = comments.add({
       file: input.file,
@@ -140,8 +264,7 @@ export function FileTabContent(props: { tab: string }) {
     comment: string
   }) => {
     comments.update(input.file, input.id, input.comment)
-    const preview =
-      input.file === path() ? selectionPreview(contents(), selectionFromLines(input.selection)) : undefined
+    const preview = input.file === path() ? buildPreview(input.file, selectionFromLines(input.selection)) : undefined
     prompt.context.updateComment(input.file, input.id, {
       comment: input.comment,
       ...(preview ? { preview } : {}),
@@ -260,102 +383,6 @@ export function FileTabContent(props: { tab: string }) {
     requestAnimationFrame(() => comments.clearFocus())
   })
 
-  const getCodeScroll = () => {
-    const el = scroll
-    if (!el) return []
-
-    const host = el.querySelector("diffs-container")
-    if (!(host instanceof HTMLElement)) return []
-
-    const root = host.shadowRoot
-    if (!root) return []
-
-    return Array.from(root.querySelectorAll("[data-code]")).filter(
-      (node): node is HTMLElement => node instanceof HTMLElement && node.clientWidth > 0,
-    )
-  }
-
-  const queueScrollUpdate = (next: { x: number; y: number }) => {
-    pending = next
-    if (scrollFrame !== undefined) return
-
-    scrollFrame = requestAnimationFrame(() => {
-      scrollFrame = undefined
-
-      const out = pending
-      pending = undefined
-      if (!out) return
-
-      view().setScroll(props.tab, out)
-    })
-  }
-
-  const handleCodeScroll = (event: Event) => {
-    const el = scroll
-    if (!el) return
-
-    const target = event.currentTarget
-    if (!(target instanceof HTMLElement)) return
-
-    queueScrollUpdate({
-      x: target.scrollLeft,
-      y: el.scrollTop,
-    })
-  }
-
-  const syncCodeScroll = () => {
-    const next = getCodeScroll()
-    if (next.length === codeScroll.length && next.every((el, i) => el === codeScroll[i])) return
-
-    for (const item of codeScroll) {
-      item.removeEventListener("scroll", handleCodeScroll)
-    }
-
-    codeScroll = next
-
-    for (const item of codeScroll) {
-      item.addEventListener("scroll", handleCodeScroll)
-    }
-  }
-
-  const restoreScroll = () => {
-    const el = scroll
-    if (!el) return
-
-    const s = view().scroll(props.tab)
-    if (!s) return
-
-    syncCodeScroll()
-
-    if (codeScroll.length > 0) {
-      for (const item of codeScroll) {
-        if (item.scrollLeft !== s.x) item.scrollLeft = s.x
-      }
-    }
-
-    if (el.scrollTop !== s.y) el.scrollTop = s.y
-    if (codeScroll.length > 0) return
-    if (el.scrollLeft !== s.x) el.scrollLeft = s.x
-  }
-
-  const queueRestore = () => {
-    if (restoreFrame !== undefined) return
-
-    restoreFrame = requestAnimationFrame(() => {
-      restoreFrame = undefined
-      restoreScroll()
-    })
-  }
-
-  const handleScroll = (event: Event & { currentTarget: HTMLDivElement }) => {
-    if (codeScroll.length === 0) syncCodeScroll()
-
-    queueScrollUpdate({
-      x: codeScroll[0]?.scrollLeft ?? event.currentTarget.scrollLeft,
-      y: event.currentTarget.scrollTop,
-    })
-  }
-
   const cancelCommenting = () => {
     const p = path()
     if (p) file.setSelectedLines(p, null)
@@ -375,16 +402,7 @@ export function FileTabContent(props: { tab: string }) {
     const restore = (loaded && !prev.loaded) || (ready && !prev.ready) || (active && loaded && !prev.active)
     prev = { loaded, ready, active }
     if (!restore) return
-    queueRestore()
-  })
-
-  onCleanup(() => {
-    for (const item of codeScroll) {
-      item.removeEventListener("scroll", handleCodeScroll)
-    }
-
-    if (scrollFrame !== undefined) cancelAnimationFrame(scrollFrame)
-    if (restoreFrame !== undefined) cancelAnimationFrame(restoreFrame)
+    scrollSync.queueRestore()
   })
 
   const renderFile = (source: string) => (
@@ -402,7 +420,7 @@ export function FileTabContent(props: { tab: string }) {
         selectedLines={activeSelection()}
         commentedLines={commentedLines()}
         onRendered={() => {
-          queueRestore()
+          scrollSync.queueRestore()
         }}
         annotations={commentsUi.annotations()}
         renderAnnotation={commentsUi.renderAnnotation}
@@ -420,7 +438,7 @@ export function FileTabContent(props: { tab: string }) {
           mode: "auto",
           path: path(),
           current: state()?.content,
-          onLoad: queueRestore,
+          onLoad: scrollSync.queueRestore,
           onError: (args: { kind: "image" | "audio" | "svg" }) => {
             if (args.kind !== "svg") return
             showToast({
@@ -435,14 +453,7 @@ export function FileTabContent(props: { tab: string }) {
 
   return (
     <Tabs.Content value={props.tab} class="mt-3 relative h-full">
-      <ScrollView
-        class="h-full"
-        viewportRef={(el: HTMLDivElement) => {
-          scroll = el
-          restoreScroll()
-        }}
-        onScroll={handleScroll as any}
-      >
+      <ScrollView class="h-full" viewportRef={scrollSync.setViewport} onScroll={scrollSync.handleScroll as any}>
         <Switch>
           <Match when={state()?.loaded}>{renderFile(contents())}</Match>
           <Match when={state()?.loading}>

packages/app/src/pages/session/use-session-commands.tsx

@@ -128,380 +128,452 @@ export const useSessionCommands = (actions: SessionCommandContext) => {
     if (sessionID) return permission.isAutoAccepting(sessionID, sdk.directory)
     return permission.isAutoAcceptingDirectory(sdk.directory)
   }
-  command.register("session", () => {
-    const share =
-      sync.data.config.share === "disabled"
-        ? []
-        : [
-            sessionCommand({
-              id: "session.share",
-              title: info()?.share?.url
-                ? language.t("session.share.copy.copyLink")
-                : language.t("command.session.share"),
-              description: info()?.share?.url
-                ? language.t("toast.session.share.success.description")
-                : language.t("command.session.share.description"),
-              slash: "share",
-              disabled: !params.id,
-              onSelect: async () => {
-                if (!params.id) return
+  const write = async (value: string) => {
+    const body = typeof document === "undefined" ? undefined : document.body
+    if (body) {
+      const textarea = document.createElement("textarea")
+      textarea.value = value
+      textarea.setAttribute("readonly", "")
+      textarea.style.position = "fixed"
+      textarea.style.opacity = "0"
+      textarea.style.pointerEvents = "none"
+      body.appendChild(textarea)
+      textarea.select()
+      const copied = document.execCommand("copy")
+      body.removeChild(textarea)
+      if (copied) return true
+    }
 
-                const write = (value: string) => {
-                  const body = typeof document === "undefined" ? undefined : document.body
-                  if (body) {
-                    const textarea = document.createElement("textarea")
-                    textarea.value = value
-                    textarea.setAttribute("readonly", "")
-                    textarea.style.position = "fixed"
-                    textarea.style.opacity = "0"
-                    textarea.style.pointerEvents = "none"
-                    body.appendChild(textarea)
-                    textarea.select()
-                    const copied = document.execCommand("copy")
-                    body.removeChild(textarea)
-                    if (copied) return Promise.resolve(true)
-                  }
+    const clipboard = typeof navigator === "undefined" ? undefined : navigator.clipboard
+    if (!clipboard?.writeText) return false
+    return clipboard.writeText(value).then(
+      () => true,
+      () => false,
+    )
+  }
 
-                  const clipboard = typeof navigator === "undefined" ? undefined : navigator.clipboard
-                  if (!clipboard?.writeText) return Promise.resolve(false)
-                  return clipboard.writeText(value).then(
-                    () => true,
-                    () => false,
-                  )
-                }
+  const copyShare = async (url: string, existing: boolean) => {
+    if (!(await write(url))) {
+      showToast({
+        title: language.t("toast.session.share.copyFailed.title"),
+        variant: "error",
+      })
+      return
+    }
 
-                const copy = async (url: string, existing: boolean) => {
-                  const ok = await write(url)
-                  if (!ok) {
-                    showToast({
-                      title: language.t("toast.session.share.copyFailed.title"),
-                      variant: "error",
-                    })
-                    return
-                  }
+    showToast({
+      title: existing ? language.t("session.share.copy.copied") : language.t("toast.session.share.success.title"),
+      description: language.t("toast.session.share.success.description"),
+      variant: "success",
+    })
+  }
 
-                  showToast({
-                    title: existing
-                      ? language.t("session.share.copy.copied")
-                      : language.t("toast.session.share.success.title"),
-                    description: language.t("toast.session.share.success.description"),
-                    variant: "success",
-                  })
-                }
+  const share = async () => {
+    const sessionID = params.id
+    if (!sessionID) return
 
-                const existing = info()?.share?.url
-                if (existing) {
-                  await copy(existing, true)
-                  return
-                }
+    const existing = info()?.share?.url
+    if (existing) {
+      await copyShare(existing, true)
+      return
+    }
 
-                const url = await sdk.client.session
-                  .share({ sessionID: params.id })
-                  .then((res) => res.data?.share?.url)
-                  .catch(() => undefined)
-                if (!url) {
-                  showToast({
-                    title: language.t("toast.session.share.failed.title"),
-                    description: language.t("toast.session.share.failed.description"),
-                    variant: "error",
-                  })
-                  return
-                }
+    const url = await sdk.client.session
+      .share({ sessionID })
+      .then((res) => res.data?.share?.url)
+      .catch(() => undefined)
+    if (!url) {
+      showToast({
+        title: language.t("toast.session.share.failed.title"),
+        description: language.t("toast.session.share.failed.description"),
+        variant: "error",
+      })
+      return
+    }
 
-                await copy(url, false)
-              },
-            }),
-            sessionCommand({
-              id: "session.unshare",
-              title: language.t("command.session.unshare"),
-              description: language.t("command.session.unshare.description"),
-              slash: "unshare",
-              disabled: !params.id || !info()?.share?.url,
-              onSelect: async () => {
-                if (!params.id) return
-                await sdk.client.session
-                  .unshare({ sessionID: params.id })
-                  .then(() =>
-                    showToast({
-                      title: language.t("toast.session.unshare.success.title"),
-                      description: language.t("toast.session.unshare.success.description"),
-                      variant: "success",
-                    }),
-                  )
-                  .catch(() =>
-                    showToast({
-                      title: language.t("toast.session.unshare.failed.title"),
-                      description: language.t("toast.session.unshare.failed.description"),
-                      variant: "error",
-                    }),
-                  )
-              },
-            }),
-          ]
+    await copyShare(url, false)
+  }
 
+  const unshare = async () => {
+    const sessionID = params.id
+    if (!sessionID) return
+
+    await sdk.client.session
+      .unshare({ sessionID })
+      .then(() =>
+        showToast({
+          title: language.t("toast.session.unshare.success.title"),
+          description: language.t("toast.session.unshare.success.description"),
+          variant: "success",
+        }),
+      )
+      .catch(() =>
+        showToast({
+          title: language.t("toast.session.unshare.failed.title"),
+          description: language.t("toast.session.unshare.failed.description"),
+          variant: "error",
+        }),
+      )
+  }
+
+  const openFile = () => {
+    void import("@/components/dialog-select-file").then((x) => {
+      dialog.show(() => <x.DialogSelectFile onOpenFile={showAllFiles} />)
+    })
+  }
+
+  const closeTab = () => {
+    const tab = closableTab()
+    if (!tab) return
+    tabs().close(tab)
+  }
+
+  const addSelection = () => {
+    const tab = activeFileTab()
+    if (!tab) return
+
+    const path = file.pathFromTab(tab)
+    if (!path) return
+
+    const range = file.selectedLines(path) as SelectedLineRange | null | undefined
+    if (!range) {
+      showToast({
+        title: language.t("toast.context.noLineSelection.title"),
+        description: language.t("toast.context.noLineSelection.description"),
+      })
+      return
+    }
+
+    addSelectionToContext(path, selectionFromLines(range))
+  }
+
+  const openTerminal = () => {
+    if (terminal.all().length > 0) terminal.new()
+    view().terminal.open()
+  }
+
+  const chooseModel = () => {
+    void import("@/components/dialog-select-model").then((x) => {
+      dialog.show(() => <x.DialogSelectModel model={local.model} />)
+    })
+  }
+
+  const chooseMcp = () => {
+    void import("@/components/dialog-select-mcp").then((x) => {
+      dialog.show(() => <x.DialogSelectMcp />)
+    })
+  }
+
+  const toggleAutoAccept = () => {
+    const sessionID = params.id
+    if (sessionID) permission.toggleAutoAccept(sessionID, sdk.directory)
+    else permission.toggleAutoAcceptDirectory(sdk.directory)
+
+    const active = sessionID
+      ? permission.isAutoAccepting(sessionID, sdk.directory)
+      : permission.isAutoAcceptingDirectory(sdk.directory)
+    showToast({
+      title: active
+        ? language.t("toast.permissions.autoaccept.on.title")
+        : language.t("toast.permissions.autoaccept.off.title"),
+      description: active
+        ? language.t("toast.permissions.autoaccept.on.description")
+        : language.t("toast.permissions.autoaccept.off.description"),
+    })
+  }
+
+  const undo = async () => {
+    const sessionID = params.id
+    if (!sessionID) return
+
+    if (status().type !== "idle") {
+      await sdk.client.session.abort({ sessionID }).catch(() => {})
+    }
+
+    const revert = info()?.revert?.messageID
+    const message = findLast(userMessages(), (x) => !revert || x.id < revert)
+    if (!message) return
+
+    await sdk.client.session.revert({ sessionID, messageID: message.id })
+    const parts = sync.data.part[message.id]
+    if (parts) {
+      const restored = extractPromptFromParts(parts, { directory: sdk.directory })
+      prompt.set(restored)
+    }
+
+    const prev = findLast(userMessages(), (x) => x.id < message.id)
+    setActiveMessage(prev)
+  }
+
+  const redo = async () => {
+    const sessionID = params.id
+    if (!sessionID) return
+
+    const revertMessageID = info()?.revert?.messageID
+    if (!revertMessageID) return
+
+    const next = userMessages().find((x) => x.id > revertMessageID)
+    if (!next) {
+      await sdk.client.session.unrevert({ sessionID })
+      prompt.reset()
+      const last = findLast(userMessages(), (x) => x.id >= revertMessageID)
+      setActiveMessage(last)
+      return
+    }
+
+    await sdk.client.session.revert({ sessionID, messageID: next.id })
+    const prev = findLast(userMessages(), (x) => x.id < next.id)
+    setActiveMessage(prev)
+  }
+
+  const compact = async () => {
+    const sessionID = params.id
+    if (!sessionID) return
+
+    const model = local.model.current()
+    if (!model) {
+      showToast({
+        title: language.t("toast.model.none.title"),
+        description: language.t("toast.model.none.description"),
+      })
+      return
+    }
+
+    await sdk.client.session.summarize({
+      sessionID,
+      modelID: model.id,
+      providerID: model.provider.id,
+    })
+  }
+
+  const fork = () => {
+    void import("@/components/dialog-fork").then((x) => {
+      dialog.show(() => <x.DialogFork />)
+    })
+  }
+
+  const shareCmds = () => {
+    if (sync.data.config.share === "disabled") return []
     return [
       sessionCommand({
-        id: "session.new",
-        title: language.t("command.session.new"),
-        keybind: "mod+shift+s",
-        slash: "new",
-        onSelect: () => navigate(`/${params.dir}/session`),
-      }),
-      fileCommand({
-        id: "file.open",
-        title: language.t("command.file.open"),
-        description: language.t("palette.search.placeholder"),
-        keybind: "mod+k,mod+p",
-        slash: "open",
-        onSelect: () => {
-          void import("@/components/dialog-select-file").then((x) => {
-            dialog.show(() => <x.DialogSelectFile onOpenFile={showAllFiles} />)
-          })
-        },
-      }),
-      fileCommand({
-        id: "tab.close",
-        title: language.t("command.tab.close"),
-        keybind: "mod+w",
-        disabled: !closableTab(),
-        onSelect: () => {
-          const tab = closableTab()
-          if (!tab) return
-          tabs().close(tab)
-        },
-      }),
-      contextCommand({
-        id: "context.addSelection",
-        title: language.t("command.context.addSelection"),
-        description: language.t("command.context.addSelection.description"),
-        keybind: "mod+shift+l",
-        disabled: !canAddSelectionContext(),
-        onSelect: () => {
-          const tab = activeFileTab()
-          if (!tab) return
-          const path = file.pathFromTab(tab)
-          if (!path) return
-
-          const range = file.selectedLines(path) as SelectedLineRange | null | undefined
-          if (!range) {
-            showToast({
-              title: language.t("toast.context.noLineSelection.title"),
-              description: language.t("toast.context.noLineSelection.description"),
-            })
-            return
-          }
-
-          addSelectionToContext(path, selectionFromLines(range))
-        },
-      }),
-      viewCommand({
-        id: "terminal.toggle",
-        title: language.t("command.terminal.toggle"),
-        keybind: "ctrl+`",
-        slash: "terminal",
-        onSelect: () => view().terminal.toggle(),
-      }),
-      viewCommand({
-        id: "review.toggle",
-        title: language.t("command.review.toggle"),
-        keybind: "mod+shift+r",
-        onSelect: () => view().reviewPanel.toggle(),
-      }),
-      viewCommand({
-        id: "fileTree.toggle",
-        title: language.t("command.fileTree.toggle"),
-        keybind: "mod+\\",
-        onSelect: () => layout.fileTree.toggle(),
-      }),
-      viewCommand({
-        id: "input.focus",
-        title: language.t("command.input.focus"),
-        keybind: "ctrl+l",
-        onSelect: focusInput,
-      }),
-      terminalCommand({
-        id: "terminal.new",
-        title: language.t("command.terminal.new"),
-        description: language.t("command.terminal.new.description"),
-        keybind: "ctrl+alt+t",
-        onSelect: () => {
-          if (terminal.all().length > 0) terminal.new()
-          view().terminal.open()
-        },
-      }),
-      sessionCommand({
-        id: "message.previous",
-        title: language.t("command.message.previous"),
-        description: language.t("command.message.previous.description"),
-        keybind: "mod+alt+[",
+        id: "session.share",
+        title: info()?.share?.url ? language.t("session.share.copy.copyLink") : language.t("command.session.share"),
+        description: info()?.share?.url
+          ? language.t("toast.session.share.success.description")
+          : language.t("command.session.share.description"),
+        slash: "share",
         disabled: !params.id,
-        onSelect: () => navigateMessageByOffset(-1),
+        onSelect: share,
       }),
       sessionCommand({
-        id: "message.next",
-        title: language.t("command.message.next"),
-        description: language.t("command.message.next.description"),
-        keybind: "mod+alt+]",
-        disabled: !params.id,
-        onSelect: () => navigateMessageByOffset(1),
+        id: "session.unshare",
+        title: language.t("command.session.unshare"),
+        description: language.t("command.session.unshare.description"),
+        slash: "unshare",
+        disabled: !params.id || !info()?.share?.url,
+        onSelect: unshare,
       }),
-      modelCommand({
-        id: "model.choose",
-        title: language.t("command.model.choose"),
-        description: language.t("command.model.choose.description"),
-        keybind: "mod+'",
-        slash: "model",
-        onSelect: () => {
-          void import("@/components/dialog-select-model").then((x) => {
-            dialog.show(() => <x.DialogSelectModel model={local.model} />)
-          })
-        },
-      }),
-      mcpCommand({
-        id: "mcp.toggle",
-        title: language.t("command.mcp.toggle"),
-        description: language.t("command.mcp.toggle.description"),
-        keybind: "mod+;",
-        slash: "mcp",
-        onSelect: () => {
-          void import("@/components/dialog-select-mcp").then((x) => {
-            dialog.show(() => <x.DialogSelectMcp />)
-          })
-        },
-      }),
-      agentCommand({
-        id: "agent.cycle",
-        title: language.t("command.agent.cycle"),
-        description: language.t("command.agent.cycle.description"),
-        keybind: "mod+.",
-        slash: "agent",
-        onSelect: () => local.agent.move(1),
-      }),
-      agentCommand({
-        id: "agent.cycle.reverse",
-        title: language.t("command.agent.cycle.reverse"),
-        description: language.t("command.agent.cycle.reverse.description"),
-        keybind: "shift+mod+.",
-        onSelect: () => local.agent.move(-1),
-      }),
-      modelCommand({
-        id: "model.variant.cycle",
-        title: language.t("command.model.variant.cycle"),
-        description: language.t("command.model.variant.cycle.description"),
-        keybind: "shift+mod+d",
-        onSelect: () => local.model.variant.cycle(),
-      }),
-      permissionsCommand({
-        id: "permissions.autoaccept",
-        title: isAutoAcceptActive()
-          ? language.t("command.permissions.autoaccept.disable")
-          : language.t("command.permissions.autoaccept.enable"),
-        keybind: "mod+shift+a",
-        disabled: false,
-        onSelect: () => {
-          const sessionID = params.id
-          if (sessionID) permission.toggleAutoAccept(sessionID, sdk.directory)
-          else permission.toggleAutoAcceptDirectory(sdk.directory)
-
-          const active = sessionID
-            ? permission.isAutoAccepting(sessionID, sdk.directory)
-            : permission.isAutoAcceptingDirectory(sdk.directory)
-          showToast({
-            title: active
-              ? language.t("toast.permissions.autoaccept.on.title")
-              : language.t("toast.permissions.autoaccept.off.title"),
-            description: active
-              ? language.t("toast.permissions.autoaccept.on.description")
-              : language.t("toast.permissions.autoaccept.off.description"),
-          })
-        },
-      }),
-      sessionCommand({
-        id: "session.undo",
-        title: language.t("command.session.undo"),
-        description: language.t("command.session.undo.description"),
-        slash: "undo",
-        disabled: !params.id || visibleUserMessages().length === 0,
-        onSelect: async () => {
-          const sessionID = params.id
-          if (!sessionID) return
-          if (status().type !== "idle") {
-            await sdk.client.session.abort({ sessionID }).catch(() => {})
-          }
-          const revert = info()?.revert?.messageID
-          const message = findLast(userMessages(), (x) => !revert || x.id < revert)
-          if (!message) return
-          await sdk.client.session.revert({ sessionID, messageID: message.id })
-          const parts = sync.data.part[message.id]
-          if (parts) {
-            const restored = extractPromptFromParts(parts, { directory: sdk.directory })
-            prompt.set(restored)
-          }
-          const priorMessage = findLast(userMessages(), (x) => x.id < message.id)
-          setActiveMessage(priorMessage)
-        },
-      }),
-      sessionCommand({
-        id: "session.redo",
-        title: language.t("command.session.redo"),
-        description: language.t("command.session.redo.description"),
-        slash: "redo",
-        disabled: !params.id || !info()?.revert?.messageID,
-        onSelect: async () => {
-          const sessionID = params.id
-          if (!sessionID) return
-          const revertMessageID = info()?.revert?.messageID
-          if (!revertMessageID) return
-          const nextMessage = userMessages().find((x) => x.id > revertMessageID)
-          if (!nextMessage) {
-            await sdk.client.session.unrevert({ sessionID })
-            prompt.reset()
-            const lastMsg = findLast(userMessages(), (x) => x.id >= revertMessageID)
-            setActiveMessage(lastMsg)
-            return
-          }
-          await sdk.client.session.revert({ sessionID, messageID: nextMessage.id })
-          const priorMsg = findLast(userMessages(), (x) => x.id < nextMessage.id)
-          setActiveMessage(priorMsg)
-        },
-      }),
-      sessionCommand({
-        id: "session.compact",
-        title: language.t("command.session.compact"),
-        description: language.t("command.session.compact.description"),
-        slash: "compact",
-        disabled: !params.id || visibleUserMessages().length === 0,
-        onSelect: async () => {
-          const sessionID = params.id
-          if (!sessionID) return
-          const model = local.model.current()
-          if (!model) {
-            showToast({
-              title: language.t("toast.model.none.title"),
-              description: language.t("toast.model.none.description"),
-            })
-            return
-          }
-          await sdk.client.session.summarize({
-            sessionID,
-            modelID: model.id,
-            providerID: model.provider.id,
-          })
-        },
-      }),
-      sessionCommand({
-        id: "session.fork",
-        title: language.t("command.session.fork"),
-        description: language.t("command.session.fork.description"),
-        slash: "fork",
-        disabled: !params.id || visibleUserMessages().length === 0,
-        onSelect: () => {
-          void import("@/components/dialog-fork").then((x) => {
-            dialog.show(() => <x.DialogFork />)
-          })
-        },
-      }),
-      ...share,
     ]
-  })
+  }
+
+  const sessionCmds = () => [
+    sessionCommand({
+      id: "session.new",
+      title: language.t("command.session.new"),
+      keybind: "mod+shift+s",
+      slash: "new",
+      onSelect: () => navigate(`/${params.dir}/session`),
+    }),
+    sessionCommand({
+      id: "session.undo",
+      title: language.t("command.session.undo"),
+      description: language.t("command.session.undo.description"),
+      slash: "undo",
+      disabled: !params.id || visibleUserMessages().length === 0,
+      onSelect: undo,
+    }),
+    sessionCommand({
+      id: "session.redo",
+      title: language.t("command.session.redo"),
+      description: language.t("command.session.redo.description"),
+      slash: "redo",
+      disabled: !params.id || !info()?.revert?.messageID,
+      onSelect: redo,
+    }),
+    sessionCommand({
+      id: "session.compact",
+      title: language.t("command.session.compact"),
+      description: language.t("command.session.compact.description"),
+      slash: "compact",
+      disabled: !params.id || visibleUserMessages().length === 0,
+      onSelect: compact,
+    }),
+    sessionCommand({
+      id: "session.fork",
+      title: language.t("command.session.fork"),
+      description: language.t("command.session.fork.description"),
+      slash: "fork",
+      disabled: !params.id || visibleUserMessages().length === 0,
+      onSelect: fork,
+    }),
+  ]
+
+  const fileCmds = () => [
+    fileCommand({
+      id: "file.open",
+      title: language.t("command.file.open"),
+      description: language.t("palette.search.placeholder"),
+      keybind: "mod+k,mod+p",
+      slash: "open",
+      onSelect: openFile,
+    }),
+    fileCommand({
+      id: "tab.close",
+      title: language.t("command.tab.close"),
+      keybind: "mod+w",
+      disabled: !closableTab(),
+      onSelect: closeTab,
+    }),
+  ]
+
+  const contextCmds = () => [
+    contextCommand({
+      id: "context.addSelection",
+      title: language.t("command.context.addSelection"),
+      description: language.t("command.context.addSelection.description"),
+      keybind: "mod+shift+l",
+      disabled: !canAddSelectionContext(),
+      onSelect: addSelection,
+    }),
+  ]
+
+  const viewCmds = () => [
+    viewCommand({
+      id: "terminal.toggle",
+      title: language.t("command.terminal.toggle"),
+      keybind: "ctrl+`",
+      slash: "terminal",
+      onSelect: () => view().terminal.toggle(),
+    }),
+    viewCommand({
+      id: "review.toggle",
+      title: language.t("command.review.toggle"),
+      keybind: "mod+shift+r",
+      onSelect: () => view().reviewPanel.toggle(),
+    }),
+    viewCommand({
+      id: "fileTree.toggle",
+      title: language.t("command.fileTree.toggle"),
+      keybind: "mod+\\",
+      onSelect: () => layout.fileTree.toggle(),
+    }),
+    viewCommand({
+      id: "input.focus",
+      title: language.t("command.input.focus"),
+      keybind: "ctrl+l",
+      onSelect: focusInput,
+    }),
+  ]
+
+  const terminalCmds = () => [
+    terminalCommand({
+      id: "terminal.new",
+      title: language.t("command.terminal.new"),
+      description: language.t("command.terminal.new.description"),
+      keybind: "ctrl+alt+t",
+      onSelect: openTerminal,
+    }),
+  ]
+
+  const messageCmds = () => [
+    sessionCommand({
+      id: "message.previous",
+      title: language.t("command.message.previous"),
+      description: language.t("command.message.previous.description"),
+      keybind: "mod+alt+[",
+      disabled: !params.id,
+      onSelect: () => navigateMessageByOffset(-1),
+    }),
+    sessionCommand({
+      id: "message.next",
+      title: language.t("command.message.next"),
+      description: language.t("command.message.next.description"),
+      keybind: "mod+alt+]",
+      disabled: !params.id,
+      onSelect: () => navigateMessageByOffset(1),
+    }),
+  ]
+
+  const modelCmds = () => [
+    modelCommand({
+      id: "model.choose",
+      title: language.t("command.model.choose"),
+      description: language.t("command.model.choose.description"),
+      keybind: "mod+'",
+      slash: "model",
+      onSelect: chooseModel,
+    }),
+    modelCommand({
+      id: "model.variant.cycle",
+      title: language.t("command.model.variant.cycle"),
+      description: language.t("command.model.variant.cycle.description"),
+      keybind: "shift+mod+d",
+      onSelect: () => local.model.variant.cycle(),
+    }),
+  ]
+
+  const mcpCmds = () => [
+    mcpCommand({
+      id: "mcp.toggle",
+      title: language.t("command.mcp.toggle"),
+      description: language.t("command.mcp.toggle.description"),
+      keybind: "mod+;",
+      slash: "mcp",
+      onSelect: chooseMcp,
+    }),
+  ]
+
+  const agentCmds = () => [
+    agentCommand({
+      id: "agent.cycle",
+      title: language.t("command.agent.cycle"),
+      description: language.t("command.agent.cycle.description"),
+      keybind: "mod+.",
+      slash: "agent",
+      onSelect: () => local.agent.move(1),
+    }),
+    agentCommand({
+      id: "agent.cycle.reverse",
+      title: language.t("command.agent.cycle.reverse"),
+      description: language.t("command.agent.cycle.reverse.description"),
+      keybind: "shift+mod+.",
+      onSelect: () => local.agent.move(-1),
+    }),
+  ]
+
+  const permissionsCmds = () => [
+    permissionsCommand({
+      id: "permissions.autoaccept",
+      title: isAutoAcceptActive()
+        ? language.t("command.permissions.autoaccept.disable")
+        : language.t("command.permissions.autoaccept.enable"),
+      keybind: "mod+shift+a",
+      disabled: false,
+      onSelect: toggleAutoAccept,
+    }),
+  ]
+
+  command.register("session", () => [
+    ...sessionCmds(),
+    ...shareCmds(),
+    ...fileCmds(),
+    ...contextCmds(),
+    ...viewCmds(),
+    ...terminalCmds(),
+    ...messageCmds(),
+    ...modelCmds(),
+    ...mcpCmds(),
+    ...agentCmds(),
+    ...permissionsCmds(),
+  ])
 }

packages/console/app/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@opencode-ai/console-app",
-  "version": "1.3.7",
+  "version": "1.3.10",
   "type": "module",
   "license": "MIT",
   "scripts": {

packages/console/app/src/routes/zen/util/handler.ts

@@ -139,19 +139,16 @@ export async function handler(
       const startTimestamp = Date.now()
       const reqUrl = providerInfo.modifyUrl(providerInfo.api, isStream)
       const reqBody = JSON.stringify(
-        providerInfo.modifyBody(
-          {
-            ...createBodyConverter(opts.format, providerInfo.format)(body),
-            model: providerInfo.model,
-            ...(providerInfo.payloadModifier ?? {}),
-            ...Object.fromEntries(
-              Object.entries(providerInfo.payloadMappings ?? {})
-                .map(([k, v]) => [k, input.request.headers.get(v)])
-                .filter(([_k, v]) => !!v),
-            ),
-          },
-          authInfo?.workspaceID,
-        ),
+        providerInfo.modifyBody({
+          ...createBodyConverter(opts.format, providerInfo.format)(body),
+          model: providerInfo.model,
+          ...(providerInfo.payloadModifier ?? {}),
+          ...Object.fromEntries(
+            Object.entries(providerInfo.payloadMappings ?? {})
+              .map(([k, v]) => [k, input.request.headers.get(v)])
+              .filter(([_k, v]) => !!v),
+          ),
+        }),
       )
       logger.debug("REQUEST URL: " + reqUrl)
       logger.debug("REQUEST: " + reqBody.substring(0, 300) + "...")
@@ -470,15 +467,17 @@ export async function handler(
       ...(() => {
         const providerProps = zenData.providers[modelProvider.id]
         const format = providerProps.format
-        const providerModel = modelProvider.model
-        if (format === "anthropic") return anthropicHelper({ reqModel, providerModel })
-        if (format === "google") return googleHelper({ reqModel, providerModel })
-        if (format === "openai") return openaiHelper({ reqModel, providerModel })
-        return oaCompatHelper({
+        const opts = {
           reqModel,
-          providerModel,
+          providerModel: modelProvider.model,
           adjustCacheUsage: providerProps.adjustCacheUsage,
-        })
+          safetyIdentifier: modelProvider.safetyIdentifier ? ip : undefined,
+          workspaceID: authInfo?.workspaceID,
+        }
+        if (format === "anthropic") return anthropicHelper(opts)
+        if (format === "google") return googleHelper(opts)
+        if (format === "openai") return openaiHelper(opts)
+        return oaCompatHelper(opts)
       })(),
     }
   }

packages/console/app/src/routes/zen/util/provider/openai-compatible.ts

@@ -21,17 +21,18 @@ type Usage = {
   }
 }
 
-export const oaCompatHelper: ProviderHelper = ({ adjustCacheUsage }) => ({
+export const oaCompatHelper: ProviderHelper = ({ adjustCacheUsage, safetyIdentifier }) => ({
   format: "oa-compat",
   modifyUrl: (providerApi: string) => providerApi + "/chat/completions",
   modifyHeaders: (headers: Headers, body: Record<string, any>, apiKey: string) => {
     headers.set("authorization", `Bearer ${apiKey}`)
     headers.set("x-session-affinity", headers.get("x-opencode-session") ?? "")
   },
-  modifyBody: (body: Record<string, any>) => {
+  modifyBody: (body: Record<string, any>, workspaceID?: string) => {
     return {
       ...body,
       ...(body.stream ? { stream_options: { include_usage: true } } : {}),
+      ...(safetyIdentifier ? { safety_identifier: safetyIdentifier } : {}),
     }
   },
   createBinaryStreamDecoder: () => undefined,

packages/console/app/src/routes/zen/util/provider/openai.ts

@@ -12,13 +12,13 @@ type Usage = {
   total_tokens?: number
 }
 
-export const openaiHelper: ProviderHelper = () => ({
+export const openaiHelper: ProviderHelper = ({ workspaceID }) => ({
   format: "openai",
   modifyUrl: (providerApi: string) => providerApi + "/responses",
   modifyHeaders: (headers: Headers, body: Record<string, any>, apiKey: string) => {
     headers.set("authorization", `Bearer ${apiKey}`)
   },
-  modifyBody: (body: Record<string, any>, workspaceID?: string) => ({
+  modifyBody: (body: Record<string, any>) => ({
     ...body,
     ...(workspaceID ? { safety_identifier: workspaceID } : {}),
   }),

packages/console/app/src/routes/zen/util/provider/provider.ts

@@ -33,11 +33,17 @@ export type UsageInfo = {
   cacheWrite1hTokens?: number
 }
 
-export type ProviderHelper = (input: { reqModel: string; providerModel: string; adjustCacheUsage?: boolean }) => {
+export type ProviderHelper = (input: {
+  reqModel: string
+  providerModel: string
+  adjustCacheUsage?: boolean
+  safetyIdentifier?: string
+  workspaceID?: string
+}) => {
   format: ZenData.Format
   modifyUrl: (providerApi: string, isStream?: boolean) => string
   modifyHeaders: (headers: Headers, body: Record<string, any>, apiKey: string) => void
-  modifyBody: (body: Record<string, any>, workspaceID?: string) => Record<string, any>
+  modifyBody: (body: Record<string, any>) => Record<string, any>
   createBinaryStreamDecoder: () => ((chunk: Uint8Array) => Uint8Array | undefined) | undefined
   streamSeparator: string
   createUsageParser: () => {

packages/console/core/package.json

@@ -1,7 +1,7 @@
 {
   "$schema": "https://json.schemastore.org/package.json",
   "name": "@opencode-ai/console-core",
-  "version": "1.3.7",
+  "version": "1.3.10",
   "private": true,
   "type": "module",
   "license": "MIT",

packages/console/core/script/promote-limits.ts

@@ -10,7 +10,7 @@ if (!stage) throw new Error("Stage is required")
 const root = path.resolve(process.cwd(), "..", "..", "..")
 
 // read the secret
-const ret = await $`bun sst secret list`.cwd(root).text()
+const ret = await $`bun sst secret list --stage frank`.cwd(root).text()
 const lines = ret.split("\n")
 const value = lines.find((line) => line.startsWith("ZEN_LIMITS"))?.split("=")[1]
 if (!value) throw new Error("ZEN_LIMITS not found")

packages/console/core/script/promote-models.ts

@@ -12,7 +12,7 @@ const root = path.resolve(process.cwd(), "..", "..", "..")
 const PARTS = 30
 
 // read the secret
-const ret = await $`bun sst secret list`.cwd(root).text()
+const ret = await $`bun sst secret list --stage frank`.cwd(root).text()
 const lines = ret.split("\n")
 const values = Array.from({ length: PARTS }, (_, i) => {
   const value = lines

packages/console/core/script/update-limits.ts

@@ -6,7 +6,7 @@ import os from "os"
 import { Subscription } from "../src/subscription"
 
 const root = path.resolve(process.cwd(), "..", "..", "..")
-const secrets = await $`bun sst secret list`.cwd(root).text()
+const secrets = await $`bun sst secret list --stage frank`.cwd(root).text()
 
 // read value
 const lines = secrets.split("\n")
@@ -25,4 +25,4 @@ const newValue = JSON.stringify(JSON.parse(await tempFile.text()))
 Subscription.validate(JSON.parse(newValue))
 
 // update the secret
-await $`bun sst secret set ZEN_LIMITS ${newValue}`
+await $`bun sst secret set ZEN_LIMITS ${newValue} --stage frank`.cwd(root)

packages/console/core/script/update-models.ts

@@ -6,7 +6,7 @@ import os from "os"
 import { ZenData } from "../src/model"
 
 const root = path.resolve(process.cwd(), "..", "..", "..")
-const models = await $`bun sst secret list`.cwd(root).text()
+const models = await $`bun sst secret list --stage frank`.cwd(root).text()
 const PARTS = 30
 
 // read the line starting with "ZEN_MODELS"
@@ -40,4 +40,4 @@ const newValues = Array.from({ length: PARTS }, (_, i) =>
 
 const envFile = Bun.file(path.join(os.tmpdir(), `models-${Date.now()}.env`))
 await envFile.write(newValues.map((v, i) => `ZEN_MODELS${i + 1}="${v.replace(/"/g, '\\"')}"`).join("\n"))
-await $`bun sst secret load ${envFile.name}`.cwd(root)
+await $`bun sst secret load ${envFile.name} --stage frank`.cwd(root)

packages/console/core/src/model.ts

@@ -37,6 +37,7 @@ export namespace ZenData {
         disabled: z.boolean().optional(),
         storeModel: z.string().optional(),
         payloadModifier: z.record(z.string(), z.any()).optional(),
+        safetyIdentifier: z.boolean().optional(),
       }),
     ),
   })

packages/console/function/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@opencode-ai/console-function",
-  "version": "1.3.7",
+  "version": "1.3.10",
   "$schema": "https://json.schemastore.org/package.json",
   "private": true,
   "type": "module",

packages/console/mail/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@opencode-ai/console-mail",
-  "version": "1.3.7",
+  "version": "1.3.10",
   "dependencies": {
     "@jsx-email/all": "2.2.3",
     "@jsx-email/cli": "1.4.3",

packages/desktop-electron/package.json

@@ -1,7 +1,7 @@
 {
   "name": "@opencode-ai/desktop-electron",
   "private": true,
-  "version": "1.3.7",
+  "version": "1.3.10",
   "type": "module",
   "license": "MIT",
   "homepage": "https://opencode.ai",

packages/desktop/package.json

@@ -1,7 +1,7 @@
 {
   "name": "@opencode-ai/desktop",
   "private": true,
-  "version": "1.3.7",
+  "version": "1.3.10",
   "type": "module",
   "license": "MIT",
   "scripts": {

packages/enterprise/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@opencode-ai/enterprise",
-  "version": "1.3.7",
+  "version": "1.3.10",
   "private": true,
   "type": "module",
   "license": "MIT",

packages/extensions/zed/extension.toml

@@ -1,7 +1,7 @@
 id = "opencode"
 name = "OpenCode"
 description = "The open source coding agent."
-version = "1.3.7"
+version = "1.3.10"
 schema_version = 1
 authors = ["Anomaly"]
 repository = "https://github.com/anomalyco/opencode"
@@ -11,26 +11,26 @@ name = "OpenCode"
 icon = "./icons/opencode.svg"
 
 [agent_servers.opencode.targets.darwin-aarch64]
-archive = "https://github.com/anomalyco/opencode/releases/download/v1.3.7/opencode-darwin-arm64.zip"
+archive = "https://github.com/anomalyco/opencode/releases/download/v1.3.10/opencode-darwin-arm64.zip"
 cmd = "./opencode"
 args = ["acp"]
 
 [agent_servers.opencode.targets.darwin-x86_64]
-archive = "https://github.com/anomalyco/opencode/releases/download/v1.3.7/opencode-darwin-x64.zip"
+archive = "https://github.com/anomalyco/opencode/releases/download/v1.3.10/opencode-darwin-x64.zip"
 cmd = "./opencode"
 args = ["acp"]
 
 [agent_servers.opencode.targets.linux-aarch64]
-archive = "https://github.com/anomalyco/opencode/releases/download/v1.3.7/opencode-linux-arm64.tar.gz"
+archive = "https://github.com/anomalyco/opencode/releases/download/v1.3.10/opencode-linux-arm64.tar.gz"
 cmd = "./opencode"
 args = ["acp"]
 
 [agent_servers.opencode.targets.linux-x86_64]
-archive = "https://github.com/anomalyco/opencode/releases/download/v1.3.7/opencode-linux-x64.tar.gz"
+archive = "https://github.com/anomalyco/opencode/releases/download/v1.3.10/opencode-linux-x64.tar.gz"
 cmd = "./opencode"
 args = ["acp"]
 
 [agent_servers.opencode.targets.windows-x86_64]
-archive = "https://github.com/anomalyco/opencode/releases/download/v1.3.7/opencode-windows-x64.zip"
+archive = "https://github.com/anomalyco/opencode/releases/download/v1.3.10/opencode-windows-x64.zip"
 cmd = "./opencode.exe"
 args = ["acp"]

packages/function/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@opencode-ai/function",
-  "version": "1.3.7",
+  "version": "1.3.10",
   "$schema": "https://json.schemastore.org/package.json",
   "private": true,
   "type": "module",

packages/opencode/package.json

@@ -1,6 +1,6 @@
 {
   "$schema": "https://json.schemastore.org/package.json",
-  "version": "1.3.7",
+  "version": "1.3.10",
   "name": "opencode",
   "type": "module",
   "license": "MIT",
@@ -87,7 +87,7 @@
     "@ai-sdk/provider-utils": "4.0.21",
     "@ai-sdk/togetherai": "2.0.41",
     "@ai-sdk/vercel": "2.0.39",
-    "@ai-sdk/xai": "3.0.74",
+    "@ai-sdk/xai": "3.0.75",
     "@aws-sdk/credential-providers": "3.993.0",
     "@clack/prompts": "1.0.0-alpha.1",
     "@effect/platform-node": "catalog:",

packages/opencode/specs/effect-migration.md

@@ -210,15 +210,13 @@ Fully migrated (single namespace, InstanceState where needed, flattened facade):
 - [x] `Vcs` — `project/vcs.ts`
 - [x] `Worktree` — `worktree/index.ts`
 
-Still open and likely worth migrating:
-
 - [x] `Session` — `session/index.ts`
-- [ ] `SessionProcessor` — blocked by AI SDK v6 PR (#18433)
-- [ ] `SessionPrompt` — blocked by AI SDK v6 PR (#18433)
-- [ ] `SessionCompaction` — blocked by AI SDK v6 PR (#18433)
-- [ ] `Provider` — blocked by AI SDK v6 PR (#18433)
+- [x] `SessionProcessor` — `session/processor.ts`
+- [x] `SessionPrompt` — `session/prompt.ts`
+- [x] `SessionCompaction` — `session/compaction.ts`
+- [x] `Provider` — `provider/provider.ts`
 
-Other services not yet migrated:
+Still open:
 
 - [ ] `SessionSummary` — `session/summary.ts`
 - [ ] `SessionTodo` — `session/todo.ts`
@@ -235,7 +233,7 @@ Once individual tools are effectified, change `Tool.Info` (`tool/tool.ts`) so `i
 
 1. Migrate each tool to return Effects
 2. Update `Tool.define()` factory to work with Effects
-3. Update `SessionPrompt` to `yield*` tool results instead of `await`ing — blocked by AI SDK v6 PR (#18433)
+3. Update `SessionPrompt` to `yield*` tool results instead of `await`ing
 
 Individual tools, ordered by value:
 

packages/opencode/specs/tui-plugins.md

@@ -84,12 +84,18 @@ export default plugin
 - TUI shape is `default export { id?, tui }`; including `server` is rejected.
 - A single module cannot export both `server` and `tui`.
 - `tui` signature is `(api, options, meta) => Promise<void>`.
-- If package `exports` contains `./tui`, the loader resolves that entrypoint. Otherwise it uses the resolved package target.
+- If package `exports` contains `./tui`, the loader resolves that entrypoint.
+- If package `exports` exists, loader only resolves `./tui` or `./server`; it never falls back to `exports["."]`.
+- For npm package specs, TUI does not use `package.json` `main` as a fallback entry.
+- `package.json` `main` is only used for server plugin entrypoint resolution.
 - If a package supports both server and TUI, use separate files and package `exports` (`./server` and `./tui`) so each target resolves to a target-only module.
 - File/path plugins must export a non-empty `id`.
 - npm plugins may omit `id`; package `name` is used.
 - Runtime identity is the resolved plugin id. Later plugins with the same id are rejected, including collisions with internal plugin ids.
-- If a path spec points at a directory, that directory must have `package.json` with `main`.
+- If a path spec points at a directory, server loading can use `package.json` `main`.
+- TUI path loading never uses `package.json` `main`.
+- Legacy compatibility: path specs like `./plugin` can resolve to `./plugin/index.ts` (or `index.js`) when `package.json` is missing.
+- The `./plugin -> ./plugin/index.*` fallback applies to both server and TUI v1 loading.
 - There is no directory auto-discovery for TUI plugins; they must be listed in `tui.json`.
 
 ## Package manifest and install
@@ -269,7 +275,9 @@ Theme install behavior:
 
 - Relative theme paths are resolved from the plugin root.
 - Theme name is the JSON basename.
-- Install is skipped if that theme name already exists.
+- First install writes only when the destination file is missing.
+- If the theme name already exists, install is skipped unless plugin metadata state is `updated`.
+- On `updated`, host only rewrites themes previously tracked for that plugin and only when source `mtime`/`size` changed.
 - Local plugins persist installed themes under the local `.opencode/themes` area near the plugin config source.
 - Global plugins persist installed themes under the global `themes` dir.
 - Invalid or unreadable theme files are ignored.

packages/opencode/src/agent/agent.ts

@@ -393,7 +393,7 @@ export namespace Agent {
   )
 
   export const defaultLayer = layer.pipe(
-    Layer.provide(Auth.layer),
+    Layer.provide(Auth.defaultLayer),
     Layer.provide(Config.defaultLayer),
     Layer.provide(Skill.defaultLayer),
   )

packages/opencode/src/auth/index.ts

@@ -3,7 +3,7 @@ import { Effect, Layer, Record, Result, Schema, ServiceMap } from "effect"
 import { makeRuntime } from "@/effect/run-service"
 import { zod } from "@/util/effect-zod"
 import { Global } from "../global"
-import { Filesystem } from "../util/filesystem"
+import { AppFileSystem } from "../filesystem"
 
 export const OAUTH_DUMMY_KEY = "opencode-oauth-dummy-key"
 
@@ -53,17 +53,13 @@ export namespace Auth {
   export const layer = Layer.effect(
     Service,
     Effect.gen(function* () {
+      const fsys = yield* AppFileSystem.Service
       const decode = Schema.decodeUnknownOption(Info)
 
-      const all = Effect.fn("Auth.all")(() =>
-        Effect.tryPromise({
-          try: async () => {
-            const data = await Filesystem.readJson<Record<string, unknown>>(file).catch(() => ({}))
-            return Record.filterMap(data, (value) => Result.fromOption(decode(value), () => undefined))
-          },
-          catch: fail("Failed to read auth data"),
-        }),
-      )
+      const all = Effect.fn("Auth.all")(function* () {
+        const data = (yield* fsys.readJson(file).pipe(Effect.orElseSucceed(() => ({})))) as Record<string, unknown>
+        return Record.filterMap(data, (value) => Result.fromOption(decode(value), () => undefined))
+      })
 
       const get = Effect.fn("Auth.get")(function* (providerID: string) {
         return (yield* all())[providerID]
@@ -74,10 +70,9 @@ export namespace Auth {
         const data = yield* all()
         if (norm !== key) delete data[key]
         delete data[norm + "/"]
-        yield* Effect.tryPromise({
-          try: () => Filesystem.writeJson(file, { ...data, [norm]: info }, 0o600),
-          catch: fail("Failed to write auth data"),
-        })
+        yield* fsys
+          .writeJson(file, { ...data, [norm]: info }, 0o600)
+          .pipe(Effect.mapError(fail("Failed to write auth data")))
       })
 
       const remove = Effect.fn("Auth.remove")(function* (key: string) {
@@ -85,17 +80,16 @@ export namespace Auth {
         const data = yield* all()
         delete data[key]
         delete data[norm]
-        yield* Effect.tryPromise({
-          try: () => Filesystem.writeJson(file, data, 0o600),
-          catch: fail("Failed to write auth data"),
-        })
+        yield* fsys.writeJson(file, data, 0o600).pipe(Effect.mapError(fail("Failed to write auth data")))
       })
 
       return Service.of({ get, all, set, remove })
     }),
   )
 
-  const { runPromise } = makeRuntime(Service, layer)
+  export const defaultLayer = layer.pipe(Layer.provide(AppFileSystem.defaultLayer))
+
+  const { runPromise } = makeRuntime(Service, defaultLayer)
 
   export async function get(providerID: string) {
     return runPromise((service) => service.get(providerID))

packages/opencode/src/cli/cmd/tui/context/theme.tsx

@@ -183,6 +183,18 @@ export function addTheme(name: string, theme: unknown) {
   return true
 }
 
+export function upsertTheme(name: string, theme: unknown) {
+  if (!name) return false
+  if (!isTheme(theme)) return false
+  if (customThemes[name] !== undefined) {
+    customThemes[name] = theme
+  } else {
+    pluginThemes[name] = theme
+  }
+  syncThemes()
+  return true
+}
+
 export function resolveTheme(theme: ThemeJson, mode: "dark" | "light") {
   const defs = theme.defs ?? {}
   function resolveColor(c: ColorValue, chain: string[] = []): RGBA {

packages/opencode/src/cli/cmd/tui/context/theme/catppuccin-frappe.json

@@ -62,8 +62,8 @@
       "light": "frappeText"
     },
     "textMuted": {
-      "dark": "frappeSubtext1",
-      "light": "frappeSubtext1"
+      "dark": "frappeOverlay2",
+      "light": "frappeOverlay2"
     },
     "background": {
       "dark": "frappeBase",

packages/opencode/src/cli/cmd/tui/context/theme/catppuccin-macchiato.json

@@ -62,8 +62,8 @@
       "light": "macText"
     },
     "textMuted": {
-      "dark": "macSubtext1",
-      "light": "macSubtext1"
+      "dark": "macOverlay2",
+      "light": "macOverlay2"
     },
     "background": {
       "dark": "macBase",

packages/opencode/src/cli/cmd/tui/context/theme/catppuccin.json

@@ -63,7 +63,7 @@
     "success": { "dark": "darkGreen", "light": "lightGreen" },
     "info": { "dark": "darkTeal", "light": "lightTeal" },
     "text": { "dark": "darkText", "light": "lightText" },
-    "textMuted": { "dark": "darkSubtext1", "light": "lightSubtext1" },
+    "textMuted": { "dark": "darkOverlay2", "light": "lightOverlay2" },
     "background": { "dark": "darkBase", "light": "lightBase" },
     "backgroundPanel": { "dark": "darkMantle", "light": "lightMantle" },
     "backgroundElement": { "dark": "darkCrust", "light": "lightCrust" },

packages/opencode/src/cli/cmd/tui/feature-plugins/home/footer.tsx

@@ -0,0 +1,93 @@
+import type { TuiPlugin, TuiPluginApi, TuiPluginModule } from "@opencode-ai/plugin/tui"
+import { createMemo, Match, Show, Switch } from "solid-js"
+import { Global } from "@/global"
+
+const id = "internal:home-footer"
+
+function Directory(props: { api: TuiPluginApi }) {
+  const theme = () => props.api.theme.current
+  const dir = createMemo(() => {
+    const dir = props.api.state.path.directory || process.cwd()
+    const out = dir.replace(Global.Path.home, "~")
+    const branch = props.api.state.vcs?.branch
+    if (branch) return out + ":" + branch
+    return out
+  })
+
+  return <text fg={theme().textMuted}>{dir()}</text>
+}
+
+function Mcp(props: { api: TuiPluginApi }) {
+  const theme = () => props.api.theme.current
+  const list = createMemo(() => props.api.state.mcp())
+  const has = createMemo(() => list().length > 0)
+  const err = createMemo(() => list().some((item) => item.status === "failed"))
+  const count = createMemo(() => list().filter((item) => item.status === "connected").length)
+
+  return (
+    <Show when={has()}>
+      <box gap={1} flexDirection="row" flexShrink={0}>
+        <text fg={theme().text}>
+          <Switch>
+            <Match when={err()}>
+              <span style={{ fg: theme().error }}>⊙ </span>
+            </Match>
+            <Match when={true}>
+              <span style={{ fg: count() > 0 ? theme().success : theme().textMuted }}>⊙ </span>
+            </Match>
+          </Switch>
+          {count()} MCP
+        </text>
+        <text fg={theme().textMuted}>/status</text>
+      </box>
+    </Show>
+  )
+}
+
+function Version(props: { api: TuiPluginApi }) {
+  const theme = () => props.api.theme.current
+
+  return (
+    <box flexShrink={0}>
+      <text fg={theme().textMuted}>{props.api.app.version}</text>
+    </box>
+  )
+}
+
+function View(props: { api: TuiPluginApi }) {
+  return (
+    <box
+      width="100%"
+      paddingTop={1}
+      paddingBottom={1}
+      paddingLeft={2}
+      paddingRight={2}
+      flexDirection="row"
+      flexShrink={0}
+      gap={2}
+    >
+      <Directory api={props.api} />
+      <Mcp api={props.api} />
+      <box flexGrow={1} />
+      <Version api={props.api} />
+    </box>
+  )
+}
+
+const tui: TuiPlugin = async (api) => {
+  api.slots.register({
+    order: 100,
+    slots: {
+      home_footer() {
+        return <View api={api} />
+      },
+    },
+  })
+}
+
+const plugin: TuiPluginModule & { id: string } = {
+  id,
+  tui,
+}
+
+export default plugin

packages/opencode/src/cli/cmd/tui/plugin/internal.ts

@@ -1,3 +1,4 @@
+import HomeFooter from "../feature-plugins/home/footer"
 import HomeTips from "../feature-plugins/home/tips"
 import SidebarContext from "../feature-plugins/sidebar/context"
 import SidebarMcp from "../feature-plugins/sidebar/mcp"
@@ -14,6 +15,7 @@ export type InternalTuiPlugin = TuiPluginModule & {
 }
 
 export const INTERNAL_TUI_PLUGINS: InternalTuiPlugin[] = [
+  HomeFooter,
   HomeTips,
   SidebarContext,
   SidebarMcp,

packages/opencode/src/cli/cmd/tui/plugin/runtime.ts

@@ -18,38 +18,29 @@ import { Log } from "@/util/log"
 import { errorData, errorMessage } from "@/util/error"
 import { isRecord } from "@/util/record"
 import { Instance } from "@/project/instance"
-import {
-  checkPluginCompatibility,
-  isDeprecatedPlugin,
-  pluginSource,
-  readPluginId,
-  readV1Plugin,
-  resolvePluginEntrypoint,
-  resolvePluginId,
-  resolvePluginTarget,
-  type PluginSource,
-} from "@/plugin/shared"
+import { pluginSource, readPluginId, readV1Plugin, resolvePluginId, type PluginSource } from "@/plugin/shared"
+import { PluginLoader } from "@/plugin/loader"
 import { PluginMeta } from "@/plugin/meta"
 import { installPlugin as installModulePlugin, patchPluginConfig, readPluginManifest } from "@/plugin/install"
-import { addTheme, hasTheme } from "../context/theme"
+import { hasTheme, upsertTheme } from "../context/theme"
 import { Global } from "@/global"
 import { Filesystem } from "@/util/filesystem"
 import { Process } from "@/util/process"
 import { Flag } from "@/flag/flag"
-import { Installation } from "@/installation"
 import { INTERNAL_TUI_PLUGINS, type InternalTuiPlugin } from "./internal"
 import { setupSlots, Slot as View } from "./slots"
 import type { HostPluginApi, HostSlots } from "./slots"
 
 type PluginLoad = {
-  item?: Config.PluginSpec
+  options: Config.PluginOptions | undefined
   spec: string
   target: string
   retry: boolean
   source: PluginSource | "internal"
   id: string
   module: TuiPluginModule
-  install_theme: TuiTheme["install"]
+  theme_meta: TuiConfig.PluginMeta
+  theme_root: string
 }
 
 type Api = HostPluginApi
@@ -64,8 +55,8 @@ type PluginEntry = {
   id: string
   load: PluginLoad
   meta: TuiPluginMeta
+  themes: Record<string, PluginMeta.Theme>
   plugin: TuiPlugin
-  options: Config.PluginOptions | undefined
   enabled: boolean
   scope?: PluginScope
 }
@@ -76,13 +67,7 @@ type RuntimeState = {
   slots: HostSlots
   plugins: PluginEntry[]
   plugins_by_id: Map<string, PluginEntry>
-  pending: Map<
-    string,
-    {
-      item: Config.PluginSpec
-      meta: TuiConfig.PluginMeta
-    }
-  >
+  pending: Map<string, TuiConfig.PluginRecord>
 }
 
 const log = Log.create({ service: "tui.plugin" })
@@ -143,12 +128,54 @@ function resolveRoot(root: string) {
   return path.resolve(process.cwd(), root)
 }
 
-function createThemeInstaller(meta: TuiConfig.PluginMeta, root: string, spec: string): TuiTheme["install"] {
+function createThemeInstaller(
+  meta: TuiConfig.PluginMeta,
+  root: string,
+  spec: string,
+  plugin: PluginEntry,
+): TuiTheme["install"] {
   return async (file) => {
     const raw = file.startsWith("file://") ? fileURLToPath(file) : file
     const src = path.isAbsolute(raw) ? raw : path.resolve(root, raw)
-    const theme = path.basename(src, path.extname(src))
-    if (hasTheme(theme)) return
+    const name = path.basename(src, path.extname(src))
+    const source_dir = path.dirname(meta.source)
+    const local_dir =
+      path.basename(source_dir) === ".opencode"
+        ? path.join(source_dir, "themes")
+        : path.join(source_dir, ".opencode", "themes")
+    const dest_dir = meta.scope === "local" ? local_dir : path.join(Global.Path.config, "themes")
+    const dest = path.join(dest_dir, `${name}.json`)
+    const stat = await Filesystem.statAsync(src)
+    const mtime = stat ? Math.floor(typeof stat.mtimeMs === "bigint" ? Number(stat.mtimeMs) : stat.mtimeMs) : undefined
+    const size = stat ? (typeof stat.size === "bigint" ? Number(stat.size) : stat.size) : undefined
+    const exists = hasTheme(name)
+    const prev = plugin.themes[name]
+
+    if (exists) {
+      if (plugin.meta.state !== "updated") return
+      if (!prev) {
+        if (await Filesystem.exists(dest)) {
+          plugin.themes[name] = {
+            src,
+            dest,
+            mtime,
+            size,
+          }
+          await PluginMeta.setTheme(plugin.id, name, plugin.themes[name]!).catch((error) => {
+            log.warn("failed to track tui plugin theme", {
+              path: spec,
+              id: plugin.id,
+              theme: src,
+              dest,
+              error,
+            })
+          })
+        }
+        return
+      }
+      if (prev.dest !== dest) return
+      if (prev.mtime === mtime && prev.size === size) return
+    }
 
     const text = await Filesystem.readText(src).catch((error) => {
       log.warn("failed to read tui plugin theme", { path: spec, theme: src, error })
@@ -170,90 +197,101 @@ function createThemeInstaller(meta: TuiConfig.PluginMeta, root: string, spec: st
       return
     }
 
-    const source_dir = path.dirname(meta.source)
-    const local_dir =
-      path.basename(source_dir) === ".opencode"
-        ? path.join(source_dir, "themes")
-        : path.join(source_dir, ".opencode", "themes")
-    const dest_dir = meta.scope === "local" ? local_dir : path.join(Global.Path.config, "themes")
-    const dest = path.join(dest_dir, `${theme}.json`)
-    if (!(await Filesystem.exists(dest))) {
+    if (exists || !(await Filesystem.exists(dest))) {
       await Filesystem.write(dest, text).catch((error) => {
         log.warn("failed to persist tui plugin theme", { path: spec, theme: src, dest, error })
       })
     }
 
-    addTheme(theme, data)
+    upsertTheme(name, data)
+    plugin.themes[name] = {
+      src,
+      dest,
+      mtime,
+      size,
+    }
+    await PluginMeta.setTheme(plugin.id, name, plugin.themes[name]!).catch((error) => {
+      log.warn("failed to track tui plugin theme", {
+        path: spec,
+        id: plugin.id,
+        theme: src,
+        dest,
+        error,
+      })
+    })
   }
 }
 
-async function loadExternalPlugin(
-  item: Config.PluginSpec,
-  meta: TuiConfig.PluginMeta | undefined,
-  retry = false,
-): Promise<PluginLoad | undefined> {
-  const spec = Config.pluginSpecifier(item)
-  if (isDeprecatedPlugin(spec)) return
-  log.info("loading tui plugin", { path: spec, retry })
-  const resolved = await resolvePluginTarget(spec).catch((error) => {
-    fail("failed to resolve tui plugin", { path: spec, retry, error })
-    return
-  })
-  if (!resolved) return
+async function loadExternalPlugin(cfg: TuiConfig.PluginRecord, retry = false): Promise<PluginLoad | undefined> {
+  const plan = PluginLoader.plan(cfg.item)
+  if (plan.deprecated) return
 
-  const source = pluginSource(spec)
-  if (source === "npm") {
-    const ok = await checkPluginCompatibility(resolved, Installation.VERSION)
-      .then(() => true)
-      .catch((error) => {
-        fail("tui plugin incompatible", { path: spec, retry, error })
-        return false
-      })
-    if (!ok) return
+  log.info("loading tui plugin", { path: plan.spec, retry })
+  const resolved = await PluginLoader.resolve(plan, "tui")
+  if (!resolved.ok) {
+    if (resolved.stage === "install") {
+      fail("failed to resolve tui plugin", { path: plan.spec, retry, error: resolved.error })
+      return
+    }
+    if (resolved.stage === "compatibility") {
+      fail("tui plugin incompatible", { path: plan.spec, retry, error: resolved.error })
+      return
+    }
+    fail("failed to resolve tui plugin entry", { path: plan.spec, retry, error: resolved.error })
+    return
   }
 
-  const target = resolved
-  if (!meta) {
-    fail("missing tui plugin metadata", {
-      path: spec,
+  const loaded = await PluginLoader.load(resolved.value)
+  if (!loaded.ok) {
+    fail("failed to load tui plugin", {
+      path: plan.spec,
+      target: resolved.value.entry,
       retry,
+      error: loaded.error,
     })
     return
   }
 
-  const root = resolveRoot(source === "file" ? spec : target)
-  const install_theme = createThemeInstaller(meta, root, spec)
-  const entry = await resolvePluginEntrypoint(spec, target, "tui").catch((error) => {
-    fail("failed to resolve tui plugin entry", { path: spec, target, retry, error })
-    return
-  })
-  if (!entry) return
-
-  const mod = await import(entry)
-    .then((raw) => {
-      return readV1Plugin(raw as Record<string, unknown>, spec, "tui") as TuiPluginModule
+  const mod = await Promise.resolve()
+    .then(() => {
+      return readV1Plugin(loaded.value.mod as Record<string, unknown>, plan.spec, "tui") as TuiPluginModule
     })
     .catch((error) => {
-      fail("failed to load tui plugin", { path: spec, target: entry, retry, error })
+      fail("failed to load tui plugin", {
+        path: plan.spec,
+        target: loaded.value.entry,
+        retry,
+        error,
+      })
       return
     })
   if (!mod) return
 
-  const id = await resolvePluginId(source, spec, target, readPluginId(mod.id, spec)).catch((error) => {
-    fail("failed to load tui plugin", { path: spec, target, retry, error })
+  const id = await resolvePluginId(
+    loaded.value.source,
+    plan.spec,
+    loaded.value.target,
+    readPluginId(mod.id, plan.spec),
+    loaded.value.pkg,
+  ).catch((error) => {
+    fail("failed to load tui plugin", { path: plan.spec, target: loaded.value.target, retry, error })
     return
   })
   if (!id) return
 
   return {
-    item,
-    spec,
-    target,
+    options: plan.options,
+    spec: plan.spec,
+    target: loaded.value.target,
     retry,
-    source,
+    source: loaded.value.source,
     id,
     module: mod,
-    install_theme,
+    theme_meta: {
+      scope: cfg.scope,
+      source: cfg.source,
+    },
+    theme_root: loaded.value.pkg?.dir ?? resolveRoot(loaded.value.target),
   }
 }
 
@@ -291,20 +329,18 @@ function loadInternalPlugin(item: InternalTuiPlugin): PluginLoad {
   const target = spec
 
   return {
+    options: undefined,
     spec,
     target,
     retry: false,
     source: "internal",
     id: item.id,
     module: item,
-    install_theme: createThemeInstaller(
-      {
-        scope: "global",
-        source: target,
-      },
-      process.cwd(),
-      spec,
-    ),
+    theme_meta: {
+      scope: "global",
+      source: target,
+    },
+    theme_root: process.cwd(),
   }
 }
 
@@ -436,10 +472,10 @@ async function activatePluginEntry(state: RuntimeState, plugin: PluginEntry, per
   if (plugin.scope) return true
 
   const scope = createPluginScope(plugin.load, plugin.id)
-  const api = pluginApi(state, plugin.load, scope, plugin.id)
+  const api = pluginApi(state, plugin, scope, plugin.id)
   const ok = await Promise.resolve()
     .then(async () => {
-      await plugin.plugin(api, plugin.options, plugin.meta)
+      await plugin.plugin(api, plugin.load.options, plugin.meta)
       return true
     })
     .catch((error) => {
@@ -479,9 +515,10 @@ async function deactivatePluginById(state: RuntimeState | undefined, id: string,
   return deactivatePluginEntry(state, plugin, persist)
 }
 
-function pluginApi(runtime: RuntimeState, load: PluginLoad, scope: PluginScope, base: string): TuiPluginApi {
+function pluginApi(runtime: RuntimeState, plugin: PluginEntry, scope: PluginScope, base: string): TuiPluginApi {
   const api = runtime.api
   const host = runtime.slots
+  const load = plugin.load
   const command: TuiPluginApi["command"] = {
     register(cb) {
       return scope.track(api.command.register(cb))
@@ -504,7 +541,7 @@ function pluginApi(runtime: RuntimeState, load: PluginLoad, scope: PluginScope,
   }
 
   const theme: TuiPluginApi["theme"] = Object.assign(Object.create(api.theme), {
-    install: load.install_theme,
+    install: createThemeInstaller(load.theme_meta, load.theme_root, load.spec, plugin),
   })
 
   const event: TuiPluginApi["event"] = {
@@ -563,20 +600,6 @@ function pluginApi(runtime: RuntimeState, load: PluginLoad, scope: PluginScope,
   }
 }
 
-function collectPluginEntries(load: PluginLoad, meta: TuiPluginMeta) {
-  const options = load.item ? Config.pluginOptions(load.item) : undefined
-  return [
-    {
-      id: load.id,
-      load,
-      meta,
-      plugin: load.module.tui,
-      options,
-      enabled: true,
-    },
-  ]
-}
-
 function addPluginEntry(state: RuntimeState, plugin: PluginEntry) {
   if (state.plugins_by_id.has(plugin.id)) {
     fail("duplicate tui plugin id", {
@@ -600,12 +623,8 @@ function applyInitialPluginEnabledState(state: RuntimeState, config: TuiConfig.I
   }
 }
 
-async function resolveExternalPlugins(
-  list: Config.PluginSpec[],
-  wait: () => Promise<void>,
-  meta: (item: Config.PluginSpec) => TuiConfig.PluginMeta | undefined,
-) {
-  const loaded = await Promise.all(list.map((item) => loadExternalPlugin(item, meta(item))))
+async function resolveExternalPlugins(list: TuiConfig.PluginRecord[], wait: () => Promise<void>) {
+  const loaded = await Promise.all(list.map((item) => loadExternalPlugin(item)))
   const ready: PluginLoad[] = []
   let deps: Promise<void> | undefined
 
@@ -614,13 +633,12 @@ async function resolveExternalPlugins(
     if (!entry) {
       const item = list[i]
       if (!item) continue
-      const spec = Config.pluginSpecifier(item)
-      if (pluginSource(spec) !== "file") continue
+      if (pluginSource(Config.pluginSpecifier(item.item)) !== "file") continue
       deps ??= wait().catch((error) => {
         log.warn("failed waiting for tui plugin dependencies", { error })
       })
       await deps
-      entry = await loadExternalPlugin(item, meta(item), true)
+      entry = await loadExternalPlugin(item, true)
     }
     if (!entry) continue
     ready.push(entry)
@@ -661,20 +679,28 @@ async function addExternalPluginEntries(state: RuntimeState, ready: PluginLoad[]
     }
 
     const row = createMeta(entry.source, entry.spec, entry.target, hit, entry.id)
-    for (const plugin of collectPluginEntries(entry, row)) {
-      if (!addPluginEntry(state, plugin)) {
-        ok = false
-        continue
-      }
-      plugins.push(plugin)
+    const themes = hit?.entry.themes ? { ...hit.entry.themes } : {}
+    const plugin: PluginEntry = {
+      id: entry.id,
+      load: entry,
+      meta: row,
+      themes,
+      plugin: entry.module.tui,
+      enabled: true,
     }
+    if (!addPluginEntry(state, plugin)) {
+      ok = false
+      continue
+    }
+    plugins.push(plugin)
   }
 
   return { plugins, ok }
 }
 
-function defaultPluginMeta(state: RuntimeState): TuiConfig.PluginMeta {
+function defaultPluginRecord(state: RuntimeState, spec: string): TuiConfig.PluginRecord {
   return {
+    item: spec,
     scope: "local",
     source: state.api.state.path.config || path.join(state.directory, ".opencode", "tui.json"),
   }
@@ -712,36 +738,28 @@ async function addPluginBySpec(state: RuntimeState | undefined, raw: string) {
   const spec = raw.trim()
   if (!spec) return false
 
-  const pending = state.pending.get(spec)
-  const item = pending?.item ?? spec
-  const nextSpec = Config.pluginSpecifier(item)
-  if (state.plugins.some((plugin) => plugin.load.spec === nextSpec)) {
+  const cfg = state.pending.get(spec) ?? defaultPluginRecord(state, spec)
+  const next = Config.pluginSpecifier(cfg.item)
+  if (state.plugins.some((plugin) => plugin.load.spec === next)) {
     state.pending.delete(spec)
     return true
   }
 
-  const meta = pending?.meta ?? defaultPluginMeta(state)
-
   const ready = await Instance.provide({
     directory: state.directory,
-    fn: () =>
-      resolveExternalPlugins(
-        [item],
-        () => TuiConfig.waitForDependencies(),
-        () => meta,
-      ),
+    fn: () => resolveExternalPlugins([cfg], () => TuiConfig.waitForDependencies()),
   }).catch((error) => {
-    fail("failed to add tui plugin", { path: nextSpec, error })
+    fail("failed to add tui plugin", { path: next, error })
     return [] as PluginLoad[]
   })
   if (!ready.length) {
-    fail("failed to add tui plugin", { path: nextSpec })
+    fail("failed to add tui plugin", { path: next })
     return false
   }
 
   const first = ready[0]
   if (!first) {
-    fail("failed to add tui plugin", { path: nextSpec })
+    fail("failed to add tui plugin", { path: next })
     return false
   }
   if (state.plugins_by_id.has(first.id)) {
@@ -758,7 +776,7 @@ async function addPluginBySpec(state: RuntimeState | undefined, raw: string) {
 
   if (ok) state.pending.delete(spec)
   if (!ok) {
-    fail("failed to add tui plugin", { path: nextSpec })
+    fail("failed to add tui plugin", { path: next })
   }
   return ok
 }
@@ -841,12 +859,11 @@ async function installPluginBySpec(
   const tui = manifest.targets.find((item) => item.kind === "tui")
   if (tui) {
     const file = patch.items.find((item) => item.kind === "tui")?.file
+    const item = tui.opts ? ([spec, tui.opts] as Config.PluginSpec) : spec
     state.pending.set(spec, {
-      item: tui.opts ? [spec, tui.opts] : spec,
-      meta: {
-        scope: global ? "global" : "local",
-        source: (file ?? dir.config) || path.join(patch.dir, "tui.json"),
-      },
+      item,
+      scope: global ? "global" : "local",
+      source: (file ?? dir.config) || path.join(patch.dir, "tui.json"),
     })
   }
 
@@ -929,25 +946,26 @@ export namespace TuiPluginRuntime {
       directory: cwd,
       fn: async () => {
         const config = await TuiConfig.get()
-        const plugins = Flag.OPENCODE_PURE ? [] : (config.plugin ?? [])
-        if (Flag.OPENCODE_PURE && config.plugin?.length) {
-          log.info("skipping external tui plugins in pure mode", { count: config.plugin.length })
+        const records = Flag.OPENCODE_PURE ? [] : (config.plugin_records ?? [])
+        if (Flag.OPENCODE_PURE && config.plugin_records?.length) {
+          log.info("skipping external tui plugins in pure mode", { count: config.plugin_records.length })
         }
 
         for (const item of INTERNAL_TUI_PLUGINS) {
           log.info("loading internal tui plugin", { id: item.id })
           const entry = loadInternalPlugin(item)
           const meta = createMeta(entry.source, entry.spec, entry.target, undefined, entry.id)
-          for (const plugin of collectPluginEntries(entry, meta)) {
-            addPluginEntry(next, plugin)
-          }
+          addPluginEntry(next, {
+            id: entry.id,
+            load: entry,
+            meta,
+            themes: {},
+            plugin: entry.module.tui,
+            enabled: true,
+          })
         }
 
-        const ready = await resolveExternalPlugins(
-          plugins,
-          () => TuiConfig.waitForDependencies(),
-          (item) => config.plugin_meta?.[Config.pluginSpecifier(item)],
-        )
+        const ready = await resolveExternalPlugins(records, () => TuiConfig.waitForDependencies())
         await addExternalPluginEntries(next, ready)
 
         applyInitialPluginEnabledState(next, config)

packages/opencode/src/cli/cmd/tui/routes/home.tsx

@@ -1,15 +1,11 @@
 import { Prompt, type PromptRef } from "@tui/component/prompt"
-import { createEffect, createMemo, Match, on, onMount, Show, Switch } from "solid-js"
-import { useTheme } from "@tui/context/theme"
+import { createEffect, on, onMount } from "solid-js"
 import { Logo } from "../component/logo"
-import { Locale } from "@/util/locale"
 import { useSync } from "../context/sync"
 import { Toast } from "../ui/toast"
 import { useArgs } from "../context/args"
-import { useDirectory } from "../context/directory"
 import { useRouteData } from "@tui/context/route"
 import { usePromptRef } from "../context/prompt"
-import { Installation } from "@/installation"
 import { useLocal } from "../context/local"
 import { TuiPluginRuntime } from "../plugin"
 
@@ -22,37 +18,8 @@ const placeholder = {
 
 export function Home() {
   const sync = useSync()
-  const { theme } = useTheme()
   const route = useRouteData("home")
   const promptRef = usePromptRef()
-  const mcp = createMemo(() => Object.keys(sync.data.mcp).length > 0)
-  const mcpError = createMemo(() => {
-    return Object.values(sync.data.mcp).some((x) => x.status === "failed")
-  })
-
-  const connectedMcpCount = createMemo(() => {
-    return Object.values(sync.data.mcp).filter((x) => x.status === "connected").length
-  })
-
-  const Hint = (
-    <box flexShrink={0} flexDirection="row" gap={1}>
-      <Show when={connectedMcpCount() > 0}>
-        <text fg={theme.text}>
-          <Switch>
-            <Match when={mcpError()}>
-              <span style={{ fg: theme.error }}>•</span> mcp errors{" "}
-              <span style={{ fg: theme.textMuted }}>ctrl+x s</span>
-            </Match>
-            <Match when={true}>
-              <span style={{ fg: theme.success }}>•</span>{" "}
-              {Locale.pluralize(connectedMcpCount(), "{} mcp server", "{} mcp servers")}
-            </Match>
-          </Switch>
-        </text>
-      </Show>
-    </box>
-  )
-
   let prompt: PromptRef | undefined
   const args = useArgs()
   const local = useLocal()
@@ -81,7 +48,6 @@ export function Home() {
       },
     ),
   )
-  const directory = useDirectory()
 
   return (
     <>
@@ -101,7 +67,6 @@ export function Home() {
                 prompt = r
                 promptRef.set(r)
               }}
-              hint={Hint}
               workspaceID={route.workspaceID}
               placeholders={placeholder}
             />
@@ -111,28 +76,8 @@ export function Home() {
         <box flexGrow={1} minHeight={0} />
         <Toast />
       </box>
-      <box paddingTop={1} paddingBottom={1} paddingLeft={2} paddingRight={2} flexDirection="row" flexShrink={0} gap={2}>
-        <text fg={theme.textMuted}>{directory()}</text>
-        <box gap={1} flexDirection="row" flexShrink={0}>
-          <Show when={mcp()}>
-            <text fg={theme.text}>
-              <Switch>
-                <Match when={mcpError()}>
-                  <span style={{ fg: theme.error }}>⊙ </span>
-                </Match>
-                <Match when={true}>
-                  <span style={{ fg: connectedMcpCount() > 0 ? theme.success : theme.textMuted }}>⊙ </span>
-                </Match>
-              </Switch>
-              {connectedMcpCount()} MCP
-            </text>
-            <text fg={theme.textMuted}>/status</text>
-          </Show>
-        </box>
-        <box flexGrow={1} />
-        <box flexShrink={0}>
-          <text fg={theme.textMuted}>{Installation.VERSION}</text>
-        </box>
+      <box width="100%" flexShrink={0}>
+        <TuiPluginRuntime.Slot name="home_footer" mode="single_winner" />
       </box>
     </>
   )

packages/opencode/src/config/config.ts

@@ -1,7 +1,6 @@
 import { Log } from "../util/log"
 import path from "path"
 import { pathToFileURL } from "url"
-import { createRequire } from "module"
 import os from "os"
 import z from "zod"
 import { ModelsDev } from "../provider/models"
@@ -366,33 +365,18 @@ export namespace Config {
   export async function resolvePluginSpec(plugin: PluginSpec, configFilepath: string): Promise<PluginSpec> {
     const spec = pluginSpecifier(plugin)
     if (!isPathPluginSpec(spec)) return plugin
-    if (spec.startsWith("file://")) {
-      const resolved = await resolvePathPluginTarget(spec).catch(() => spec)
-      if (Array.isArray(plugin)) return [resolved, plugin[1]]
-      return resolved
-    }
-    if (path.isAbsolute(spec) || /^[A-Za-z]:[\\/]/.test(spec)) {
-      const base = pathToFileURL(spec).href
-      const resolved = await resolvePathPluginTarget(base).catch(() => base)
-      if (Array.isArray(plugin)) return [resolved, plugin[1]]
-      return resolved
-    }
-    try {
-      const base = import.meta.resolve!(spec, configFilepath)
-      const resolved = await resolvePathPluginTarget(base).catch(() => base)
-      if (Array.isArray(plugin)) return [resolved, plugin[1]]
-      return resolved
-    } catch {
-      try {
-        const require = createRequire(configFilepath)
-        const base = pathToFileURL(require.resolve(spec)).href
-        const resolved = await resolvePathPluginTarget(base).catch(() => base)
-        if (Array.isArray(plugin)) return [resolved, plugin[1]]
-        return resolved
-      } catch {
-        return plugin
-      }
-    }
+
+    const base = path.dirname(configFilepath)
+    const file = (() => {
+      if (spec.startsWith("file://")) return spec
+      if (path.isAbsolute(spec) || /^[A-Za-z]:[\\/]/.test(spec)) return pathToFileURL(spec).href
+      return pathToFileURL(path.resolve(base, spec)).href
+    })()
+
+    const resolved = await resolvePathPluginTarget(file).catch(() => file)
+
+    if (Array.isArray(plugin)) return [resolved, plugin[1]]
+    return resolved
   }
 
   /**
@@ -1556,7 +1540,7 @@ export namespace Config {
 
   export const defaultLayer = layer.pipe(
     Layer.provide(AppFileSystem.defaultLayer),
-    Layer.provide(Auth.layer),
+    Layer.provide(Auth.defaultLayer),
     Layer.provide(Account.defaultLayer),
   )
 

packages/opencode/src/config/tui.ts

@@ -22,6 +22,12 @@ export namespace TuiConfig {
     source: string
   }
 
+  export type PluginRecord = {
+    item: Config.PluginSpec
+    scope: PluginMeta["scope"]
+    source: string
+  }
+
   type PluginEntry = {
     item: Config.PluginSpec
     meta: PluginMeta
@@ -33,7 +39,8 @@ export namespace TuiConfig {
   }
 
   export type Info = z.output<typeof Info> & {
-    plugin_meta?: Record<string, PluginMeta>
+    // Internal resolved plugin list used by runtime loading.
+    plugin_records?: PluginRecord[]
   }
 
   function pluginScope(file: string): PluginMeta["scope"] {
@@ -149,10 +156,13 @@ export namespace TuiConfig {
 
     const merged = dedupePlugins(acc.entries)
     acc.result.keybinds = Config.Keybinds.parse(acc.result.keybinds ?? {})
-    acc.result.plugin = merged.map((item) => item.item)
-    acc.result.plugin_meta = merged.length
-      ? Object.fromEntries(merged.map((item) => [Config.pluginSpecifier(item.item), item.meta]))
-      : undefined
+    const list = merged.map((item) => ({
+      item: item.item,
+      scope: item.meta.scope,
+      source: item.meta.source,
+    }))
+    acc.result.plugin = list.map((item) => item.item)
+    acc.result.plugin_records = list.length ? list : undefined
 
     const deps: Promise<void>[] = []
     if (acc.result.plugin?.length) {

packages/opencode/src/effect/runner.ts

@@ -0,0 +1,216 @@
+import { Cause, Deferred, Effect, Exit, Fiber, Option, Schema, Scope, SynchronizedRef } from "effect"
+
+export interface Runner<A, E = never> {
+  readonly state: Runner.State<A, E>
+  readonly busy: boolean
+  readonly ensureRunning: (work: Effect.Effect<A, E>) => Effect.Effect<A, E>
+  readonly startShell: (work: (signal: AbortSignal) => Effect.Effect<A, E>) => Effect.Effect<A, E>
+  readonly cancel: Effect.Effect<void>
+}
+
+export namespace Runner {
+  export class Cancelled extends Schema.TaggedErrorClass<Cancelled>()("RunnerCancelled", {}) {}
+
+  interface RunHandle<A, E> {
+    id: number
+    done: Deferred.Deferred<A, E | Cancelled>
+    fiber: Fiber.Fiber<A, E>
+  }
+
+  interface ShellHandle<A, E> {
+    id: number
+    fiber: Fiber.Fiber<A, E>
+    abort: AbortController
+  }
+
+  interface PendingHandle<A, E> {
+    id: number
+    done: Deferred.Deferred<A, E | Cancelled>
+    work: Effect.Effect<A, E>
+  }
+
+  export type State<A, E> =
+    | { readonly _tag: "Idle" }
+    | { readonly _tag: "Running"; readonly run: RunHandle<A, E> }
+    | { readonly _tag: "Shell"; readonly shell: ShellHandle<A, E> }
+    | { readonly _tag: "ShellThenRun"; readonly shell: ShellHandle<A, E>; readonly run: PendingHandle<A, E> }
+
+  export const make = <A, E = never>(
+    scope: Scope.Scope,
+    opts?: {
+      onIdle?: Effect.Effect<void>
+      onBusy?: Effect.Effect<void>
+      onInterrupt?: Effect.Effect<A, E>
+      busy?: () => never
+    },
+  ): Runner<A, E> => {
+    const ref = SynchronizedRef.makeUnsafe<State<A, E>>({ _tag: "Idle" })
+    const idle = opts?.onIdle ?? Effect.void
+    const busy = opts?.onBusy ?? Effect.void
+    const onInterrupt = opts?.onInterrupt
+    let ids = 0
+
+    const state = () => SynchronizedRef.getUnsafe(ref)
+    const next = () => {
+      ids += 1
+      return ids
+    }
+
+    const complete = (done: Deferred.Deferred<A, E | Cancelled>, exit: Exit.Exit<A, E>) =>
+      Exit.isFailure(exit) && Cause.hasInterruptsOnly(exit.cause)
+        ? Deferred.fail(done, new Cancelled()).pipe(Effect.asVoid)
+        : Deferred.done(done, exit).pipe(Effect.asVoid)
+
+    const idleIfCurrent = () =>
+      SynchronizedRef.modify(ref, (st) => [st._tag === "Idle" ? idle : Effect.void, st] as const).pipe(Effect.flatten)
+
+    const finishRun = (id: number, done: Deferred.Deferred<A, E | Cancelled>, exit: Exit.Exit<A, E>) =>
+      SynchronizedRef.modify(
+        ref,
+        (st) =>
+          [
+            Effect.gen(function* () {
+              if (st._tag === "Running" && st.run.id === id) yield* idle
+              yield* complete(done, exit)
+            }),
+            st._tag === "Running" && st.run.id === id ? ({ _tag: "Idle" } as const) : st,
+          ] as const,
+      ).pipe(Effect.flatten)
+
+    const startRun = (work: Effect.Effect<A, E>, done: Deferred.Deferred<A, E | Cancelled>) =>
+      Effect.gen(function* () {
+        const id = next()
+        const fiber = yield* work.pipe(
+          Effect.onExit((exit) => finishRun(id, done, exit)),
+          Effect.forkIn(scope),
+        )
+        return { id, done, fiber } satisfies RunHandle<A, E>
+      })
+
+    const finishShell = (id: number) =>
+      SynchronizedRef.modifyEffect(
+        ref,
+        Effect.fnUntraced(function* (st) {
+          if (st._tag === "Shell" && st.shell.id === id) return [idle, { _tag: "Idle" }] as const
+          if (st._tag === "ShellThenRun" && st.shell.id === id) {
+            const run = yield* startRun(st.run.work, st.run.done)
+            return [Effect.void, { _tag: "Running", run }] as const
+          }
+          return [Effect.void, st] as const
+        }),
+      ).pipe(Effect.flatten)
+
+    const stopShell = (shell: ShellHandle<A, E>) =>
+      Effect.gen(function* () {
+        shell.abort.abort()
+        const exit = yield* Fiber.await(shell.fiber).pipe(Effect.timeoutOption("100 millis"))
+        if (Option.isNone(exit)) yield* Fiber.interrupt(shell.fiber)
+        yield* Fiber.await(shell.fiber).pipe(Effect.exit, Effect.asVoid)
+      })
+
+    const ensureRunning = (work: Effect.Effect<A, E>) =>
+      SynchronizedRef.modifyEffect(
+        ref,
+        Effect.fnUntraced(function* (st) {
+          switch (st._tag) {
+            case "Running":
+            case "ShellThenRun":
+              return [Deferred.await(st.run.done), st] as const
+            case "Shell": {
+              const run = {
+                id: next(),
+                done: yield* Deferred.make<A, E | Cancelled>(),
+                work,
+              } satisfies PendingHandle<A, E>
+              return [Deferred.await(run.done), { _tag: "ShellThenRun", shell: st.shell, run }] as const
+            }
+            case "Idle": {
+              const done = yield* Deferred.make<A, E | Cancelled>()
+              const run = yield* startRun(work, done)
+              return [Deferred.await(done), { _tag: "Running", run }] as const
+            }
+          }
+        }),
+      ).pipe(
+        Effect.flatten,
+        Effect.catch(
+          (e): Effect.Effect<A, E> => (e instanceof Cancelled ? (onInterrupt ?? Effect.die(e)) : Effect.fail(e as E)),
+        ),
+      )
+
+    const startShell = (work: (signal: AbortSignal) => Effect.Effect<A, E>) =>
+      SynchronizedRef.modifyEffect(
+        ref,
+        Effect.fnUntraced(function* (st) {
+          if (st._tag !== "Idle") {
+            return [
+              Effect.sync(() => {
+                if (opts?.busy) opts.busy()
+                throw new Error("Runner is busy")
+              }),
+              st,
+            ] as const
+          }
+          yield* busy
+          const id = next()
+          const abort = new AbortController()
+          const fiber = yield* work(abort.signal).pipe(Effect.ensuring(finishShell(id)), Effect.forkChild)
+          const shell = { id, fiber, abort } satisfies ShellHandle<A, E>
+          return [
+            Effect.gen(function* () {
+              const exit = yield* Fiber.await(fiber)
+              if (Exit.isSuccess(exit)) return exit.value
+              if (Cause.hasInterruptsOnly(exit.cause) && onInterrupt) return yield* onInterrupt
+              return yield* Effect.failCause(exit.cause)
+            }),
+            { _tag: "Shell", shell },
+          ] as const
+        }),
+      ).pipe(Effect.flatten)
+
+    const cancel = SynchronizedRef.modify(ref, (st) => {
+      switch (st._tag) {
+        case "Idle":
+          return [Effect.void, st] as const
+        case "Running":
+          return [
+            Effect.gen(function* () {
+              yield* Fiber.interrupt(st.run.fiber)
+              yield* Deferred.await(st.run.done).pipe(Effect.exit, Effect.asVoid)
+              yield* idleIfCurrent()
+            }),
+            { _tag: "Idle" } as const,
+          ] as const
+        case "Shell":
+          return [
+            Effect.gen(function* () {
+              yield* stopShell(st.shell)
+              yield* idleIfCurrent()
+            }),
+            { _tag: "Idle" } as const,
+          ] as const
+        case "ShellThenRun":
+          return [
+            Effect.gen(function* () {
+              yield* Deferred.fail(st.run.done, new Cancelled()).pipe(Effect.asVoid)
+              yield* stopShell(st.shell)
+              yield* idleIfCurrent()
+            }),
+            { _tag: "Idle" } as const,
+          ] as const
+      }
+    }).pipe(Effect.flatten)
+
+    return {
+      get state() {
+        return state()
+      },
+      get busy() {
+        return state()._tag !== "Idle"
+      },
+      ensureRunning,
+      startShell,
+      cancel,
+    }
+  }
+}

packages/opencode/src/file/index.ts

@@ -541,7 +541,7 @@ export namespace File {
         const exists = yield* appFs.existsSafe(full)
         if (!exists) return { type: "text" as const, content: "" }
 
-        const mimeType = Filesystem.mimeType(full)
+        const mimeType = AppFileSystem.mimeType(full)
         const encode = knownText ? false : shouldEncode(mimeType)
 
         if (encode && !isImage(mimeType)) return { type: "binary" as const, content: "", mimeType }

packages/opencode/src/file/time.ts

@@ -1,9 +1,9 @@
-import { DateTime, Effect, Layer, Semaphore, ServiceMap } from "effect"
+import { DateTime, Effect, Layer, Option, Semaphore, ServiceMap } from "effect"
 import { InstanceState } from "@/effect/instance-state"
 import { makeRuntime } from "@/effect/run-service"
+import { AppFileSystem } from "@/filesystem"
 import { Flag } from "@/flag/flag"
 import type { SessionID } from "@/session/schema"
-import { Filesystem } from "../util/filesystem"
 import { Log } from "../util/log"
 
 export namespace FileTime {
@@ -12,21 +12,9 @@ export namespace FileTime {
   export type Stamp = {
     readonly read: Date
     readonly mtime: number | undefined
-    readonly ctime: number | undefined
     readonly size: number | undefined
   }
 
-  const stamp = Effect.fnUntraced(function* (file: string) {
-    const stat = Filesystem.stat(file)
-    const size = typeof stat?.size === "bigint" ? Number(stat.size) : stat?.size
-    return {
-      read: yield* DateTime.nowAsDate,
-      mtime: stat?.mtime?.getTime(),
-      ctime: stat?.ctime?.getTime(),
-      size,
-    }
-  })
-
   const session = (reads: Map<SessionID, Map<string, Stamp>>, sessionID: SessionID) => {
     const value = reads.get(sessionID)
     if (value) return value
@@ -53,7 +41,17 @@ export namespace FileTime {
   export const layer = Layer.effect(
     Service,
     Effect.gen(function* () {
+      const fsys = yield* AppFileSystem.Service
       const disableCheck = yield* Flag.OPENCODE_DISABLE_FILETIME_CHECK
+
+      const stamp = Effect.fnUntraced(function* (file: string) {
+        const info = yield* fsys.stat(file).pipe(Effect.catch(() => Effect.succeed(undefined)))
+        return {
+          read: yield* DateTime.nowAsDate,
+          mtime: info ? Option.getOrUndefined(info.mtime)?.getTime() : undefined,
+          size: info ? Number(info.size) : undefined,
+        }
+      })
       const state = yield* InstanceState.make<State>(
         Effect.fn("FileTime.state")(() =>
           Effect.succeed({
@@ -92,7 +90,7 @@ export namespace FileTime {
         if (!time) throw new Error(`You must read file ${filepath} before overwriting it. Use the Read tool first`)
 
         const next = yield* stamp(filepath)
-        const changed = next.mtime !== time.mtime || next.ctime !== time.ctime || next.size !== time.size
+        const changed = next.mtime !== time.mtime || next.size !== time.size
         if (!changed) return
 
         throw new Error(
@@ -108,7 +106,9 @@ export namespace FileTime {
     }),
   ).pipe(Layer.orDie)
 
-  const { runPromise } = makeRuntime(Service, layer)
+  export const defaultLayer = layer.pipe(Layer.provide(AppFileSystem.defaultLayer))
+
+  const { runPromise } = makeRuntime(Service, defaultLayer)
 
   export function read(sessionID: SessionID, file: string) {
     return runPromise((s) => s.read(sessionID, file))

packages/opencode/src/plugin/index.ts

@@ -14,19 +14,8 @@ import { Effect, Layer, ServiceMap, Stream } from "effect"
 import { InstanceState } from "@/effect/instance-state"
 import { makeRuntime } from "@/effect/run-service"
 import { errorMessage } from "@/util/error"
-import { Installation } from "@/installation"
-import {
-  checkPluginCompatibility,
-  isDeprecatedPlugin,
-  parsePluginSpecifier,
-  pluginSource,
-  readPluginId,
-  readV1Plugin,
-  resolvePluginEntrypoint,
-  resolvePluginId,
-  resolvePluginTarget,
-  type PluginSource,
-} from "./shared"
+import { PluginLoader } from "./loader"
+import { parsePluginSpecifier, readPluginId, readV1Plugin, resolvePluginId } from "./shared"
 
 export namespace Plugin {
   const log = Log.create({ service: "plugin" })
@@ -36,11 +25,7 @@ export namespace Plugin {
   }
 
   type Loaded = {
-    item: Config.PluginSpec
-    spec: string
-    target: string
-    source: PluginSource
-    mod: Record<string, unknown>
+    row: PluginLoader.Loaded
   }
 
   // Hook names that follow the (input, output) => Promise<void> trigger pattern
@@ -93,91 +78,22 @@ export namespace Plugin {
     return result
   }
 
-  async function resolvePlugin(spec: string) {
-    const parsed = parsePluginSpecifier(spec)
-    const target = await resolvePluginTarget(spec, parsed).catch((err) => {
-      const cause = err instanceof Error ? err.cause : err
-      const detail = errorMessage(cause ?? err)
-      log.error("failed to install plugin", { pkg: parsed.pkg, version: parsed.version, error: detail })
-      Bus.publish(Session.Event.Error, {
-        error: new NamedError.Unknown({
-          message: `Failed to install plugin ${parsed.pkg}@${parsed.version}: ${detail}`,
-        }).toObject(),
-      })
-      return ""
-    })
-    if (!target) return
-    return target
-  }
-
-  async function prepPlugin(item: Config.PluginSpec): Promise<Loaded | undefined> {
-    const spec = Config.pluginSpecifier(item)
-    if (isDeprecatedPlugin(spec)) return
-    log.info("loading plugin", { path: spec })
-    const resolved = await resolvePlugin(spec)
-    if (!resolved) return
-
-    const source = pluginSource(spec)
-    if (source === "npm") {
-      const incompatible = await checkPluginCompatibility(resolved, Installation.VERSION)
-        .then(() => false)
-        .catch((err) => {
-          const message = errorMessage(err)
-          log.warn("plugin incompatible", { path: spec, error: message })
-          Bus.publish(Session.Event.Error, {
-            error: new NamedError.Unknown({
-              message: `Plugin ${spec} skipped: ${message}`,
-            }).toObject(),
-          })
-          return true
-        })
-      if (incompatible) return
-    }
-
-    const target = resolved
-    const entry = await resolvePluginEntrypoint(spec, target, "server").catch((err) => {
-      const message = errorMessage(err)
-      log.error("failed to resolve plugin server entry", { path: spec, target, error: message })
-      Bus.publish(Session.Event.Error, {
-        error: new NamedError.Unknown({
-          message: `Failed to load plugin ${spec}: ${message}`,
-        }).toObject(),
-      })
-      return
-    })
-    if (!entry) return
-
-    const mod = await import(entry).catch((err) => {
-      const message = errorMessage(err)
-      log.error("failed to load plugin", { path: spec, target: entry, error: message })
-      Bus.publish(Session.Event.Error, {
-        error: new NamedError.Unknown({
-          message: `Failed to load plugin ${spec}: ${message}`,
-        }).toObject(),
-      })
-      return
-    })
-    if (!mod) return
-
-    return {
-      item,
-      spec,
-      target,
-      source,
-      mod,
-    }
-  }
-
   async function applyPlugin(load: Loaded, input: PluginInput, hooks: Hooks[]) {
-    const plugin = readV1Plugin(load.mod, load.spec, "server", "detect")
+    const plugin = readV1Plugin(load.row.mod, load.row.spec, "server", "detect")
     if (plugin) {
-      await resolvePluginId(load.source, load.spec, load.target, readPluginId(plugin.id, load.spec))
-      hooks.push(await (plugin as PluginModule).server(input, Config.pluginOptions(load.item)))
+      await resolvePluginId(
+        load.row.source,
+        load.row.spec,
+        load.row.target,
+        readPluginId(plugin.id, load.row.spec),
+        load.row.pkg,
+      )
+      hooks.push(await (plugin as PluginModule).server(input, load.row.options))
       return
     }
 
-    for (const server of getLegacyPlugins(load.mod)) {
-      hooks.push(await server(input, Config.pluginOptions(load.item)))
+    for (const server of getLegacyPlugins(load.row.mod)) {
+      hooks.push(await server(input, load.row.options))
     }
   }
 
@@ -232,7 +148,74 @@ export namespace Plugin {
           }
           if (plugins.length) yield* config.waitForDependencies()
 
-          const loaded = yield* Effect.promise(() => Promise.all(plugins.map((item) => prepPlugin(item))))
+          const loaded = yield* Effect.promise(() =>
+            Promise.all(
+              plugins.map(async (item) => {
+                const plan = PluginLoader.plan(item)
+                if (plan.deprecated) return
+                log.info("loading plugin", { path: plan.spec })
+
+                const resolved = await PluginLoader.resolve(plan, "server")
+                if (!resolved.ok) {
+                  const cause =
+                    resolved.error instanceof Error ? (resolved.error.cause ?? resolved.error) : resolved.error
+                  const message = errorMessage(cause)
+
+                  if (resolved.stage === "install") {
+                    const parsed = parsePluginSpecifier(plan.spec)
+                    log.error("failed to install plugin", {
+                      pkg: parsed.pkg,
+                      version: parsed.version,
+                      error: message,
+                    })
+                    Bus.publish(Session.Event.Error, {
+                      error: new NamedError.Unknown({
+                        message: `Failed to install plugin ${parsed.pkg}@${parsed.version}: ${message}`,
+                      }).toObject(),
+                    })
+                    return
+                  }
+
+                  if (resolved.stage === "compatibility") {
+                    log.warn("plugin incompatible", { path: plan.spec, error: message })
+                    Bus.publish(Session.Event.Error, {
+                      error: new NamedError.Unknown({
+                        message: `Plugin ${plan.spec} skipped: ${message}`,
+                      }).toObject(),
+                    })
+                    return
+                  }
+
+                  log.error("failed to resolve plugin server entry", {
+                    path: plan.spec,
+                    error: message,
+                  })
+                  Bus.publish(Session.Event.Error, {
+                    error: new NamedError.Unknown({
+                      message: `Failed to load plugin ${plan.spec}: ${message}`,
+                    }).toObject(),
+                  })
+                  return
+                }
+
+                const mod = await PluginLoader.load(resolved.value)
+                if (!mod.ok) {
+                  const message = errorMessage(mod.error)
+                  log.error("failed to load plugin", { path: plan.spec, target: resolved.value.entry, error: message })
+                  Bus.publish(Session.Event.Error, {
+                    error: new NamedError.Unknown({
+                      message: `Failed to load plugin ${plan.spec}: ${message}`,
+                    }).toObject(),
+                  })
+                  return
+                }
+
+                return {
+                  row: mod.value,
+                }
+              }),
+            ),
+          )
           for (const load of loaded) {
             if (!load) continue
 
@@ -242,14 +225,14 @@ export namespace Plugin {
               try: () => applyPlugin(load, input, hooks),
               catch: (err) => {
                 const message = errorMessage(err)
-                log.error("failed to load plugin", { path: load.spec, error: message })
+                log.error("failed to load plugin", { path: load.row.spec, error: message })
                 return message
               },
             }).pipe(
               Effect.catch((message) =>
                 bus.publish(Session.Event.Error, {
                   error: new NamedError.Unknown({
-                    message: `Failed to load plugin ${load.spec}: ${message}`,
+                    message: `Failed to load plugin ${load.row.spec}: ${message}`,
                   }).toObject(),
                 }),
               ),

packages/opencode/src/plugin/loader.ts

@@ -0,0 +1,135 @@
+import { Config } from "@/config/config"
+import { Installation } from "@/installation"
+import {
+  checkPluginCompatibility,
+  createPluginEntry,
+  isDeprecatedPlugin,
+  resolvePluginTarget,
+  type PluginKind,
+  type PluginPackage,
+  type PluginSource,
+} from "./shared"
+
+export namespace PluginLoader {
+  export type Plan = {
+    item: Config.PluginSpec
+    spec: string
+    options: Config.PluginOptions | undefined
+    deprecated: boolean
+  }
+
+  export type Resolved = Plan & {
+    source: PluginSource
+    target: string
+    entry: string
+    pkg?: PluginPackage
+  }
+
+  export type Loaded = Resolved & {
+    mod: Record<string, unknown>
+  }
+
+  export function plan(item: Config.PluginSpec): Plan {
+    const spec = Config.pluginSpecifier(item)
+    return {
+      item,
+      spec,
+      options: Config.pluginOptions(item),
+      deprecated: isDeprecatedPlugin(spec),
+    }
+  }
+
+  export async function resolve(
+    plan: Plan,
+    kind: PluginKind,
+  ): Promise<
+    { ok: true; value: Resolved } | { ok: false; stage: "install" | "entry" | "compatibility"; error: unknown }
+  > {
+    let target = ""
+    try {
+      target = await resolvePluginTarget(plan.spec)
+    } catch (error) {
+      return {
+        ok: false,
+        stage: "install",
+        error,
+      }
+    }
+    if (!target) {
+      return {
+        ok: false,
+        stage: "install",
+        error: new Error(`Plugin ${plan.spec} target is empty`),
+      }
+    }
+
+    let base
+    try {
+      base = await createPluginEntry(plan.spec, target, kind)
+    } catch (error) {
+      return {
+        ok: false,
+        stage: "entry",
+        error,
+      }
+    }
+
+    if (!base.entry) {
+      return {
+        ok: false,
+        stage: "entry",
+        error: new Error(`Plugin ${plan.spec} entry is empty`),
+      }
+    }
+
+    if (base.source === "npm") {
+      try {
+        await checkPluginCompatibility(base.target, Installation.VERSION, base.pkg)
+      } catch (error) {
+        return {
+          ok: false,
+          stage: "compatibility",
+          error,
+        }
+      }
+    }
+
+    return {
+      ok: true,
+      value: {
+        ...plan,
+        source: base.source,
+        target: base.target,
+        entry: base.entry,
+        pkg: base.pkg,
+      },
+    }
+  }
+
+  export async function load(row: Resolved): Promise<{ ok: true; value: Loaded } | { ok: false; error: unknown }> {
+    let mod
+    try {
+      mod = await import(row.entry)
+    } catch (error) {
+      return {
+        ok: false,
+        error,
+      }
+    }
+
+    if (!mod) {
+      return {
+        ok: false,
+        error: new Error(`Plugin ${row.spec} module is empty`),
+      }
+    }
+
+    return {
+      ok: true,
+      value: {
+        ...row,
+        mod,
+      },
+    }
+  }
+}

packages/opencode/src/plugin/meta.ts

@@ -11,6 +11,13 @@ import { parsePluginSpecifier, pluginSource } from "./shared"
 export namespace PluginMeta {
   type Source = "file" | "npm"
 
+  export type Theme = {
+    src: string
+    dest: string
+    mtime?: number
+    size?: number
+  }
+
   export type Entry = {
     id: string
     source: Source
@@ -24,6 +31,7 @@ export namespace PluginMeta {
     time_changed: number
     load_count: number
     fingerprint: string
+    themes?: Record<string, Theme>
   }
 
   export type State = "first" | "updated" | "same"
@@ -35,7 +43,7 @@ export namespace PluginMeta {
   }
 
   type Store = Record<string, Entry>
-  type Core = Omit<Entry, "first_time" | "last_time" | "time_changed" | "load_count" | "fingerprint">
+  type Core = Omit<Entry, "first_time" | "last_time" | "time_changed" | "load_count" | "fingerprint" | "themes">
   type Row = Touch & { core: Core }
 
   function storePath() {
@@ -52,11 +60,11 @@ export namespace PluginMeta {
     return
   }
 
-  function modifiedAt(file: string) {
-    const stat = Filesystem.stat(file)
+  async function modifiedAt(file: string) {
+    const stat = await Filesystem.statAsync(file)
     if (!stat) return
-    const value = stat.mtimeMs
-    return Math.floor(typeof value === "bigint" ? Number(value) : value)
+    const mtime = stat.mtimeMs
+    return Math.floor(typeof mtime === "bigint" ? Number(mtime) : mtime)
   }
 
   function resolvedTarget(target: string) {
@@ -66,7 +74,7 @@ export namespace PluginMeta {
 
   async function npmVersion(target: string) {
     const resolved = resolvedTarget(target)
-    const stat = Filesystem.stat(resolved)
+    const stat = await Filesystem.statAsync(resolved)
     const dir = stat?.isDirectory() ? resolved : path.dirname(resolved)
     return Filesystem.readJson<{ version?: string }>(path.join(dir, "package.json"))
       .then((item) => item.version)
@@ -84,7 +92,7 @@ export namespace PluginMeta {
         source,
         spec,
         target,
-        modified: file ? modifiedAt(file) : undefined,
+        modified: file ? await modifiedAt(file) : undefined,
       }
     }
 
@@ -122,6 +130,7 @@ export namespace PluginMeta {
       time_changed: prev?.time_changed ?? now,
       load_count: (prev?.load_count ?? 0) + 1,
       fingerprint: fingerprint(core),
+      themes: prev?.themes,
     }
     const state: State = !prev ? "first" : prev.fingerprint === entry.fingerprint ? "same" : "updated"
     if (state === "updated") entry.time_changed = now
@@ -158,6 +167,20 @@ export namespace PluginMeta {
     })
   }
 
+  export async function setTheme(id: string, name: string, theme: Theme): Promise<void> {
+    const file = storePath()
+    await Flock.withLock(lock(file), async () => {
+      const store = await read(file)
+      const entry = store[id]
+      if (!entry) return
+      entry.themes = {
+        ...(entry.themes ?? {}),
+        [name]: theme,
+      }
+      await Filesystem.writeJson(file, store)
+    })
+  }
+
   export async function list(): Promise<Store> {
     const file = storePath()
     return Flock.withLock(lock(file), async () => read(file))

packages/opencode/src/plugin/shared.ts

@@ -23,19 +23,31 @@ export type PluginSource = "file" | "npm"
 export type PluginKind = "server" | "tui"
 type PluginMode = "strict" | "detect"
 
-export function pluginSource(spec: string): PluginSource {
-  return spec.startsWith("file://") ? "file" : "npm"
+export type PluginPackage = {
+  dir: string
+  pkg: string
+  json: Record<string, unknown>
 }
 
-function hasEntrypoint(json: Record<string, unknown>, kind: PluginKind) {
-  if (!isRecord(json.exports)) return false
-  return `./${kind}` in json.exports
+export type PluginEntry = {
+  spec: string
+  source: PluginSource
+  target: string
+  pkg?: PluginPackage
+  entry: string
+}
+
+const INDEX_FILES = ["index.ts", "index.tsx", "index.js", "index.mjs", "index.cjs"]
+
+export function pluginSource(spec: string): PluginSource {
+  if (isPathPluginSpec(spec)) return "file"
+  return "npm"
 }
 
 function resolveExportPath(raw: string, dir: string) {
-  if (raw.startsWith("./") || raw.startsWith("../")) return path.resolve(dir, raw)
   if (raw.startsWith("file://")) return fileURLToPath(raw)
-  return raw
+  if (path.isAbsolute(raw)) return raw
+  return path.resolve(dir, raw)
 }
 
 function extractExportValue(value: unknown): string | undefined {
@@ -48,26 +60,97 @@ function extractExportValue(value: unknown): string | undefined {
   return undefined
 }
 
-export async function resolvePluginEntrypoint(spec: string, target: string, kind: PluginKind) {
-  const pkg = await readPluginPackage(target).catch(() => undefined)
-  if (!pkg) return target
-  if (!hasEntrypoint(pkg.json, kind)) return target
-
-  const exports = pkg.json.exports
-  if (!isRecord(exports)) return target
-  const raw = extractExportValue(exports[`./${kind}`])
-  if (!raw) return target
+function packageMain(pkg: PluginPackage) {
+  const value = pkg.json.main
+  if (typeof value !== "string") return
+  const next = value.trim()
+  if (!next) return
+  return next
+}
 
+function resolvePackagePath(spec: string, raw: string, kind: PluginKind, pkg: PluginPackage) {
   const resolved = resolveExportPath(raw, pkg.dir)
   const root = Filesystem.resolve(pkg.dir)
   const next = Filesystem.resolve(resolved)
   if (!Filesystem.contains(root, next)) {
     throw new Error(`Plugin ${spec} resolved ${kind} entry outside plugin directory`)
   }
-
   return pathToFileURL(next).href
 }
 
+function resolvePackageEntrypoint(spec: string, kind: PluginKind, pkg: PluginPackage) {
+  const exports = pkg.json.exports
+  if (isRecord(exports)) {
+    const raw = extractExportValue(exports[`./${kind}`])
+    if (raw) return resolvePackagePath(spec, raw, kind, pkg)
+  }
+
+  if (kind !== "server") return
+  const main = packageMain(pkg)
+  if (!main) return
+  return resolvePackagePath(spec, main, kind, pkg)
+}
+
+function targetPath(target: string) {
+  if (target.startsWith("file://")) return fileURLToPath(target)
+  if (path.isAbsolute(target)) return target
+}
+
+async function resolveDirectoryIndex(dir: string) {
+  for (const name of INDEX_FILES) {
+    const file = path.join(dir, name)
+    if (await Filesystem.exists(file)) return file
+  }
+}
+
+async function resolveTargetDirectory(target: string) {
+  const file = targetPath(target)
+  if (!file) return
+  const stat = await Filesystem.stat(file)
+  if (!stat?.isDirectory()) return
+  return file
+}
+
+async function resolvePluginEntrypoint(spec: string, target: string, kind: PluginKind, pkg?: PluginPackage) {
+  const source = pluginSource(spec)
+  const hit =
+    pkg ?? (source === "npm" ? await readPluginPackage(target) : await readPluginPackage(target).catch(() => undefined))
+  if (!hit) return target
+
+  const entry = resolvePackageEntrypoint(spec, kind, hit)
+  if (entry) return entry
+
+  const dir = await resolveTargetDirectory(target)
+
+  if (kind === "tui") {
+    if (source === "file" && dir) {
+      const index = await resolveDirectoryIndex(dir)
+      if (index) return pathToFileURL(index).href
+    }
+
+    if (source === "npm") {
+      throw new TypeError(`Plugin ${spec} must define package.json exports["./tui"]`)
+    }
+
+    if (dir) {
+      throw new TypeError(`Plugin ${spec} must define package.json exports["./tui"] or include index file`)
+    }
+
+    return target
+  }
+
+  if (dir && isRecord(hit.json.exports)) {
+    if (source === "file") {
+      const index = await resolveDirectoryIndex(dir)
+      if (index) return pathToFileURL(index).href
+    }
+
+    throw new TypeError(`Plugin ${spec} must define package.json exports["./server"] or package.json main`)
+  }
+
+  return target
+}
+
 export function isPathPluginSpec(spec: string) {
   return spec.startsWith("file://") || spec.startsWith(".") || path.isAbsolute(spec) || /^[A-Za-z]:[\\/]/.test(spec)
 }
@@ -81,19 +164,21 @@ export async function resolvePathPluginTarget(spec: string) {
     return pathToFileURL(file).href
   }
 
-  const pkg = await Filesystem.readJson<Record<string, unknown>>(path.join(file, "package.json")).catch(() => undefined)
-  if (!pkg) throw new Error(`Plugin directory ${file} is missing package.json`)
-  if (typeof pkg.main !== "string" || !pkg.main.trim()) {
-    throw new Error(`Plugin directory ${file} must define package.json main`)
+  if (await Filesystem.exists(path.join(file, "package.json"))) {
+    return pathToFileURL(file).href
   }
-  return pathToFileURL(path.resolve(file, pkg.main)).href
+
+  const index = await resolveDirectoryIndex(file)
+  if (index) return pathToFileURL(index).href
+
+  throw new Error(`Plugin directory ${file} is missing package.json or index file`)
 }
 
-export async function checkPluginCompatibility(target: string, opencodeVersion: string) {
+export async function checkPluginCompatibility(target: string, opencodeVersion: string, pkg?: PluginPackage) {
   if (!semver.valid(opencodeVersion) || semver.major(opencodeVersion) === 0) return
-  const pkg = await readPluginPackage(target).catch(() => undefined)
-  if (!pkg) return
-  const engines = pkg.json.engines
+  const hit = pkg ?? (await readPluginPackage(target).catch(() => undefined))
+  if (!hit) return
+  const engines = hit.json.engines
   if (!isRecord(engines)) return
   const range = engines.opencode
   if (typeof range !== "string") return
@@ -107,7 +192,7 @@ export async function resolvePluginTarget(spec: string, parsed = parsePluginSpec
   return BunProc.install(parsed.pkg, parsed.version)
 }
 
-export async function readPluginPackage(target: string) {
+export async function readPluginPackage(target: string): Promise<PluginPackage> {
   const file = target.startsWith("file://") ? fileURLToPath(target) : target
   const stat = await Filesystem.stat(file)
   const dir = stat?.isDirectory() ? file : path.dirname(file)
@@ -116,6 +201,20 @@ export async function readPluginPackage(target: string) {
   return { dir, pkg, json }
 }
 
+export async function createPluginEntry(spec: string, target: string, kind: PluginKind): Promise<PluginEntry> {
+  const source = pluginSource(spec)
+  const pkg =
+    source === "npm" ? await readPluginPackage(target) : await readPluginPackage(target).catch(() => undefined)
+  const entry = await resolvePluginEntrypoint(spec, target, kind, pkg)
+  return {
+    spec,
+    source,
+    target,
+    pkg,
+    entry,
+  }
+}
+
 export function readPluginId(id: unknown, spec: string) {
   if (id === undefined) return
   if (typeof id !== "string") throw new TypeError(`Plugin ${spec} has invalid id type ${typeof id}`)
@@ -158,15 +257,21 @@ export function readV1Plugin(
   return value
 }
 
-export async function resolvePluginId(source: PluginSource, spec: string, target: string, id: string | undefined) {
+export async function resolvePluginId(
+  source: PluginSource,
+  spec: string,
+  target: string,
+  id: string | undefined,
+  pkg?: PluginPackage,
+) {
   if (source === "file") {
     if (id) return id
     throw new TypeError(`Path plugin ${spec} must export id`)
   }
   if (id) return id
-  const pkg = await readPluginPackage(target)
-  if (typeof pkg.json.name !== "string" || !pkg.json.name.trim()) {
-    throw new TypeError(`Plugin package ${pkg.pkg} is missing name`)
+  const hit = pkg ?? (await readPluginPackage(target))
+  if (typeof hit.json.name !== "string" || !hit.json.name.trim()) {
+    throw new TypeError(`Plugin package ${hit.pkg} is missing name`)
   }
-  return pkg.json.name.trim()
+  return hit.json.name.trim()
 }

packages/opencode/src/provider/auth.ts

@@ -230,7 +230,7 @@ export namespace ProviderAuth {
     }),
   )
 
-  export const defaultLayer = layer.pipe(Layer.provide(Auth.layer))
+  export const defaultLayer = layer.pipe(Layer.provide(Auth.defaultLayer))
 
   const { runPromise } = makeRuntime(Service, defaultLayer)
 

packages/opencode/src/server/middleware.ts

@@ -1,6 +1,7 @@
 import { Provider } from "../provider/provider"
 import { NamedError } from "@opencode-ai/util/error"
 import { NotFoundError } from "../storage/db"
+import { Session } from "../session"
 import type { ContentfulStatusCode } from "hono/utils/http-status"
 import type { ErrorHandler } from "hono"
 import { HTTPException } from "hono/http-exception"
@@ -20,6 +21,9 @@ export function errorHandler(log: Log.Logger): ErrorHandler {
       else status = 500
       return c.json(err.toObject(), { status })
     }
+    if (err instanceof Session.BusyError) {
+      return c.json(new NamedError.Unknown({ message: err.message }).toObject(), { status: 400 })
+    }
     if (err instanceof HTTPException) return err.getResponse()
     const message = err instanceof Error && err.stack ? err.stack : err.toString()
     return c.json(new NamedError.Unknown({ message }).toObject(), {

packages/opencode/src/server/routes/session.ts

@@ -381,7 +381,7 @@ export const SessionRoutes = lazy(() =>
         }),
       ),
       async (c) => {
-        SessionPrompt.cancel(c.req.valid("param").sessionID)
+        await SessionPrompt.cancel(c.req.valid("param").sessionID)
         return c.json(true)
       },
     )
@@ -699,7 +699,7 @@ export const SessionRoutes = lazy(() =>
       ),
       async (c) => {
         const params = c.req.valid("param")
-        SessionPrompt.assertNotBusy(params.sessionID)
+        await SessionPrompt.assertNotBusy(params.sessionID)
         await Session.removeMessage({
           sessionID: params.sessionID,
           messageID: params.messageID,

packages/opencode/src/session/compaction.ts

@@ -15,7 +15,7 @@ import { Plugin } from "@/plugin"
 import { Config } from "@/config/config"
 import { NotFoundError } from "@/storage/db"
 import { ModelID, ProviderID } from "@/provider/schema"
-import { Cause, Effect, Exit, Layer, ServiceMap } from "effect"
+import { Effect, Layer, ServiceMap } from "effect"
 import { makeRuntime } from "@/effect/run-service"
 import { isOverflow as overflow } from "./overflow"
 
@@ -45,7 +45,6 @@ export namespace SessionCompaction {
       parentID: MessageID
       messages: MessageV2.WithParts[]
       sessionID: SessionID
-      abort: AbortSignal
       auto: boolean
       overflow?: boolean
     }) => Effect.Effect<"continue" | "stop">
@@ -135,20 +134,28 @@ export namespace SessionCompaction {
         parentID: MessageID
         messages: MessageV2.WithParts[]
         sessionID: SessionID
-        abort: AbortSignal
         auto: boolean
         overflow?: boolean
       }) {
-        const userMessage = input.messages.findLast((m) => m.info.id === input.parentID)!.info as MessageV2.User
+        const parent = input.messages.findLast((m) => m.info.id === input.parentID)
+        if (!parent || parent.info.role !== "user") {
+          throw new Error(`Compaction parent must be a user message: ${input.parentID}`)
+        }
+        const userMessage = parent.info
 
         let messages = input.messages
-        let replay: MessageV2.WithParts | undefined
+        let replay:
+          | {
+              info: MessageV2.User
+              parts: MessageV2.Part[]
+            }
+          | undefined
         if (input.overflow) {
           const idx = input.messages.findIndex((m) => m.info.id === input.parentID)
           for (let i = idx - 1; i >= 0; i--) {
             const msg = input.messages[i]
             if (msg.info.role === "user" && !msg.parts.some((p) => p.type === "compaction")) {
-              replay = msg
+              replay = { info: msg.info, parts: msg.parts }
               messages = input.messages.slice(0, i)
               break
             }
@@ -206,7 +213,7 @@ When constructing the summary, try to stick to this template:
         const msgs = structuredClone(messages)
         yield* plugin.trigger("experimental.chat.messages.transform", {}, { messages: msgs })
         const modelMessages = yield* Effect.promise(() => MessageV2.toModelMessages(msgs, model, { stripMedia: true }))
-        const msg = (yield* session.updateMessage({
+        const msg: MessageV2.Assistant = {
           id: MessageID.ascending(),
           role: "assistant",
           parentID: input.parentID,
@@ -231,25 +238,17 @@ When constructing the summary, try to stick to this template:
           time: {
             created: Date.now(),
           },
-        })) as MessageV2.Assistant
+        }
+        yield* session.updateMessage(msg)
         const processor = yield* processors.create({
           assistantMessage: msg,
           sessionID: input.sessionID,
           model,
-          abort: input.abort,
-        })
-        const cancel = Effect.fn("SessionCompaction.cancel")(function* () {
-          if (!input.abort.aborted || msg.time.completed) return
-          msg.error = msg.error ?? new MessageV2.AbortedError({ message: "Aborted" }).toObject()
-          msg.finish = msg.finish ?? "error"
-          msg.time.completed = Date.now()
-          yield* session.updateMessage(msg)
         })
         const result = yield* processor
           .process({
             user: userMessage,
             agent,
-            abort: input.abort,
             sessionID: input.sessionID,
             tools: {},
             system: [],
@@ -262,7 +261,7 @@ When constructing the summary, try to stick to this template:
             ],
             model,
           })
-          .pipe(Effect.ensuring(cancel()))
+          .pipe(Effect.onInterrupt(() => processor.abort()))
 
         if (result === "compact") {
           processor.message.error = new MessageV2.ContextOverflowError({
@@ -277,7 +276,7 @@ When constructing the summary, try to stick to this template:
 
         if (result === "continue" && input.auto) {
           if (replay) {
-            const original = replay.info as MessageV2.User
+            const original = replay.info
             const replayMsg = yield* session.updateMessage({
               id: MessageID.ascending(),
               role: "user",
@@ -386,7 +385,7 @@ When constructing the summary, try to stick to this template:
     ),
   )
 
-  const { runPromise, runPromiseExit } = makeRuntime(Service, defaultLayer)
+  const { runPromise } = makeRuntime(Service, defaultLayer)
 
   export async function isOverflow(input: { tokens: MessageV2.Assistant["tokens"]; model: Provider.Model }) {
     return runPromise((svc) => svc.isOverflow(input))
@@ -396,21 +395,16 @@ When constructing the summary, try to stick to this template:
     return runPromise((svc) => svc.prune(input))
   }
 
-  export async function process(input: {
-    parentID: MessageID
-    messages: MessageV2.WithParts[]
-    sessionID: SessionID
-    abort: AbortSignal
-    auto: boolean
-    overflow?: boolean
-  }) {
-    const exit = await runPromiseExit((svc) => svc.process(input), { signal: input.abort })
-    if (Exit.isFailure(exit)) {
-      if (Cause.hasInterrupts(exit.cause) && input.abort.aborted) return "stop"
-      throw Cause.squash(exit.cause)
-    }
-    return exit.value
-  }
+  export const process = fn(
+    z.object({
+      parentID: MessageID.zod,
+      messages: z.custom<MessageV2.WithParts[]>(),
+      sessionID: SessionID.zod,
+      auto: z.boolean(),
+      overflow: z.boolean().optional(),
+    }),
+    (input) => runPromise((svc) => svc.process(input)),
+  )
 
   export const create = fn(
     z.object({

packages/opencode/src/session/index.ts

@@ -334,14 +334,14 @@ export namespace Session {
     readonly messages: (input: { sessionID: SessionID; limit?: number }) => Effect.Effect<MessageV2.WithParts[]>
     readonly children: (parentID: SessionID) => Effect.Effect<Info[]>
     readonly remove: (sessionID: SessionID) => Effect.Effect<void>
-    readonly updateMessage: (msg: MessageV2.Info) => Effect.Effect<MessageV2.Info>
+    readonly updateMessage: <T extends MessageV2.Info>(msg: T) => Effect.Effect<T>
     readonly removeMessage: (input: { sessionID: SessionID; messageID: MessageID }) => Effect.Effect<MessageID>
     readonly removePart: (input: {
       sessionID: SessionID
       messageID: MessageID
       partID: PartID
     }) => Effect.Effect<PartID>
-    readonly updatePart: (part: MessageV2.Part) => Effect.Effect<MessageV2.Part>
+    readonly updatePart: <T extends MessageV2.Part>(part: T) => Effect.Effect<T>
     readonly updatePartDelta: (input: {
       sessionID: SessionID
       messageID: MessageID
@@ -469,26 +469,23 @@ export namespace Session {
         }
       })
 
-      const updateMessage = Effect.fn("Session.updateMessage")(function* (msg: MessageV2.Info) {
-        yield* Effect.sync(() =>
-          SyncEvent.run(MessageV2.Event.Updated, {
-            sessionID: msg.sessionID,
-            info: msg,
-          }),
-        )
-        return msg
-      })
+      const updateMessage = <T extends MessageV2.Info>(msg: T): Effect.Effect<T> =>
+        Effect.gen(function* () {
+          yield* Effect.sync(() => SyncEvent.run(MessageV2.Event.Updated, { sessionID: msg.sessionID, info: msg }))
+          return msg
+        }).pipe(Effect.withSpan("Session.updateMessage"))
 
-      const updatePart = Effect.fn("Session.updatePart")(function* (part: MessageV2.Part) {
-        yield* Effect.sync(() =>
-          SyncEvent.run(MessageV2.Event.PartUpdated, {
-            sessionID: part.sessionID,
-            part: structuredClone(part),
-            time: Date.now(),
-          }),
-        )
-        return part
-      })
+      const updatePart = <T extends MessageV2.Part>(part: T): Effect.Effect<T> =>
+        Effect.gen(function* () {
+          yield* Effect.sync(() =>
+            SyncEvent.run(MessageV2.Event.PartUpdated, {
+              sessionID: part.sessionID,
+              part: structuredClone(part),
+              time: Date.now(),
+            }),
+          )
+          return part
+        }).pipe(Effect.withSpan("Session.updatePart"))
 
       const create = Effect.fn("Session.create")(function* (input?: {
         parentID?: SessionID
@@ -851,7 +848,10 @@ export namespace Session {
 
   export const children = fn(SessionID.zod, (id) => runPromise((svc) => svc.children(id)))
   export const remove = fn(SessionID.zod, (id) => runPromise((svc) => svc.remove(id)))
-  export const updateMessage = fn(MessageV2.Info, (msg) => runPromise((svc) => svc.updateMessage(msg)))
+  export async function updateMessage<T extends MessageV2.Info>(msg: T): Promise<T> {
+    MessageV2.Info.parse(msg)
+    return runPromise((svc) => svc.updateMessage(msg))
+  }
 
   export const removeMessage = fn(z.object({ sessionID: SessionID.zod, messageID: MessageID.zod }), (input) =>
     runPromise((svc) => svc.removeMessage(input)),
@@ -862,7 +862,10 @@ export namespace Session {
     (input) => runPromise((svc) => svc.removePart(input)),
   )
 
-  export const updatePart = fn(MessageV2.Part, (part) => runPromise((svc) => svc.updatePart(part)))
+  export async function updatePart<T extends MessageV2.Part>(part: T): Promise<T> {
+    MessageV2.Part.parse(part)
+    return runPromise((svc) => svc.updatePart(part))
+  }
 
   export const updatePartDelta = fn(
     z.object({

packages/opencode/src/session/llm.ts

@@ -1,6 +1,7 @@
 import { Provider } from "@/provider/provider"
 import { Log } from "@/util/log"
-import { Effect, Layer, ServiceMap } from "effect"
+import { Cause, Effect, Layer, Record, ServiceMap } from "effect"
+import * as Queue from "effect/Queue"
 import * as Stream from "effect/Stream"
 import { streamText, wrapLanguageModel, type ModelMessage, type Tool, tool, jsonSchema } from "ai"
 import { mergeDeep, pipe } from "remeda"
@@ -28,7 +29,6 @@ export namespace LLM {
     agent: Agent.Info
     permission?: Permission.Ruleset
     system: string[]
-    abort: AbortSignal
     messages: ModelMessage[]
     small?: boolean
     tools: Record<string, Tool>
@@ -36,6 +36,10 @@ export namespace LLM {
     toolChoice?: "auto" | "required" | "none"
   }
 
+  export type StreamRequest = StreamInput & {
+    abort: AbortSignal
+  }
+
   export type Event = Awaited<ReturnType<typeof stream>>["fullStream"] extends AsyncIterable<infer T> ? T : never
 
   export interface Interface {
@@ -49,15 +53,32 @@ export namespace LLM {
     Effect.gen(function* () {
       return Service.of({
         stream(input) {
-          return Stream.unwrap(
-            Effect.promise(() => LLM.stream(input)).pipe(
-              Effect.map((result) =>
-                Stream.fromAsyncIterable(result.fullStream, (err) => err).pipe(
-                  Stream.mapEffect((event) => Effect.succeed(event)),
-                ),
-              ),
+          const stream: Stream.Stream<Event, unknown> = Stream.scoped(
+            Stream.unwrap(
+              Effect.gen(function* () {
+                const ctrl = yield* Effect.acquireRelease(
+                  Effect.sync(() => new AbortController()),
+                  (ctrl) => Effect.sync(() => ctrl.abort()),
+                )
+                const queue = yield* Queue.unbounded<Event, unknown | Cause.Done>()
+
+                yield* Effect.promise(async () => {
+                  const result = await LLM.stream({ ...input, abort: ctrl.signal })
+                  for await (const event of result.fullStream) {
+                    if (!Queue.offerUnsafe(queue, event)) break
+                  }
+                  Queue.endUnsafe(queue)
+                }).pipe(
+                  Effect.catchCause((cause) => Effect.sync(() => void Queue.failCauseUnsafe(queue, cause))),
+                  Effect.onInterrupt(() => Effect.sync(() => ctrl.abort())),
+                  Effect.forkScoped,
+                )
+
+                return Stream.fromQueue(queue)
+              }),
             ),
           )
+          return stream
         },
       })
     }),
@@ -65,7 +86,7 @@ export namespace LLM {
 
   export const defaultLayer = layer
 
-  export async function stream(input: StreamInput) {
+  export async function stream(input: StreamRequest) {
     const l = log
       .clone()
       .tag("providerID", input.model.providerID)
@@ -152,7 +173,7 @@ export namespace LLM {
       "chat.params",
       {
         sessionID: input.sessionID,
-        agent: input.agent,
+        agent: input.agent.name,
         model: input.model,
         provider,
         message: input.user,
@@ -171,7 +192,7 @@ export namespace LLM {
       "chat.headers",
       {
         sessionID: input.sessionID,
-        agent: input.agent,
+        agent: input.agent.name,
         model: input.model,
         provider,
         message: input.user,
@@ -322,17 +343,12 @@ export namespace LLM {
     })
   }
 
-  async function resolveTools(input: Pick<StreamInput, "tools" | "agent" | "permission" | "user">) {
+  function resolveTools(input: Pick<StreamInput, "tools" | "agent" | "permission" | "user">) {
     const disabled = Permission.disabled(
       Object.keys(input.tools),
       Permission.merge(input.agent.permission, input.permission ?? []),
     )
-    for (const tool of Object.keys(input.tools)) {
-      if (input.user.tools?.[tool] === false || disabled.has(tool)) {
-        delete input.tools[tool]
-      }
-    }
-    return input.tools
+    return Record.filter(input.tools, (_, k) => input.user.tools?.[k] !== false && !disabled.has(k))
   }
 
   // Check if messages contain any tool-call content

packages/opencode/src/session/processor.ts

@@ -1,8 +1,7 @@
-import { Cause, Effect, Exit, Layer, ServiceMap } from "effect"
+import { Cause, Effect, Layer, ServiceMap } from "effect"
 import * as Stream from "effect/Stream"
 import { Agent } from "@/agent/agent"
 import { Bus } from "@/bus"
-import { makeRuntime } from "@/effect/run-service"
 import { Config } from "@/config/config"
 import { Permission } from "@/permission"
 import { Plugin } from "@/plugin"
@@ -35,17 +34,10 @@ export namespace SessionProcessor {
     readonly process: (streamInput: LLM.StreamInput) => Effect.Effect<Result>
   }
 
-  export interface Info {
-    readonly message: MessageV2.Assistant
-    readonly partFromToolCall: (toolCallID: string) => MessageV2.ToolPart | undefined
-    readonly process: (streamInput: LLM.StreamInput) => Promise<Result>
-  }
-
   type Input = {
     assistantMessage: MessageV2.Assistant
     sessionID: SessionID
     model: Provider.Model
-    abort: AbortSignal
   }
 
   export interface Interface {
@@ -96,7 +88,6 @@ export namespace SessionProcessor {
           assistantMessage: input.assistantMessage,
           sessionID: input.sessionID,
           model: input.model,
-          abort: input.abort,
           toolcalls: {},
           shouldBreak: false,
           snapshot: undefined,
@@ -105,11 +96,12 @@ export namespace SessionProcessor {
           currentText: undefined,
           reasoningMap: {},
         }
+        let aborted = false
 
         const parse = (e: unknown) =>
           MessageV2.fromError(e, {
             providerID: input.model.providerID,
-            aborted: input.abort.aborted,
+            aborted,
           })
 
         const handleEvent = Effect.fn("SessionProcessor.handleEvent")(function* (value: StreamEvent) {
@@ -155,7 +147,10 @@ export namespace SessionProcessor {
               return
 
             case "tool-input-start":
-              ctx.toolcalls[value.id] = (yield* session.updatePart({
+              if (ctx.assistantMessage.summary) {
+                throw new Error(`Tool call not allowed while generating summary: ${value.toolName}`)
+              }
+              ctx.toolcalls[value.id] = yield* session.updatePart({
                 id: ctx.toolcalls[value.id]?.id ?? PartID.ascending(),
                 messageID: ctx.assistantMessage.id,
                 sessionID: ctx.assistantMessage.sessionID,
@@ -163,7 +158,7 @@ export namespace SessionProcessor {
                 tool: value.toolName,
                 callID: value.id,
                 state: { status: "pending", input: {}, raw: "" },
-              })) as MessageV2.ToolPart
+              } satisfies MessageV2.ToolPart)
               return
 
             case "tool-input-delta":
@@ -173,14 +168,17 @@ export namespace SessionProcessor {
               return
 
             case "tool-call": {
+              if (ctx.assistantMessage.summary) {
+                throw new Error(`Tool call not allowed while generating summary: ${value.toolName}`)
+              }
               const match = ctx.toolcalls[value.toolCallId]
               if (!match) return
-              ctx.toolcalls[value.toolCallId] = (yield* session.updatePart({
+              ctx.toolcalls[value.toolCallId] = yield* session.updatePart({
                 ...match,
                 tool: value.toolName,
                 state: { status: "running", input: value.input, time: { start: Date.now() } },
                 metadata: value.providerMetadata,
-              })) as MessageV2.ToolPart
+              } satisfies MessageV2.ToolPart)
 
               const parts = yield* Effect.promise(() => MessageV2.parts(ctx.assistantMessage.id))
               const recentParts = parts.slice(-DOOM_LOOP_THRESHOLD)
@@ -414,7 +412,7 @@ export namespace SessionProcessor {
         })
 
         const halt = Effect.fn("SessionProcessor.halt")(function* (e: unknown) {
-          log.error("process", { error: e, stack: JSON.stringify((e as any)?.stack) })
+          log.error("process", { error: e, stack: e instanceof Error ? e.stack : undefined })
           const error = parse(e)
           if (MessageV2.ContextOverflowError.isInstance(error)) {
             ctx.needsCompaction = true
@@ -429,59 +427,6 @@ export namespace SessionProcessor {
           yield* status.set(ctx.sessionID, { type: "idle" })
         })
 
-        const process = Effect.fn("SessionProcessor.process")(function* (streamInput: LLM.StreamInput) {
-          log.info("process")
-          ctx.needsCompaction = false
-          ctx.shouldBreak = (yield* config.get()).experimental?.continue_loop_on_deny !== true
-
-          yield* Effect.gen(function* () {
-            ctx.currentText = undefined
-            ctx.reasoningMap = {}
-            const stream = llm.stream(streamInput)
-
-            yield* stream.pipe(
-              Stream.tap((event) =>
-                Effect.gen(function* () {
-                  input.abort.throwIfAborted()
-                  yield* handleEvent(event)
-                }),
-              ),
-              Stream.takeUntil(() => ctx.needsCompaction),
-              Stream.runDrain,
-            )
-          }).pipe(
-            Effect.catchCauseIf(
-              (cause) => !Cause.hasInterruptsOnly(cause),
-              (cause) => Effect.fail(Cause.squash(cause)),
-            ),
-            Effect.retry(
-              SessionRetry.policy({
-                parse,
-                set: (info) =>
-                  status.set(ctx.sessionID, {
-                    type: "retry",
-                    attempt: info.attempt,
-                    message: info.message,
-                    next: info.next,
-                  }),
-              }),
-            ),
-            Effect.catchCause((cause) =>
-              Cause.hasInterruptsOnly(cause)
-                ? halt(new DOMException("Aborted", "AbortError"))
-                : halt(Cause.squash(cause)),
-            ),
-            Effect.ensuring(cleanup()),
-          )
-
-          if (input.abort.aborted && !ctx.assistantMessage.error) {
-            yield* abort()
-          }
-          if (ctx.needsCompaction) return "compact"
-          if (ctx.blocked || ctx.assistantMessage.error || input.abort.aborted) return "stop"
-          return "continue"
-        })
-
         const abort = Effect.fn("SessionProcessor.abort")(() =>
           Effect.gen(function* () {
             if (!ctx.assistantMessage.error) {
@@ -495,6 +440,53 @@ export namespace SessionProcessor {
           }),
         )
 
+        const process = Effect.fn("SessionProcessor.process")(function* (streamInput: LLM.StreamInput) {
+          log.info("process")
+          ctx.needsCompaction = false
+          ctx.shouldBreak = (yield* config.get()).experimental?.continue_loop_on_deny !== true
+
+          return yield* Effect.gen(function* () {
+            yield* Effect.gen(function* () {
+              ctx.currentText = undefined
+              ctx.reasoningMap = {}
+              const stream = llm.stream(streamInput)
+
+              yield* stream.pipe(
+                Stream.tap((event) => handleEvent(event)),
+                Stream.takeUntil(() => ctx.needsCompaction),
+                Stream.runDrain,
+              )
+            }).pipe(
+              Effect.onInterrupt(() => Effect.sync(() => void (aborted = true))),
+              Effect.catchCauseIf(
+                (cause) => !Cause.hasInterruptsOnly(cause),
+                (cause) => Effect.fail(Cause.squash(cause)),
+              ),
+              Effect.retry(
+                SessionRetry.policy({
+                  parse,
+                  set: (info) =>
+                    status.set(ctx.sessionID, {
+                      type: "retry",
+                      attempt: info.attempt,
+                      message: info.message,
+                      next: info.next,
+                    }),
+                }),
+              ),
+              Effect.catch(halt),
+              Effect.ensuring(cleanup()),
+            )
+
+            if (aborted && !ctx.assistantMessage.error) {
+              yield* abort()
+            }
+            if (ctx.needsCompaction) return "compact"
+            if (ctx.blocked || ctx.assistantMessage.error || aborted) return "stop"
+            return "continue"
+          }).pipe(Effect.onInterrupt(() => abort().pipe(Effect.asVoid)))
+        })
+
         return {
           get message() {
             return ctx.assistantMessage
@@ -526,29 +518,4 @@ export namespace SessionProcessor {
       ),
     ),
   )
-
-  const { runPromise } = makeRuntime(Service, defaultLayer)
-
-  export async function create(input: Input): Promise<Info> {
-    const hit = await runPromise((svc) => svc.create(input))
-    return {
-      get message() {
-        return hit.message
-      },
-      partFromToolCall(toolCallID: string) {
-        return hit.partFromToolCall(toolCallID)
-      },
-      async process(streamInput: LLM.StreamInput) {
-        const exit = await Effect.runPromiseExit(hit.process(streamInput), { signal: input.abort })
-        if (Exit.isFailure(exit)) {
-          if (Cause.hasInterrupts(exit.cause) && input.abort.aborted) {
-            await Effect.runPromise(hit.abort())
-            return "stop"
-          }
-          throw Cause.squash(exit.cause)
-        }
-        return exit.value
-      },
-    }
-  }
 }

packages/opencode/src/session/revert.ts

@@ -21,7 +21,7 @@ export namespace SessionRevert {
   export type RevertInput = z.infer<typeof RevertInput>
 
   export async function revert(input: RevertInput) {
-    SessionPrompt.assertNotBusy(input.sessionID)
+    await SessionPrompt.assertNotBusy(input.sessionID)
     const all = await Session.messages({ sessionID: input.sessionID })
     let lastUser: MessageV2.User | undefined
     const session = await Session.get(input.sessionID)
@@ -80,7 +80,7 @@ export namespace SessionRevert {
 
   export async function unrevert(input: { sessionID: SessionID }) {
     log.info("unreverting", input)
-    SessionPrompt.assertNotBusy(input.sessionID)
+    await SessionPrompt.assertNotBusy(input.sessionID)
     const session = await Session.get(input.sessionID)
     if (!session.revert) return session
     if (session.revert.snapshot) await Snapshot.restore(session.revert.snapshot)
@@ -92,12 +92,10 @@ export namespace SessionRevert {
     const sessionID = session.id
     const msgs = await Session.messages({ sessionID })
     const messageID = session.revert.messageID
-    const preserve = [] as MessageV2.WithParts[]
     const remove = [] as MessageV2.WithParts[]
     let target: MessageV2.WithParts | undefined
     for (const msg of msgs) {
       if (msg.info.id < messageID) {
-        preserve.push(msg)
         continue
       }
       if (msg.info.id > messageID) {
@@ -105,7 +103,6 @@ export namespace SessionRevert {
         continue
       }
       if (session.revert.partID) {
-        preserve.push(msg)
         target = msg
         continue
       }

packages/opencode/src/session/summary.ts

@@ -3,7 +3,6 @@ import z from "zod"
 import { Session } from "."
 
 import { MessageV2 } from "./message-v2"
-import { Identifier } from "@/id/id"
 import { SessionID, MessageID } from "./schema"
 import { Snapshot } from "@/snapshot"
 
@@ -110,8 +109,8 @@ export namespace SessionSummary {
       (m) => m.info.id === input.messageID || (m.info.role === "assistant" && m.info.parentID === input.messageID),
     )
     const msgWithParts = messages.find((m) => m.info.id === input.messageID)
-    if (!msgWithParts) return
-    const userMsg = msgWithParts.info as MessageV2.User
+    if (!msgWithParts || msgWithParts.info.role !== "user") return
+    const userMsg = msgWithParts.info
     const diffs = await computeDiff({ messages })
     userMsg.summary = {
       ...userMsg.summary,

packages/opencode/src/skill/index.ts

@@ -11,7 +11,7 @@ import { makeRuntime } from "@/effect/run-service"
 import { Flag } from "@/flag/flag"
 import { Global } from "@/global"
 import { Permission } from "@/permission"
-import { Filesystem } from "@/util/filesystem"
+import { AppFileSystem } from "@/filesystem"
 import { Config } from "../config/config"
 import { ConfigMarkdown } from "../config/markdown"
 import { Glob } from "../util/glob"
@@ -139,28 +139,20 @@ export namespace Skill {
     config: Config.Interface,
     discovery: Discovery.Interface,
     bus: Bus.Interface,
+    fsys: AppFileSystem.Interface,
     directory: string,
     worktree: string,
   ) {
     if (!Flag.OPENCODE_DISABLE_EXTERNAL_SKILLS) {
       for (const dir of EXTERNAL_DIRS) {
         const root = path.join(Global.Path.home, dir)
-        const isDir = yield* Effect.promise(() => Filesystem.isDir(root))
-        if (!isDir) continue
+        if (!(yield* fsys.isDir(root))) continue
         yield* scan(state, bus, root, EXTERNAL_SKILL_PATTERN, { dot: true, scope: "global" })
       }
 
-      const upDirs = yield* Effect.promise(async () => {
-        const dirs: string[] = []
-        for await (const root of Filesystem.up({
-          targets: EXTERNAL_DIRS,
-          start: directory,
-          stop: worktree,
-        })) {
-          dirs.push(root)
-        }
-        return dirs
-      })
+      const upDirs = yield* fsys
+        .up({ targets: EXTERNAL_DIRS, start: directory, stop: worktree })
+        .pipe(Effect.catch(() => Effect.succeed([] as string[])))
 
       for (const root of upDirs) {
         yield* scan(state, bus, root, EXTERNAL_SKILL_PATTERN, { dot: true, scope: "project" })
@@ -176,8 +168,7 @@ export namespace Skill {
     for (const item of cfg.skills?.paths ?? []) {
       const expanded = item.startsWith("~/") ? path.join(os.homedir(), item.slice(2)) : item
       const dir = path.isAbsolute(expanded) ? expanded : path.join(directory, expanded)
-      const isDir = yield* Effect.promise(() => Filesystem.isDir(dir))
-      if (!isDir) {
+      if (!(yield* fsys.isDir(dir))) {
         log.warn("skill path not found", { path: dir })
         continue
       }
@@ -198,16 +189,17 @@ export namespace Skill {
 
   export class Service extends ServiceMap.Service<Service, Interface>()("@opencode/Skill") {}
 
-  export const layer: Layer.Layer<Service, never, Discovery.Service | Config.Service | Bus.Service> = Layer.effect(
+  export const layer = Layer.effect(
     Service,
     Effect.gen(function* () {
       const discovery = yield* Discovery.Service
       const config = yield* Config.Service
       const bus = yield* Bus.Service
+      const fsys = yield* AppFileSystem.Service
       const state = yield* InstanceState.make(
         Effect.fn("Skill.state")(function* (ctx) {
           const s: State = { skills: {}, dirs: new Set() }
-          yield* loadSkills(s, config, discovery, bus, ctx.directory, ctx.worktree)
+          yield* loadSkills(s, config, discovery, bus, fsys, ctx.directory, ctx.worktree)
           return s
         }),
       )
@@ -238,10 +230,11 @@ export namespace Skill {
     }),
   )
 
-  export const defaultLayer: Layer.Layer<Service> = layer.pipe(
+  export const defaultLayer = layer.pipe(
     Layer.provide(Discovery.defaultLayer),
     Layer.provide(Config.defaultLayer),
     Layer.provide(Bus.layer),
+    Layer.provide(AppFileSystem.defaultLayer),
   )
 
   export function fmt(list: Info[], opts: { verbose: boolean }) {

packages/opencode/src/storage/storage.ts

@@ -1,19 +1,17 @@
 import { Log } from "../util/log"
 import path from "path"
-import fs from "fs/promises"
 import { Global } from "../global"
-import { Filesystem } from "../util/filesystem"
-import { lazy } from "../util/lazy"
-import { Lock } from "../util/lock"
 import { NamedError } from "@opencode-ai/util/error"
 import z from "zod"
-import { Glob } from "../util/glob"
 import { git } from "@/util/git"
+import { AppFileSystem } from "@/filesystem"
+import { makeRuntime } from "@/effect/run-service"
+import { Effect, Exit, Layer, Option, RcMap, Schema, ServiceMap, TxReentrantLock } from "effect"
 
 export namespace Storage {
   const log = Log.create({ service: "storage" })
 
-  type Migration = (dir: string) => Promise<void>
+  type Migration = (dir: string, fs: AppFileSystem.Interface) => Effect.Effect<void, AppFileSystem.Error>
 
   export const NotFoundError = NamedError.create(
     "NotFoundError",
@@ -22,36 +20,101 @@ export namespace Storage {
     }),
   )
 
+  export type Error = AppFileSystem.Error | InstanceType<typeof NotFoundError>
+
+  const RootFile = Schema.Struct({
+    path: Schema.optional(
+      Schema.Struct({
+        root: Schema.optional(Schema.String),
+      }),
+    ),
+  })
+
+  const SessionFile = Schema.Struct({
+    id: Schema.String,
+  })
+
+  const MessageFile = Schema.Struct({
+    id: Schema.String,
+  })
+
+  const DiffFile = Schema.Struct({
+    additions: Schema.Number,
+    deletions: Schema.Number,
+  })
+
+  const SummaryFile = Schema.Struct({
+    id: Schema.String,
+    projectID: Schema.String,
+    summary: Schema.Struct({ diffs: Schema.Array(DiffFile) }),
+  })
+
+  const decodeRoot = Schema.decodeUnknownOption(RootFile)
+  const decodeSession = Schema.decodeUnknownOption(SessionFile)
+  const decodeMessage = Schema.decodeUnknownOption(MessageFile)
+  const decodeSummary = Schema.decodeUnknownOption(SummaryFile)
+
+  export interface Interface {
+    readonly remove: (key: string[]) => Effect.Effect<void, AppFileSystem.Error>
+    readonly read: <T>(key: string[]) => Effect.Effect<T, Error>
+    readonly update: <T>(key: string[], fn: (draft: T) => void) => Effect.Effect<T, Error>
+    readonly write: <T>(key: string[], content: T) => Effect.Effect<void, AppFileSystem.Error>
+    readonly list: (prefix: string[]) => Effect.Effect<string[][], AppFileSystem.Error>
+  }
+
+  export class Service extends ServiceMap.Service<Service, Interface>()("@opencode/Storage") {}
+
+  function file(dir: string, key: string[]) {
+    return path.join(dir, ...key) + ".json"
+  }
+
+  function missing(err: unknown) {
+    if (!err || typeof err !== "object") return false
+    if ("code" in err && err.code === "ENOENT") return true
+    if ("reason" in err && err.reason && typeof err.reason === "object" && "_tag" in err.reason) {
+      return err.reason._tag === "NotFound"
+    }
+    return false
+  }
+
+  function parseMigration(text: string) {
+    const value = Number.parseInt(text, 10)
+    return Number.isNaN(value) ? 0 : value
+  }
+
   const MIGRATIONS: Migration[] = [
-    async (dir) => {
+    Effect.fn("Storage.migration.1")(function* (dir: string, fs: AppFileSystem.Interface) {
       const project = path.resolve(dir, "../project")
-      if (!(await Filesystem.isDir(project))) return
-      const projectDirs = await Glob.scan("*", {
+      if (!(yield* fs.isDir(project))) return
+      const projectDirs = yield* fs.glob("*", {
         cwd: project,
         include: "all",
       })
       for (const projectDir of projectDirs) {
-        const fullPath = path.join(project, projectDir)
-        if (!(await Filesystem.isDir(fullPath))) continue
+        const full = path.join(project, projectDir)
+        if (!(yield* fs.isDir(full))) continue
         log.info(`migrating project ${projectDir}`)
         let projectID = projectDir
-        const fullProjectDir = path.join(project, projectDir)
         let worktree = "/"
 
         if (projectID !== "global") {
-          for (const msgFile of await Glob.scan("storage/session/message/*/*.json", {
-            cwd: path.join(project, projectDir),
+          for (const msgFile of yield* fs.glob("storage/session/message/*/*.json", {
+            cwd: full,
             absolute: true,
           })) {
-            const json = await Filesystem.readJson<any>(msgFile)
-            worktree = json.path?.root
-            if (worktree) break
+            const json = decodeRoot(yield* fs.readJson(msgFile), { onExcessProperty: "preserve" })
+            const root = Option.isSome(json) ? json.value.path?.root : undefined
+            if (!root) continue
+            worktree = root
+            break
           }
           if (!worktree) continue
-          if (!(await Filesystem.isDir(worktree))) continue
-          const result = await git(["rev-list", "--max-parents=0", "--all"], {
-            cwd: worktree,
-          })
+          if (!(yield* fs.isDir(worktree))) continue
+          const result = yield* Effect.promise(() =>
+            git(["rev-list", "--max-parents=0", "--all"], {
+              cwd: worktree,
+            }),
+          )
           const [id] = result
             .text()
             .split("\n")
@@ -61,157 +124,230 @@ export namespace Storage {
           if (!id) continue
           projectID = id
 
-          await Filesystem.writeJson(path.join(dir, "project", projectID + ".json"), {
-            id,
-            vcs: "git",
-            worktree,
-            time: {
-              created: Date.now(),
-              initialized: Date.now(),
-            },
-          })
+          yield* fs.writeWithDirs(
+            path.join(dir, "project", projectID + ".json"),
+            JSON.stringify(
+              {
+                id,
+                vcs: "git",
+                worktree,
+                time: {
+                  created: Date.now(),
+                  initialized: Date.now(),
+                },
+              },
+              null,
+              2,
+            ),
+          )
 
           log.info(`migrating sessions for project ${projectID}`)
-          for (const sessionFile of await Glob.scan("storage/session/info/*.json", {
-            cwd: fullProjectDir,
+          for (const sessionFile of yield* fs.glob("storage/session/info/*.json", {
+            cwd: full,
             absolute: true,
           })) {
             const dest = path.join(dir, "session", projectID, path.basename(sessionFile))
-            log.info("copying", {
-              sessionFile,
-              dest,
-            })
-            const session = await Filesystem.readJson<any>(sessionFile)
-            await Filesystem.writeJson(dest, session)
-            log.info(`migrating messages for session ${session.id}`)
-            for (const msgFile of await Glob.scan(`storage/session/message/${session.id}/*.json`, {
-              cwd: fullProjectDir,
+            log.info("copying", { sessionFile, dest })
+            const session = yield* fs.readJson(sessionFile)
+            const info = decodeSession(session, { onExcessProperty: "preserve" })
+            yield* fs.writeWithDirs(dest, JSON.stringify(session, null, 2))
+            if (Option.isNone(info)) continue
+            log.info(`migrating messages for session ${info.value.id}`)
+            for (const msgFile of yield* fs.glob(`storage/session/message/${info.value.id}/*.json`, {
+              cwd: full,
               absolute: true,
             })) {
-              const dest = path.join(dir, "message", session.id, path.basename(msgFile))
+              const next = path.join(dir, "message", info.value.id, path.basename(msgFile))
               log.info("copying", {
                 msgFile,
-                dest,
+                dest: next,
               })
-              const message = await Filesystem.readJson<any>(msgFile)
-              await Filesystem.writeJson(dest, message)
+              const message = yield* fs.readJson(msgFile)
+              const item = decodeMessage(message, { onExcessProperty: "preserve" })
+              yield* fs.writeWithDirs(next, JSON.stringify(message, null, 2))
+              if (Option.isNone(item)) continue
 
-              log.info(`migrating parts for message ${message.id}`)
-              for (const partFile of await Glob.scan(`storage/session/part/${session.id}/${message.id}/*.json`, {
-                cwd: fullProjectDir,
+              log.info(`migrating parts for message ${item.value.id}`)
+              for (const partFile of yield* fs.glob(`storage/session/part/${info.value.id}/${item.value.id}/*.json`, {
+                cwd: full,
                 absolute: true,
               })) {
-                const dest = path.join(dir, "part", message.id, path.basename(partFile))
-                const part = await Filesystem.readJson(partFile)
+                const out = path.join(dir, "part", item.value.id, path.basename(partFile))
+                const part = yield* fs.readJson(partFile)
                 log.info("copying", {
                   partFile,
-                  dest,
+                  dest: out,
                 })
-                await Filesystem.writeJson(dest, part)
+                yield* fs.writeWithDirs(out, JSON.stringify(part, null, 2))
               }
             }
           }
         }
       }
-    },
-    async (dir) => {
-      for (const item of await Glob.scan("session/*/*.json", {
+    }),
+    Effect.fn("Storage.migration.2")(function* (dir: string, fs: AppFileSystem.Interface) {
+      for (const item of yield* fs.glob("session/*/*.json", {
         cwd: dir,
         absolute: true,
       })) {
-        const session = await Filesystem.readJson<any>(item)
-        if (!session.projectID) continue
-        if (!session.summary?.diffs) continue
-        const { diffs } = session.summary
-        await Filesystem.write(path.join(dir, "session_diff", session.id + ".json"), JSON.stringify(diffs))
-        await Filesystem.writeJson(path.join(dir, "session", session.projectID, session.id + ".json"), {
-          ...session,
-          summary: {
-            additions: diffs.reduce((sum: any, x: any) => sum + x.additions, 0),
-            deletions: diffs.reduce((sum: any, x: any) => sum + x.deletions, 0),
-          },
-        })
+        const raw = yield* fs.readJson(item)
+        const session = decodeSummary(raw, { onExcessProperty: "preserve" })
+        if (Option.isNone(session)) continue
+        const diffs = session.value.summary.diffs
+        yield* fs.writeWithDirs(
+          path.join(dir, "session_diff", session.value.id + ".json"),
+          JSON.stringify(diffs, null, 2),
+        )
+        yield* fs.writeWithDirs(
+          path.join(dir, "session", session.value.projectID, session.value.id + ".json"),
+          JSON.stringify(
+            {
+              ...(raw as Record<string, unknown>),
+              summary: {
+                additions: diffs.reduce((sum, x) => sum + x.additions, 0),
+                deletions: diffs.reduce((sum, x) => sum + x.deletions, 0),
+              },
+            },
+            null,
+            2,
+          ),
+        )
       }
-    },
+    }),
   ]
 
-  const state = lazy(async () => {
-    const dir = path.join(Global.Path.data, "storage")
-    const migration = await Filesystem.readJson<string>(path.join(dir, "migration"))
-      .then((x) => parseInt(x))
-      .catch(() => 0)
-    for (let index = migration; index < MIGRATIONS.length; index++) {
-      log.info("running migration", { index })
-      const migration = MIGRATIONS[index]
-      await migration(dir).catch(() => log.error("failed to run migration", { index }))
-      await Filesystem.write(path.join(dir, "migration"), (index + 1).toString())
-    }
-    return {
-      dir,
-    }
-  })
+  export const layer = Layer.effect(
+    Service,
+    Effect.gen(function* () {
+      const fs = yield* AppFileSystem.Service
+      const locks = yield* RcMap.make({
+        lookup: () => TxReentrantLock.make(),
+        idleTimeToLive: 0,
+      })
+      const state = yield* Effect.cached(
+        Effect.gen(function* () {
+          const dir = path.join(Global.Path.data, "storage")
+          const marker = path.join(dir, "migration")
+          const migration = yield* fs.readFileString(marker).pipe(
+            Effect.map(parseMigration),
+            Effect.catchIf(missing, () => Effect.succeed(0)),
+            Effect.orElseSucceed(() => 0),
+          )
+          for (let i = migration; i < MIGRATIONS.length; i++) {
+            log.info("running migration", { index: i })
+            const step = MIGRATIONS[i]!
+            const exit = yield* Effect.exit(step(dir, fs))
+            if (Exit.isFailure(exit)) {
+              log.error("failed to run migration", { index: i, cause: exit.cause })
+              break
+            }
+            yield* fs.writeWithDirs(marker, String(i + 1))
+          }
+          return { dir }
+        }),
+      )
+
+      const fail = (target: string): Effect.Effect<never, InstanceType<typeof NotFoundError>> =>
+        Effect.fail(new NotFoundError({ message: `Resource not found: ${target}` }))
+
+      const wrap = <A>(target: string, body: Effect.Effect<A, AppFileSystem.Error>) =>
+        body.pipe(Effect.catchIf(missing, () => fail(target)))
+
+      const writeJson = Effect.fnUntraced(function* (target: string, content: unknown) {
+        yield* fs.writeWithDirs(target, JSON.stringify(content, null, 2))
+      })
+
+      const withResolved = <A, E>(
+        key: string[],
+        fn: (target: string, rw: TxReentrantLock.TxReentrantLock) => Effect.Effect<A, E>,
+      ): Effect.Effect<A, E | AppFileSystem.Error> =>
+        Effect.scoped(
+          Effect.gen(function* () {
+            const target = file((yield* state).dir, key)
+            return yield* fn(target, yield* RcMap.get(locks, target))
+          }),
+        )
+
+      const remove: Interface["remove"] = Effect.fn("Storage.remove")(function* (key: string[]) {
+        yield* withResolved(key, (target, rw) =>
+          TxReentrantLock.withWriteLock(rw, fs.remove(target).pipe(Effect.catchIf(missing, () => Effect.void))),
+        )
+      })
+
+      const read: Interface["read"] = <T>(key: string[]) =>
+        Effect.gen(function* () {
+          const value = yield* withResolved(key, (target, rw) =>
+            TxReentrantLock.withReadLock(rw, wrap(target, fs.readJson(target))),
+          )
+          return value as T
+        })
+
+      const update: Interface["update"] = <T>(key: string[], fn: (draft: T) => void) =>
+        Effect.gen(function* () {
+          const value = yield* withResolved(key, (target, rw) =>
+            TxReentrantLock.withWriteLock(
+              rw,
+              Effect.gen(function* () {
+                const content = yield* wrap(target, fs.readJson(target))
+                fn(content as T)
+                yield* writeJson(target, content)
+                return content
+              }),
+            ),
+          )
+          return value as T
+        })
+
+      const write: Interface["write"] = (key: string[], content: unknown) =>
+        Effect.gen(function* () {
+          yield* withResolved(key, (target, rw) => TxReentrantLock.withWriteLock(rw, writeJson(target, content)))
+        })
+
+      const list: Interface["list"] = Effect.fn("Storage.list")(function* (prefix: string[]) {
+        const dir = (yield* state).dir
+        const cwd = path.join(dir, ...prefix)
+        const result = yield* fs
+          .glob("**/*", {
+            cwd,
+            include: "file",
+          })
+          .pipe(Effect.catch(() => Effect.succeed<string[]>([])))
+        return result
+          .map((x) => [...prefix, ...x.slice(0, -5).split(path.sep)])
+          .toSorted((a, b) => a.join("/").localeCompare(b.join("/")))
+      })
+
+      return Service.of({
+        remove,
+        read,
+        update,
+        write,
+        list,
+      })
+    }),
+  )
+
+  export const defaultLayer = layer.pipe(Layer.provide(AppFileSystem.defaultLayer))
+
+  const { runPromise } = makeRuntime(Service, defaultLayer)
 
   export async function remove(key: string[]) {
-    const dir = await state().then((x) => x.dir)
-    const target = path.join(dir, ...key) + ".json"
-    return withErrorHandling(async () => {
-      await fs.unlink(target).catch(() => {})
-    })
+    return runPromise((svc) => svc.remove(key))
   }
 
   export async function read<T>(key: string[]) {
-    const dir = await state().then((x) => x.dir)
-    const target = path.join(dir, ...key) + ".json"
-    return withErrorHandling(async () => {
-      using _ = await Lock.read(target)
-      const result = await Filesystem.readJson<T>(target)
-      return result as T
-    })
+    return runPromise((svc) => svc.read<T>(key))
   }
 
   export async function update<T>(key: string[], fn: (draft: T) => void) {
-    const dir = await state().then((x) => x.dir)
-    const target = path.join(dir, ...key) + ".json"
-    return withErrorHandling(async () => {
-      using _ = await Lock.write(target)
-      const content = await Filesystem.readJson<T>(target)
-      fn(content as T)
-      await Filesystem.writeJson(target, content)
-      return content
-    })
+    return runPromise((svc) => svc.update<T>(key, fn))
   }
 
   export async function write<T>(key: string[], content: T) {
-    const dir = await state().then((x) => x.dir)
-    const target = path.join(dir, ...key) + ".json"
-    return withErrorHandling(async () => {
-      using _ = await Lock.write(target)
-      await Filesystem.writeJson(target, content)
-    })
-  }
-
-  async function withErrorHandling<T>(body: () => Promise<T>) {
-    return body().catch((e) => {
-      if (!(e instanceof Error)) throw e
-      const errnoException = e as NodeJS.ErrnoException
-      if (errnoException.code === "ENOENT") {
-        throw new NotFoundError({ message: `Resource not found: ${errnoException.path}` })
-      }
-      throw e
-    })
+    return runPromise((svc) => svc.write(key, content))
   }
 
   export async function list(prefix: string[]) {
-    const dir = await state().then((x) => x.dir)
-    try {
-      const result = await Glob.scan("**/*", {
-        cwd: path.join(dir, ...prefix),
-        include: "file",
-      }).then((results) => results.map((x) => [...prefix, ...x.slice(0, -5).split(path.sep)]))
-      result.sort()
-      return result
-    } catch {
-      return []
-    }
+    return runPromise((svc) => svc.list(prefix))
   }
 }

packages/opencode/src/tool/registry.ts

@@ -46,12 +46,12 @@ export namespace ToolRegistry {
     readonly tools: (
       model: { providerID: ProviderID; modelID: ModelID },
       agent?: Agent.Info,
-    ) => Effect.Effect<(Awaited<ReturnType<Tool.Info["init"]>> & { id: string })[]>
+    ) => Effect.Effect<(Tool.Def & { id: string })[]>
   }
 
   export class Service extends ServiceMap.Service<Service, Interface>()("@opencode/ToolRegistry") {}
 
-  export const layer = Layer.effect(
+  export const layer: Layer.Layer<Service, never, Config.Service | Plugin.Service> = Layer.effect(
     Service,
     Effect.gen(function* () {
       const config = yield* Config.Service
@@ -174,7 +174,7 @@ export namespace ToolRegistry {
         })
         return yield* Effect.forEach(
           filtered,
-          Effect.fnUntraced(function* (tool) {
+          Effect.fnUntraced(function* (tool: Tool.Info) {
             using _ = log.time(tool.id)
             const next = yield* Effect.promise(() => tool.init({ agent }))
             const output = {
@@ -184,10 +184,11 @@ export namespace ToolRegistry {
             yield* plugin.trigger("tool.definition", { toolID: tool.id }, output)
             return {
               id: tool.id,
-              ...next,
               description: output.description,
               parameters: output.parameters,
-            } as Awaited<ReturnType<Tool.Info["init"]>> & { id: string }
+              execute: next.execute,
+              formatValidationError: next.formatValidationError,
+            }
           }),
           { concurrency: "unbounded" },
         )
@@ -217,7 +218,7 @@ export namespace ToolRegistry {
       modelID: ModelID
     },
     agent?: Agent.Info,
-  ): Promise<(Awaited<ReturnType<Tool.Info["init"]>> & { id: string })[]> {
+  ): Promise<(Tool.Def & { id: string })[]> {
     return runPromise((svc) => svc.tools(model, agent))
   }
 }

packages/opencode/src/tool/tool.ts

@@ -25,22 +25,24 @@ export namespace Tool {
     metadata(input: { title?: string; metadata?: M }): void
     ask(input: Omit<Permission.Request, "id" | "sessionID" | "tool">): Promise<void>
   }
+  export interface Def<Parameters extends z.ZodType = z.ZodType, M extends Metadata = Metadata> {
+    description: string
+    parameters: Parameters
+    execute(
+      args: z.infer<Parameters>,
+      ctx: Context,
+    ): Promise<{
+      title: string
+      metadata: M
+      output: string
+      attachments?: Omit<MessageV2.FilePart, "id" | "sessionID" | "messageID">[]
+    }>
+    formatValidationError?(error: z.ZodError): string
+  }
+
   export interface Info<Parameters extends z.ZodType = z.ZodType, M extends Metadata = Metadata> {
     id: string
-    init: (ctx?: InitContext) => Promise<{
-      description: string
-      parameters: Parameters
-      execute(
-        args: z.infer<Parameters>,
-        ctx: Context,
-      ): Promise<{
-        title: string
-        metadata: M
-        output: string
-        attachments?: Omit<MessageV2.FilePart, "id" | "sessionID" | "messageID">[]
-      }>
-      formatValidationError?(error: z.ZodError): string
-    }>
+    init: (ctx?: InitContext) => Promise<Def<Parameters, M>>
   }
 
   export type InferParameters<T extends Info> = T extends Info<infer P> ? z.infer<P> : never
@@ -48,7 +50,7 @@ export namespace Tool {
 
   export function define<Parameters extends z.ZodType, Result extends Metadata>(
     id: string,
-    init: Info<Parameters, Result>["init"] | Awaited<ReturnType<Info<Parameters, Result>["init"]>>,
+    init: Info<Parameters, Result>["init"] | Def<Parameters, Result>,
   ): Info<Parameters, Result> {
     return {
       id,

packages/opencode/src/util/filesystem.ts

@@ -1,4 +1,4 @@
-import { chmod, mkdir, readFile, writeFile } from "fs/promises"
+import { chmod, mkdir, readFile, stat as statFile, writeFile } from "fs/promises"
 import { createWriteStream, existsSync, statSync } from "fs"
 import { lookup } from "mime-types"
 import { realpathSync } from "fs"
@@ -25,6 +25,13 @@ export namespace Filesystem {
     return statSync(p, { throwIfNoEntry: false }) ?? undefined
   }
 
+  export async function statAsync(p: string): Promise<ReturnType<typeof statSync> | undefined> {
+    return statFile(p).catch((e) => {
+      if (isEnoent(e)) return undefined
+      throw e
+    })
+  }
+
   export async function size(p: string): Promise<number> {
     const s = stat(p)?.size ?? 0
     return typeof s === "bigint" ? Number(s) : s

packages/opencode/test/cli/tui/plugin-add.test.ts

@@ -33,7 +33,7 @@ test("adds tui plugin at runtime from spec", async () => {
   process.env.OPENCODE_PLUGIN_META_FILE = path.join(tmp.path, "plugin-meta.json")
   const get = spyOn(TuiConfig, "get").mockResolvedValue({
     plugin: [],
-    plugin_meta: undefined,
+    plugin_records: undefined,
   })
   const wait = spyOn(TuiConfig, "waitForDependencies").mockResolvedValue()
   const cwd = spyOn(process, "cwd").mockImplementation(() => tmp.path)

packages/opencode/test/cli/tui/plugin-install.test.ts

@@ -48,7 +48,7 @@ test("installs plugin without loading it", async () => {
   process.env.OPENCODE_PLUGIN_META_FILE = path.join(tmp.path, "plugin-meta.json")
   let cfg: Awaited<ReturnType<typeof TuiConfig.get>> = {
     plugin: [],
-    plugin_meta: undefined,
+    plugin_records: undefined,
   }
   const get = spyOn(TuiConfig, "get").mockImplementation(async () => cfg)
   const wait = spyOn(TuiConfig, "waitForDependencies").mockResolvedValue()
@@ -68,12 +68,13 @@ test("installs plugin without loading it", async () => {
     await TuiPluginRuntime.init(api)
     cfg = {
       plugin: [[tmp.extra.spec, { marker: tmp.extra.marker }]],
-      plugin_meta: {
-        [tmp.extra.spec]: {
+      plugin_records: [
+        {
+          item: [tmp.extra.spec, { marker: tmp.extra.marker }],
           scope: "local",
           source: path.join(tmp.path, "tui.json"),
         },
-      },
+      ],
     }
 
     const out = await TuiPluginRuntime.installPlugin(tmp.extra.spec)

packages/opencode/test/cli/tui/plugin-loader-entrypoint.test.ts

@@ -1,6 +1,7 @@
 import { expect, spyOn, test } from "bun:test"
 import fs from "fs/promises"
 import path from "path"
+import { pathToFileURL } from "url"
 import { tmpdir } from "../../fixture/fixture"
 import { createTuiPluginApi } from "../../fixture/tui-plugin"
 import { TuiConfig } from "../../../src/config/tui"
@@ -45,9 +46,13 @@ test("loads npm tui plugin from package ./tui export", async () => {
   process.env.OPENCODE_PLUGIN_META_FILE = path.join(tmp.path, "plugin-meta.json")
   const get = spyOn(TuiConfig, "get").mockResolvedValue({
     plugin: [[tmp.extra.spec, { marker: tmp.extra.marker }]],
-    plugin_meta: {
-      [tmp.extra.spec]: { scope: "local", source: path.join(tmp.path, "tui.json") },
-    },
+    plugin_records: [
+      {
+        item: [tmp.extra.spec, { marker: tmp.extra.marker }],
+        scope: "local",
+        source: path.join(tmp.path, "tui.json"),
+      },
+    ],
   })
   const wait = spyOn(TuiConfig, "waitForDependencies").mockResolvedValue()
   const cwd = spyOn(process, "cwd").mockImplementation(() => tmp.path)
@@ -70,6 +75,65 @@ test("loads npm tui plugin from package ./tui export", async () => {
   }
 })
 
+test("does not use npm package exports dot for tui entry", async () => {
+  await using tmp = await tmpdir({
+    init: async (dir) => {
+      const mod = path.join(dir, "mods", "acme-plugin")
+      const marker = path.join(dir, "dot-called.txt")
+      await fs.mkdir(mod, { recursive: true })
+
+      await Bun.write(
+        path.join(mod, "package.json"),
+        JSON.stringify({
+          name: "acme-plugin",
+          type: "module",
+          exports: { ".": "./index.js" },
+        }),
+      )
+      await Bun.write(
+        path.join(mod, "index.js"),
+        `export default {
+  id: "demo.dot",
+  tui: async () => {
+    await Bun.write(${JSON.stringify(marker)}, "called")
+  },
+}
+`,
+      )
+
+      return { mod, marker, spec: "acme-plugin@1.0.0" }
+    },
+  })
+
+  process.env.OPENCODE_PLUGIN_META_FILE = path.join(tmp.path, "plugin-meta.json")
+  const get = spyOn(TuiConfig, "get").mockResolvedValue({
+    plugin: [tmp.extra.spec],
+    plugin_records: [
+      {
+        item: tmp.extra.spec,
+        scope: "local",
+        source: path.join(tmp.path, "tui.json"),
+      },
+    ],
+  })
+  const wait = spyOn(TuiConfig, "waitForDependencies").mockResolvedValue()
+  const cwd = spyOn(process, "cwd").mockImplementation(() => tmp.path)
+  const install = spyOn(BunProc, "install").mockResolvedValue(tmp.extra.mod)
+
+  try {
+    await TuiPluginRuntime.init(createTuiPluginApi())
+    await expect(fs.readFile(tmp.extra.marker, "utf8")).rejects.toThrow()
+    expect(TuiPluginRuntime.list().some((item) => item.spec === tmp.extra.spec)).toBe(false)
+  } finally {
+    await TuiPluginRuntime.dispose()
+    install.mockRestore()
+    cwd.mockRestore()
+    get.mockRestore()
+    wait.mockRestore()
+    delete process.env.OPENCODE_PLUGIN_META_FILE
+  }
+})
+
 test("rejects npm tui export that resolves outside plugin directory", async () => {
   await using tmp = await tmpdir({
     init: async (dir) => {
@@ -107,9 +171,13 @@ test("rejects npm tui export that resolves outside plugin directory", async () =
   process.env.OPENCODE_PLUGIN_META_FILE = path.join(tmp.path, "plugin-meta.json")
   const get = spyOn(TuiConfig, "get").mockResolvedValue({
     plugin: [tmp.extra.spec],
-    plugin_meta: {
-      [tmp.extra.spec]: { scope: "local", source: path.join(tmp.path, "tui.json") },
-    },
+    plugin_records: [
+      {
+        item: tmp.extra.spec,
+        scope: "local",
+        source: path.join(tmp.path, "tui.json"),
+      },
+    ],
   })
   const wait = spyOn(TuiConfig, "waitForDependencies").mockResolvedValue()
   const cwd = spyOn(process, "cwd").mockImplementation(() => tmp.path)
@@ -166,9 +234,13 @@ test("rejects npm tui plugin that exports server and tui together", async () =>
   process.env.OPENCODE_PLUGIN_META_FILE = path.join(tmp.path, "plugin-meta.json")
   const get = spyOn(TuiConfig, "get").mockResolvedValue({
     plugin: [tmp.extra.spec],
-    plugin_meta: {
-      [tmp.extra.spec]: { scope: "local", source: path.join(tmp.path, "tui.json") },
-    },
+    plugin_records: [
+      {
+        item: tmp.extra.spec,
+        scope: "local",
+        source: path.join(tmp.path, "tui.json"),
+      },
+    ],
   })
   const wait = spyOn(TuiConfig, "waitForDependencies").mockResolvedValue()
   const cwd = spyOn(process, "cwd").mockImplementation(() => tmp.path)
@@ -187,3 +259,228 @@ test("rejects npm tui plugin that exports server and tui together", async () =>
     delete process.env.OPENCODE_PLUGIN_META_FILE
   }
 })
+
+test("does not use npm package main for tui entry", async () => {
+  await using tmp = await tmpdir({
+    init: async (dir) => {
+      const mod = path.join(dir, "mods", "acme-plugin")
+      const marker = path.join(dir, "main-called.txt")
+      await fs.mkdir(mod, { recursive: true })
+
+      await Bun.write(
+        path.join(mod, "package.json"),
+        JSON.stringify({
+          name: "acme-plugin",
+          type: "module",
+          main: "./index.js",
+        }),
+      )
+      await Bun.write(
+        path.join(mod, "index.js"),
+        `export default {
+  id: "demo.main",
+  tui: async () => {
+    await Bun.write(${JSON.stringify(marker)}, "called")
+  },
+}
+`,
+      )
+
+      return { mod, marker, spec: "acme-plugin@1.0.0" }
+    },
+  })
+
+  process.env.OPENCODE_PLUGIN_META_FILE = path.join(tmp.path, "plugin-meta.json")
+  const get = spyOn(TuiConfig, "get").mockResolvedValue({
+    plugin: [tmp.extra.spec],
+    plugin_records: [
+      {
+        item: tmp.extra.spec,
+        scope: "local",
+        source: path.join(tmp.path, "tui.json"),
+      },
+    ],
+  })
+  const wait = spyOn(TuiConfig, "waitForDependencies").mockResolvedValue()
+  const cwd = spyOn(process, "cwd").mockImplementation(() => tmp.path)
+  const install = spyOn(BunProc, "install").mockResolvedValue(tmp.extra.mod)
+
+  try {
+    await TuiPluginRuntime.init(createTuiPluginApi())
+    await expect(fs.readFile(tmp.extra.marker, "utf8")).rejects.toThrow()
+    expect(TuiPluginRuntime.list().some((item) => item.spec === tmp.extra.spec)).toBe(false)
+  } finally {
+    await TuiPluginRuntime.dispose()
+    install.mockRestore()
+    cwd.mockRestore()
+    get.mockRestore()
+    wait.mockRestore()
+    delete process.env.OPENCODE_PLUGIN_META_FILE
+  }
+})
+
+test("does not use directory package main for tui entry", async () => {
+  await using tmp = await tmpdir({
+    init: async (dir) => {
+      const mod = path.join(dir, "mods", "dir-plugin")
+      const spec = pathToFileURL(mod).href
+      const marker = path.join(dir, "dir-main-called.txt")
+      await fs.mkdir(mod, { recursive: true })
+
+      await Bun.write(
+        path.join(mod, "package.json"),
+        JSON.stringify({
+          name: "dir-plugin",
+          type: "module",
+          main: "./main.js",
+        }),
+      )
+      await Bun.write(
+        path.join(mod, "main.js"),
+        `export default {
+  id: "demo.dir.main",
+  tui: async () => {
+    await Bun.write(${JSON.stringify(marker)}, "called")
+  },
+}
+`,
+      )
+
+      return { marker, spec }
+    },
+  })
+
+  process.env.OPENCODE_PLUGIN_META_FILE = path.join(tmp.path, "plugin-meta.json")
+  const get = spyOn(TuiConfig, "get").mockResolvedValue({
+    plugin: [tmp.extra.spec],
+    plugin_records: [
+      {
+        item: tmp.extra.spec,
+        scope: "local",
+        source: path.join(tmp.path, "tui.json"),
+      },
+    ],
+  })
+  const wait = spyOn(TuiConfig, "waitForDependencies").mockResolvedValue()
+  const cwd = spyOn(process, "cwd").mockImplementation(() => tmp.path)
+
+  try {
+    await TuiPluginRuntime.init(createTuiPluginApi())
+    await expect(fs.readFile(tmp.extra.marker, "utf8")).rejects.toThrow()
+    expect(TuiPluginRuntime.list().some((item) => item.spec === tmp.extra.spec)).toBe(false)
+  } finally {
+    await TuiPluginRuntime.dispose()
+    cwd.mockRestore()
+    get.mockRestore()
+    wait.mockRestore()
+    delete process.env.OPENCODE_PLUGIN_META_FILE
+  }
+})
+
+test("uses directory index fallback for tui when package.json is missing", async () => {
+  await using tmp = await tmpdir({
+    init: async (dir) => {
+      const mod = path.join(dir, "mods", "dir-index")
+      const spec = pathToFileURL(mod).href
+      const marker = path.join(dir, "dir-index-called.txt")
+      await fs.mkdir(mod, { recursive: true })
+      await Bun.write(
+        path.join(mod, "index.ts"),
+        `export default {
+  id: "demo.dir.index",
+  tui: async () => {
+    await Bun.write(${JSON.stringify(marker)}, "called")
+  },
+}
+`,
+      )
+      return { marker, spec }
+    },
+  })
+
+  process.env.OPENCODE_PLUGIN_META_FILE = path.join(tmp.path, "plugin-meta.json")
+  const get = spyOn(TuiConfig, "get").mockResolvedValue({
+    plugin: [tmp.extra.spec],
+    plugin_records: [
+      {
+        item: tmp.extra.spec,
+        scope: "local",
+        source: path.join(tmp.path, "tui.json"),
+      },
+    ],
+  })
+  const wait = spyOn(TuiConfig, "waitForDependencies").mockResolvedValue()
+  const cwd = spyOn(process, "cwd").mockImplementation(() => tmp.path)
+
+  try {
+    await TuiPluginRuntime.init(createTuiPluginApi())
+    await expect(fs.readFile(tmp.extra.marker, "utf8")).resolves.toBe("called")
+    expect(TuiPluginRuntime.list().find((item) => item.id === "demo.dir.index")?.active).toBe(true)
+  } finally {
+    await TuiPluginRuntime.dispose()
+    cwd.mockRestore()
+    get.mockRestore()
+    wait.mockRestore()
+    delete process.env.OPENCODE_PLUGIN_META_FILE
+  }
+})
+
+test("uses npm package name when tui plugin id is omitted", async () => {
+  await using tmp = await tmpdir({
+    init: async (dir) => {
+      const mod = path.join(dir, "mods", "acme-plugin")
+      const marker = path.join(dir, "name-id-called.txt")
+      await fs.mkdir(mod, { recursive: true })
+
+      await Bun.write(
+        path.join(mod, "package.json"),
+        JSON.stringify({
+          name: "acme-plugin",
+          type: "module",
+          exports: { ".": "./index.js", "./tui": "./tui.js" },
+        }),
+      )
+      await Bun.write(path.join(mod, "index.js"), "export default {}\n")
+      await Bun.write(
+        path.join(mod, "tui.js"),
+        `export default {
+  tui: async (_api, options) => {
+    if (!options?.marker) return
+    await Bun.write(options.marker, "called")
+  },
+}
+`,
+      )
+
+      return { mod, marker, spec: "acme-plugin@1.0.0" }
+    },
+  })
+
+  process.env.OPENCODE_PLUGIN_META_FILE = path.join(tmp.path, "plugin-meta.json")
+  const get = spyOn(TuiConfig, "get").mockResolvedValue({
+    plugin: [[tmp.extra.spec, { marker: tmp.extra.marker }]],
+    plugin_records: [
+      {
+        item: [tmp.extra.spec, { marker: tmp.extra.marker }],
+        scope: "local",
+        source: path.join(tmp.path, "tui.json"),
+      },
+    ],
+  })
+  const wait = spyOn(TuiConfig, "waitForDependencies").mockResolvedValue()
+  const cwd = spyOn(process, "cwd").mockImplementation(() => tmp.path)
+  const install = spyOn(BunProc, "install").mockResolvedValue(tmp.extra.mod)
+
+  try {
+    await TuiPluginRuntime.init(createTuiPluginApi())
+    await expect(fs.readFile(tmp.extra.marker, "utf8")).resolves.toBe("called")
+    expect(TuiPluginRuntime.list().find((item) => item.spec === tmp.extra.spec)?.id).toBe("acme-plugin")
+  } finally {
+    await TuiPluginRuntime.dispose()
+    install.mockRestore()
+    cwd.mockRestore()
+    get.mockRestore()
+    wait.mockRestore()
+    delete process.env.OPENCODE_PLUGIN_META_FILE
+  }
+})

packages/opencode/test/cli/tui/plugin-loader-pure.test.ts

@@ -39,12 +39,13 @@ test("skips external tui plugins in pure mode", async () => {
 
   const get = spyOn(TuiConfig, "get").mockResolvedValue({
     plugin: [[tmp.extra.spec, { marker: tmp.extra.marker }]],
-    plugin_meta: {
-      [tmp.extra.spec]: {
+    plugin_records: [
+      {
+        item: [tmp.extra.spec, { marker: tmp.extra.marker }],
         scope: "local",
         source: path.join(tmp.path, "tui.json"),
       },
-    },
+    ],
   })
   const wait = spyOn(TuiConfig, "waitForDependencies").mockResolvedValue()
   const cwd = spyOn(process, "cwd").mockImplementation(() => tmp.path)

packages/opencode/test/cli/tui/plugin-loader.test.ts

@@ -468,10 +468,18 @@ test("continues loading when a plugin is missing config metadata", async () => {
       [tmp.extra.goodSpec, { marker: tmp.extra.goodMarker }],
       tmp.extra.bareSpec,
     ],
-    plugin_meta: {
-      [tmp.extra.goodSpec]: { scope: "local", source: path.join(tmp.path, "tui.json") },
-      [tmp.extra.bareSpec]: { scope: "local", source: path.join(tmp.path, "tui.json") },
-    },
+    plugin_records: [
+      {
+        item: [tmp.extra.goodSpec, { marker: tmp.extra.goodMarker }],
+        scope: "local",
+        source: path.join(tmp.path, "tui.json"),
+      },
+      {
+        item: tmp.extra.bareSpec,
+        scope: "local",
+        source: path.join(tmp.path, "tui.json"),
+      },
+    ],
   })
   const wait = spyOn(TuiConfig, "waitForDependencies").mockResolvedValue()
   const cwd = spyOn(process, "cwd").mockImplementation(() => tmp.path)
@@ -493,6 +501,84 @@ test("continues loading when a plugin is missing config metadata", async () => {
   }
 })
 
+test("initializes external tui plugins in config order", async () => {
+  const globalJson = path.join(Global.Path.config, "tui.json")
+  const globalJsonc = path.join(Global.Path.config, "tui.jsonc")
+  const backupJson = await Bun.file(globalJson)
+    .text()
+    .catch(() => undefined)
+  const backupJsonc = await Bun.file(globalJsonc)
+    .text()
+    .catch(() => undefined)
+
+  await fs.rm(globalJson, { force: true }).catch(() => {})
+  await fs.rm(globalJsonc, { force: true }).catch(() => {})
+
+  await using tmp = await tmpdir({
+    init: async (dir) => {
+      const a = path.join(dir, "order-a.ts")
+      const b = path.join(dir, "order-b.ts")
+      const aSpec = pathToFileURL(a).href
+      const bSpec = pathToFileURL(b).href
+      const marker = path.join(dir, "tui-order.txt")
+
+      await Bun.write(
+        a,
+        `import fs from "fs/promises"
+
+export default {
+  id: "demo.tui.order.a",
+  tui: async () => {
+    await fs.appendFile(${JSON.stringify(marker)}, "a-start\\n")
+    await Bun.sleep(25)
+    await fs.appendFile(${JSON.stringify(marker)}, "a-end\\n")
+  },
+}
+`,
+      )
+      await Bun.write(
+        b,
+        `import fs from "fs/promises"
+
+export default {
+  id: "demo.tui.order.b",
+  tui: async () => {
+    await fs.appendFile(${JSON.stringify(marker)}, "b\\n")
+  },
+}
+`,
+      )
+      await Bun.write(path.join(dir, "tui.json"), JSON.stringify({ plugin: [aSpec, bSpec] }, null, 2))
+
+      return { marker }
+    },
+  })
+
+  process.env.OPENCODE_PLUGIN_META_FILE = path.join(tmp.path, "plugin-meta.json")
+  const cwd = spyOn(process, "cwd").mockImplementation(() => tmp.path)
+
+  try {
+    await TuiPluginRuntime.init(createTuiPluginApi())
+    const lines = (await fs.readFile(tmp.extra.marker, "utf8")).trim().split("\n")
+    expect(lines).toEqual(["a-start", "a-end", "b"])
+  } finally {
+    await TuiPluginRuntime.dispose()
+    cwd.mockRestore()
+    delete process.env.OPENCODE_PLUGIN_META_FILE
+
+    if (backupJson === undefined) {
+      await fs.rm(globalJson, { force: true }).catch(() => {})
+    } else {
+      await Bun.write(globalJson, backupJson)
+    }
+    if (backupJsonc === undefined) {
+      await fs.rm(globalJsonc, { force: true }).catch(() => {})
+    } else {
+      await Bun.write(globalJsonc, backupJsonc)
+    }
+  }
+})
+
 describe("tui.plugin.loader", () => {
   let data: Data
 
@@ -561,3 +647,106 @@ describe("tui.plugin.loader", () => {
     expect(data.leaked_global_to_local).toBe(false)
   })
 })
+
+test("updates installed theme when plugin metadata changes", async () => {
+  await using tmp = await tmpdir<{
+    spec: string
+    pluginPath: string
+    themePath: string
+    dest: string
+    themeName: string
+  }>({
+    init: async (dir) => {
+      const pluginPath = path.join(dir, "theme-update-plugin.ts")
+      const spec = pathToFileURL(pluginPath).href
+      const themeFile = "theme-update.json"
+      const themePath = path.join(dir, themeFile)
+      const dest = path.join(dir, ".opencode", "themes", themeFile)
+      const themeName = themeFile.replace(/\.json$/, "")
+      const configPath = path.join(dir, "tui.json")
+
+      await Bun.write(themePath, JSON.stringify({ theme: { primary: "#111111" } }, null, 2))
+      await Bun.write(
+        pluginPath,
+        `export default {
+  id: "demo.theme-update",
+  tui: async (api, options) => {
+    if (!options?.theme_path) return
+    await api.theme.install(options.theme_path)
+  },
+}
+`,
+      )
+      await Bun.write(
+        configPath,
+        JSON.stringify(
+          {
+            plugin: [[spec, { theme_path: `./${themeFile}` }]],
+          },
+          null,
+          2,
+        ),
+      )
+
+      return {
+        spec,
+        pluginPath,
+        themePath,
+        dest,
+        themeName,
+      }
+    },
+  })
+
+  process.env.OPENCODE_PLUGIN_META_FILE = path.join(tmp.path, "plugin-meta.json")
+  const cwd = spyOn(process, "cwd").mockImplementation(() => tmp.path)
+  const wait = spyOn(TuiConfig, "waitForDependencies").mockResolvedValue()
+  const install = spyOn(Config, "installDependencies").mockResolvedValue()
+
+  const api = () =>
+    createTuiPluginApi({
+      theme: {
+        has(name) {
+          return allThemes()[name] !== undefined
+        },
+      },
+    })
+
+  try {
+    await TuiPluginRuntime.init(api())
+    await TuiPluginRuntime.dispose()
+    await expect(fs.readFile(tmp.extra.dest, "utf8")).resolves.toContain("#111111")
+
+    await Bun.write(tmp.extra.themePath, JSON.stringify({ theme: { primary: "#222222" } }, null, 2))
+    await Bun.write(
+      tmp.extra.pluginPath,
+      `export default {
+  id: "demo.theme-update",
+  tui: async (api, options) => {
+    if (!options?.theme_path) return
+    await api.theme.install(options.theme_path)
+  },
+}
+// v2
+`,
+    )
+    const stamp = new Date(Date.now() + 10_000)
+    await fs.utimes(tmp.extra.pluginPath, stamp, stamp)
+    await fs.utimes(tmp.extra.themePath, stamp, stamp)
+
+    await TuiPluginRuntime.init(api())
+    const text = await fs.readFile(tmp.extra.dest, "utf8")
+    expect(text).toContain("#222222")
+    expect(text).not.toContain("#111111")
+    const list = await Filesystem.readJson<Record<string, { themes?: Record<string, { dest: string }> }>>(
+      process.env.OPENCODE_PLUGIN_META_FILE!,
+    )
+    expect(list["demo.theme-update"]?.themes?.[tmp.extra.themeName]?.dest).toBe(tmp.extra.dest)
+  } finally {
+    await TuiPluginRuntime.dispose()
+    cwd.mockRestore()
+    wait.mockRestore()
+    install.mockRestore()
+    delete process.env.OPENCODE_PLUGIN_META_FILE
+  }
+})

packages/opencode/test/cli/tui/plugin-toggle.test.ts

@@ -44,12 +44,13 @@ test("toggles plugin runtime state by exported id", async () => {
     plugin_enabled: {
       "demo.toggle": false,
     },
-    plugin_meta: {
-      [tmp.extra.spec]: {
+    plugin_records: [
+      {
+        item: [tmp.extra.spec, { marker: tmp.extra.marker }],
         scope: "local",
         source: path.join(tmp.path, "tui.json"),
       },
-    },
+    ],
   })
   const wait = spyOn(TuiConfig, "waitForDependencies").mockResolvedValue()
   const cwd = spyOn(process, "cwd").mockImplementation(() => tmp.path)
@@ -121,12 +122,13 @@ test("kv plugin_enabled overrides tui config on startup", async () => {
     plugin_enabled: {
       "demo.startup": false,
     },
-    plugin_meta: {
-      [tmp.extra.spec]: {
+    plugin_records: [
+      {
+        item: [tmp.extra.spec, { marker: tmp.extra.marker }],
         scope: "local",
         source: path.join(tmp.path, "tui.json"),
       },
-    },
+    ],
   })
   const wait = spyOn(TuiConfig, "waitForDependencies").mockResolvedValue()
   const cwd = spyOn(process, "cwd").mockImplementation(() => tmp.path)

packages/opencode/test/config/config.test.ts

@@ -1822,6 +1822,22 @@ describe("resolvePluginSpec", () => {
     expect(await Config.resolvePluginSpec("@scope/pkg", file)).toBe("@scope/pkg")
   })
 
+  test("resolves windows-style relative plugin directory specs", async () => {
+    if (process.platform !== "win32") return
+
+    await using tmp = await tmpdir({
+      init: async (dir) => {
+        const plugin = path.join(dir, "plugin")
+        await fs.mkdir(plugin, { recursive: true })
+        await Filesystem.write(path.join(plugin, "index.ts"), "export default {}")
+      },
+    })
+
+    const file = path.join(tmp.path, "opencode.json")
+    const hit = await Config.resolvePluginSpec(".\\plugin", file)
+    expect(Config.pluginSpecifier(hit)).toBe(pathToFileURL(path.join(tmp.path, "plugin", "index.ts")).href)
+  })
+
   test("resolves relative file plugin paths to file urls", async () => {
     await using tmp = await tmpdir({
       init: async (dir) => {
@@ -1834,7 +1850,7 @@ describe("resolvePluginSpec", () => {
     expect(Config.pluginSpecifier(hit)).toBe(pathToFileURL(path.join(tmp.path, "plugin.ts")).href)
   })
 
-  test("resolves plugin directory paths to package main files", async () => {
+  test("resolves plugin directory paths to directory urls", async () => {
     await using tmp = await tmpdir({
       init: async (dir) => {
         const plugin = path.join(dir, "plugin")
@@ -1848,6 +1864,20 @@ describe("resolvePluginSpec", () => {
       },
     })
 
+    const file = path.join(tmp.path, "opencode.json")
+    const hit = await Config.resolvePluginSpec("./plugin", file)
+    expect(Config.pluginSpecifier(hit)).toBe(pathToFileURL(path.join(tmp.path, "plugin")).href)
+  })
+
+  test("resolves plugin directories without package.json to index.ts", async () => {
+    await using tmp = await tmpdir({
+      init: async (dir) => {
+        const plugin = path.join(dir, "plugin")
+        await fs.mkdir(plugin, { recursive: true })
+        await Filesystem.write(path.join(plugin, "index.ts"), "export default {}")
+      },
+    })
+
     const file = path.join(tmp.path, "opencode.json")
     const hit = await Config.resolvePluginSpec("./plugin", file)
     expect(Config.pluginSpecifier(hit)).toBe(pathToFileURL(path.join(tmp.path, "plugin", "index.ts")).href)

packages/opencode/test/config/tui.test.ts

@@ -476,12 +476,13 @@ test("loads managed tui config and gives it highest precedence", async () => {
       const config = await TuiConfig.get()
       expect(config.theme).toBe("managed-theme")
       expect(config.plugin).toEqual(["shared-plugin@2.0.0"])
-      expect(config.plugin_meta).toEqual({
-        "shared-plugin@2.0.0": {
+      expect(config.plugin_records).toEqual([
+        {
+          item: "shared-plugin@2.0.0",
           scope: "global",
           source: path.join(managedConfigDir, "tui.json"),
         },
-      })
+      ])
     },
   })
 })
@@ -539,12 +540,13 @@ test("supports tuple plugin specs with options in tui.json", async () => {
     fn: async () => {
       const config = await TuiConfig.get()
       expect(config.plugin).toEqual([["acme-plugin@1.2.3", { enabled: true, label: "demo" }]])
-      expect(config.plugin_meta).toEqual({
-        "acme-plugin@1.2.3": {
+      expect(config.plugin_records).toEqual([
+        {
+          item: ["acme-plugin@1.2.3", { enabled: true, label: "demo" }],
           scope: "local",
           source: path.join(tmp.path, "tui.json"),
         },
-      })
+      ])
     },
   })
 })
@@ -578,16 +580,18 @@ test("deduplicates tuple plugin specs by name with higher precedence winning", a
         ["acme-plugin@2.0.0", { source: "project" }],
         ["second-plugin@3.0.0", { source: "project" }],
       ])
-      expect(config.plugin_meta).toEqual({
-        "acme-plugin@2.0.0": {
+      expect(config.plugin_records).toEqual([
+        {
+          item: ["acme-plugin@2.0.0", { source: "project" }],
           scope: "local",
           source: path.join(tmp.path, "tui.json"),
         },
-        "second-plugin@3.0.0": {
+        {
+          item: ["second-plugin@3.0.0", { source: "project" }],
           scope: "local",
           source: path.join(tmp.path, "tui.json"),
         },
-      })
+      ])
     },
   })
 })
@@ -615,16 +619,18 @@ test("tracks global and local plugin metadata in merged tui config", async () =>
     fn: async () => {
       const config = await TuiConfig.get()
       expect(config.plugin).toEqual(["global-plugin@1.0.0", "local-plugin@2.0.0"])
-      expect(config.plugin_meta).toEqual({
-        "global-plugin@1.0.0": {
+      expect(config.plugin_records).toEqual([
+        {
+          item: "global-plugin@1.0.0",
           scope: "global",
           source: path.join(Global.Path.config, "tui.json"),
         },
-        "local-plugin@2.0.0": {
+        {
+          item: "local-plugin@2.0.0",
           scope: "local",
           source: path.join(tmp.path, "tui.json"),
         },
-      })
+      ])
     },
   })
 })

packages/opencode/test/effect/runner.test.ts

@@ -0,0 +1,523 @@
+import { describe, expect, test } from "bun:test"
+import { Deferred, Effect, Exit, Fiber, Ref, Scope } from "effect"
+import { Runner } from "../../src/effect/runner"
+import { it } from "../lib/effect"
+
+describe("Runner", () => {
+  // --- ensureRunning semantics ---
+
+  it.effect(
+    "ensureRunning starts work and returns result",
+    Effect.gen(function* () {
+      const s = yield* Scope.Scope
+      const runner = Runner.make<string>(s)
+      const result = yield* runner.ensureRunning(Effect.succeed("hello"))
+      expect(result).toBe("hello")
+      expect(runner.state._tag).toBe("Idle")
+      expect(runner.busy).toBe(false)
+    }),
+  )
+
+  it.effect(
+    "ensureRunning propagates work failures",
+    Effect.gen(function* () {
+      const s = yield* Scope.Scope
+      const runner = Runner.make<string, string>(s)
+      const exit = yield* runner.ensureRunning(Effect.fail("boom")).pipe(Effect.exit)
+      expect(Exit.isFailure(exit)).toBe(true)
+      expect(runner.state._tag).toBe("Idle")
+    }),
+  )
+
+  it.effect(
+    "concurrent callers share the same run",
+    Effect.gen(function* () {
+      const s = yield* Scope.Scope
+      const runner = Runner.make<string>(s)
+      const calls = yield* Ref.make(0)
+      const work = Effect.gen(function* () {
+        yield* Ref.update(calls, (n) => n + 1)
+        yield* Effect.sleep("10 millis")
+        return "shared"
+      })
+
+      const [a, b] = yield* Effect.all([runner.ensureRunning(work), runner.ensureRunning(work)], {
+        concurrency: "unbounded",
+      })
+
+      expect(a).toBe("shared")
+      expect(b).toBe("shared")
+      expect(yield* Ref.get(calls)).toBe(1)
+    }),
+  )
+
+  it.effect(
+    "concurrent callers all receive same error",
+    Effect.gen(function* () {
+      const s = yield* Scope.Scope
+      const runner = Runner.make<string, string>(s)
+      const work = Effect.gen(function* () {
+        yield* Effect.sleep("10 millis")
+        return yield* Effect.fail("boom")
+      })
+
+      const [a, b] = yield* Effect.all(
+        [runner.ensureRunning(work).pipe(Effect.exit), runner.ensureRunning(work).pipe(Effect.exit)],
+        { concurrency: "unbounded" },
+      )
+
+      expect(Exit.isFailure(a)).toBe(true)
+      expect(Exit.isFailure(b)).toBe(true)
+    }),
+  )
+
+  it.effect(
+    "ensureRunning can be called again after previous run completes",
+    Effect.gen(function* () {
+      const s = yield* Scope.Scope
+      const runner = Runner.make<string>(s)
+      expect(yield* runner.ensureRunning(Effect.succeed("first"))).toBe("first")
+      expect(yield* runner.ensureRunning(Effect.succeed("second"))).toBe("second")
+    }),
+  )
+
+  it.effect(
+    "second ensureRunning ignores new work if already running",
+    Effect.gen(function* () {
+      const s = yield* Scope.Scope
+      const runner = Runner.make<string>(s)
+      const ran = yield* Ref.make<string[]>([])
+
+      const first = Effect.gen(function* () {
+        yield* Ref.update(ran, (a) => [...a, "first"])
+        yield* Effect.sleep("50 millis")
+        return "first-result"
+      })
+      const second = Effect.gen(function* () {
+        yield* Ref.update(ran, (a) => [...a, "second"])
+        return "second-result"
+      })
+
+      const [a, b] = yield* Effect.all([runner.ensureRunning(first), runner.ensureRunning(second)], {
+        concurrency: "unbounded",
+      })
+
+      expect(a).toBe("first-result")
+      expect(b).toBe("first-result")
+      expect(yield* Ref.get(ran)).toEqual(["first"])
+    }),
+  )
+
+  // --- cancel semantics ---
+
+  it.effect(
+    "cancel interrupts running work",
+    Effect.gen(function* () {
+      const s = yield* Scope.Scope
+      const runner = Runner.make<string>(s)
+      const fiber = yield* runner.ensureRunning(Effect.never.pipe(Effect.as("never"))).pipe(Effect.forkChild)
+      yield* Effect.sleep("10 millis")
+      expect(runner.busy).toBe(true)
+      expect(runner.state._tag).toBe("Running")
+
+      yield* runner.cancel
+      expect(runner.busy).toBe(false)
+
+      const exit = yield* Fiber.await(fiber)
+      expect(Exit.isFailure(exit)).toBe(true)
+    }),
+  )
+
+  it.effect(
+    "cancel on idle is a no-op",
+    Effect.gen(function* () {
+      const s = yield* Scope.Scope
+      const runner = Runner.make<string>(s)
+      yield* runner.cancel
+      expect(runner.busy).toBe(false)
+    }),
+  )
+
+  it.effect(
+    "cancel with onInterrupt resolves callers gracefully",
+    Effect.gen(function* () {
+      const s = yield* Scope.Scope
+      const runner = Runner.make<string>(s, { onInterrupt: Effect.succeed("fallback") })
+      const fiber = yield* runner.ensureRunning(Effect.never.pipe(Effect.as("never"))).pipe(Effect.forkChild)
+      yield* Effect.sleep("10 millis")
+
+      yield* runner.cancel
+
+      const exit = yield* Fiber.await(fiber)
+      expect(Exit.isSuccess(exit)).toBe(true)
+      if (Exit.isSuccess(exit)) expect(exit.value).toBe("fallback")
+    }),
+  )
+
+  it.effect(
+    "cancel with queued callers resolves all",
+    Effect.gen(function* () {
+      const s = yield* Scope.Scope
+      const runner = Runner.make<string>(s, { onInterrupt: Effect.succeed("fallback") })
+
+      const a = yield* runner.ensureRunning(Effect.never.pipe(Effect.as("x"))).pipe(Effect.forkChild)
+      yield* Effect.sleep("10 millis")
+      const b = yield* runner.ensureRunning(Effect.succeed("y")).pipe(Effect.forkChild)
+      yield* Effect.sleep("10 millis")
+
+      yield* runner.cancel
+
+      const [exitA, exitB] = yield* Effect.all([Fiber.await(a), Fiber.await(b)])
+      expect(Exit.isSuccess(exitA)).toBe(true)
+      expect(Exit.isSuccess(exitB)).toBe(true)
+      if (Exit.isSuccess(exitA)) expect(exitA.value).toBe("fallback")
+      if (Exit.isSuccess(exitB)) expect(exitB.value).toBe("fallback")
+    }),
+  )
+
+  it.effect(
+    "work can be started after cancel",
+    Effect.gen(function* () {
+      const s = yield* Scope.Scope
+      const runner = Runner.make<string>(s)
+      const fiber = yield* runner.ensureRunning(Effect.never.pipe(Effect.as("x"))).pipe(Effect.forkChild)
+      yield* Effect.sleep("10 millis")
+      yield* runner.cancel
+      yield* Fiber.await(fiber)
+
+      const result = yield* runner.ensureRunning(Effect.succeed("after-cancel"))
+      expect(result).toBe("after-cancel")
+    }),
+  )
+
+  test("cancel does not deadlock when replacement work starts before interrupted run exits", async () => {
+    function defer() {
+      let resolve!: () => void
+      const promise = new Promise<void>((done) => {
+        resolve = done
+      })
+      return { promise, resolve }
+    }
+
+    function fail(ms: number, msg: string) {
+      return new Promise<never>((_, reject) => {
+        setTimeout(() => reject(new Error(msg)), ms)
+      })
+    }
+
+    const s = await Effect.runPromise(Scope.make())
+    const hit = defer()
+    const hold = defer()
+    const done = defer()
+    try {
+      const runner = Runner.make<string>(s)
+      const first = Effect.never.pipe(
+        Effect.onInterrupt(() => Effect.sync(() => hit.resolve())),
+        Effect.ensuring(Effect.promise(() => hold.promise)),
+        Effect.as("first"),
+      )
+
+      const a = Effect.runPromiseExit(runner.ensureRunning(first))
+      await Bun.sleep(10)
+
+      const stop = Effect.runPromise(runner.cancel)
+      await Promise.race([hit.promise, fail(250, "cancel did not interrupt running work")])
+
+      const b = Effect.runPromise(runner.ensureRunning(Effect.promise(() => done.promise).pipe(Effect.as("second"))))
+      expect(runner.busy).toBe(true)
+
+      hold.resolve()
+      await Promise.race([stop, fail(250, "cancel deadlocked while replacement run was active")])
+
+      expect(runner.busy).toBe(true)
+      done.resolve()
+      expect(await b).toBe("second")
+      expect(runner.busy).toBe(false)
+
+      const exit = await a
+      expect(Exit.isFailure(exit)).toBe(true)
+    } finally {
+      hold.resolve()
+      done.resolve()
+      await Promise.race([Effect.runPromise(Scope.close(s, Exit.void)), fail(1000, "runner scope did not close")])
+    }
+  })
+
+  // --- shell semantics ---
+
+  it.effect(
+    "shell runs exclusively",
+    Effect.gen(function* () {
+      const s = yield* Scope.Scope
+      const runner = Runner.make<string>(s)
+      const result = yield* runner.startShell((_signal) => Effect.succeed("shell-done"))
+      expect(result).toBe("shell-done")
+      expect(runner.busy).toBe(false)
+    }),
+  )
+
+  it.effect(
+    "shell rejects when run is active",
+    Effect.gen(function* () {
+      const s = yield* Scope.Scope
+      const runner = Runner.make<string>(s)
+      const fiber = yield* runner.ensureRunning(Effect.never.pipe(Effect.as("x"))).pipe(Effect.forkChild)
+      yield* Effect.sleep("10 millis")
+
+      const exit = yield* runner.startShell((_s) => Effect.succeed("nope")).pipe(Effect.exit)
+      expect(Exit.isFailure(exit)).toBe(true)
+
+      yield* runner.cancel
+      yield* Fiber.await(fiber)
+    }),
+  )
+
+  it.effect(
+    "shell rejects when another shell is running",
+    Effect.gen(function* () {
+      const s = yield* Scope.Scope
+      const runner = Runner.make<string>(s)
+      const gate = yield* Deferred.make<void>()
+
+      const sh = yield* runner
+        .startShell((_signal) => Deferred.await(gate).pipe(Effect.as("first")))
+        .pipe(Effect.forkChild)
+      yield* Effect.sleep("10 millis")
+
+      const exit = yield* runner.startShell((_s) => Effect.succeed("second")).pipe(Effect.exit)
+      expect(Exit.isFailure(exit)).toBe(true)
+
+      yield* Deferred.succeed(gate, undefined)
+      yield* Fiber.await(sh)
+    }),
+  )
+
+  it.effect(
+    "shell rejects via busy callback and cancel still stops the first shell",
+    Effect.gen(function* () {
+      const s = yield* Scope.Scope
+      const runner = Runner.make<string>(s, {
+        busy: () => {
+          throw new Error("busy")
+        },
+      })
+
+      const sh = yield* runner
+        .startShell((signal) =>
+          Effect.promise(
+            () =>
+              new Promise<string>((resolve) => {
+                signal.addEventListener("abort", () => resolve("aborted"), { once: true })
+              }),
+          ),
+        )
+        .pipe(Effect.forkChild)
+      yield* Effect.sleep("10 millis")
+
+      const exit = yield* runner.startShell((_s) => Effect.succeed("second")).pipe(Effect.exit)
+      expect(Exit.isFailure(exit)).toBe(true)
+
+      yield* runner.cancel
+      const done = yield* Fiber.await(sh)
+      expect(Exit.isSuccess(done)).toBe(true)
+    }),
+  )
+
+  it.effect(
+    "cancel interrupts shell that ignores abort signal",
+    Effect.gen(function* () {
+      const s = yield* Scope.Scope
+      const runner = Runner.make<string>(s)
+      const gate = yield* Deferred.make<void>()
+
+      const sh = yield* runner
+        .startShell((_signal) => Deferred.await(gate).pipe(Effect.as("ignored")))
+        .pipe(Effect.forkChild)
+      yield* Effect.sleep("10 millis")
+
+      const stop = yield* runner.cancel.pipe(Effect.forkChild)
+      const stopExit = yield* Fiber.await(stop).pipe(Effect.timeout("250 millis"))
+      expect(Exit.isSuccess(stopExit)).toBe(true)
+      expect(runner.busy).toBe(false)
+
+      const shellExit = yield* Fiber.await(sh)
+      expect(Exit.isFailure(shellExit)).toBe(true)
+
+      yield* Deferred.succeed(gate, undefined).pipe(Effect.ignore)
+    }),
+  )
+
+  // --- shell→run handoff ---
+
+  it.effect(
+    "ensureRunning queues behind shell then runs after",
+    Effect.gen(function* () {
+      const s = yield* Scope.Scope
+      const runner = Runner.make<string>(s)
+      const gate = yield* Deferred.make<void>()
+
+      const sh = yield* runner
+        .startShell((_signal) => Deferred.await(gate).pipe(Effect.as("shell-result")))
+        .pipe(Effect.forkChild)
+      yield* Effect.sleep("10 millis")
+      expect(runner.state._tag).toBe("Shell")
+
+      const run = yield* runner.ensureRunning(Effect.succeed("run-result")).pipe(Effect.forkChild)
+      yield* Effect.sleep("10 millis")
+      expect(runner.state._tag).toBe("ShellThenRun")
+
+      yield* Deferred.succeed(gate, undefined)
+      yield* Fiber.await(sh)
+
+      const exit = yield* Fiber.await(run)
+      expect(Exit.isSuccess(exit)).toBe(true)
+      if (Exit.isSuccess(exit)) expect(exit.value).toBe("run-result")
+      expect(runner.state._tag).toBe("Idle")
+    }),
+  )
+
+  it.effect(
+    "multiple ensureRunning callers share the queued run behind shell",
+    Effect.gen(function* () {
+      const s = yield* Scope.Scope
+      const runner = Runner.make<string>(s)
+      const calls = yield* Ref.make(0)
+      const gate = yield* Deferred.make<void>()
+
+      const sh = yield* runner
+        .startShell((_signal) => Deferred.await(gate).pipe(Effect.as("shell")))
+        .pipe(Effect.forkChild)
+      yield* Effect.sleep("10 millis")
+
+      const work = Effect.gen(function* () {
+        yield* Ref.update(calls, (n) => n + 1)
+        return "run"
+      })
+      const a = yield* runner.ensureRunning(work).pipe(Effect.forkChild)
+      const b = yield* runner.ensureRunning(work).pipe(Effect.forkChild)
+      yield* Effect.sleep("10 millis")
+
+      yield* Deferred.succeed(gate, undefined)
+      yield* Fiber.await(sh)
+
+      const [exitA, exitB] = yield* Effect.all([Fiber.await(a), Fiber.await(b)])
+      expect(Exit.isSuccess(exitA)).toBe(true)
+      expect(Exit.isSuccess(exitB)).toBe(true)
+      expect(yield* Ref.get(calls)).toBe(1)
+    }),
+  )
+
+  it.effect(
+    "cancel during shell_then_run cancels both",
+    Effect.gen(function* () {
+      const s = yield* Scope.Scope
+      const runner = Runner.make<string>(s)
+      const gate = yield* Deferred.make<void>()
+
+      const sh = yield* runner
+        .startShell((signal) =>
+          Effect.promise(
+            () =>
+              new Promise<string>((resolve) => {
+                signal.addEventListener("abort", () => resolve("aborted"), { once: true })
+              }),
+          ),
+        )
+        .pipe(Effect.forkChild)
+      yield* Effect.sleep("10 millis")
+
+      const run = yield* runner.ensureRunning(Effect.succeed("y")).pipe(Effect.forkChild)
+      yield* Effect.sleep("10 millis")
+      expect(runner.state._tag).toBe("ShellThenRun")
+
+      yield* runner.cancel
+      expect(runner.busy).toBe(false)
+
+      yield* Fiber.await(sh)
+      const exit = yield* Fiber.await(run)
+      expect(Exit.isFailure(exit)).toBe(true)
+    }),
+  )
+
+  // --- lifecycle callbacks ---
+
+  it.effect(
+    "onIdle fires when returning to idle from running",
+    Effect.gen(function* () {
+      const s = yield* Scope.Scope
+      const count = yield* Ref.make(0)
+      const runner = Runner.make<string>(s, {
+        onIdle: Ref.update(count, (n) => n + 1),
+      })
+      yield* runner.ensureRunning(Effect.succeed("ok"))
+      expect(yield* Ref.get(count)).toBe(1)
+    }),
+  )
+
+  it.effect(
+    "onIdle fires on cancel",
+    Effect.gen(function* () {
+      const s = yield* Scope.Scope
+      const count = yield* Ref.make(0)
+      const runner = Runner.make<string>(s, {
+        onIdle: Ref.update(count, (n) => n + 1),
+      })
+      const fiber = yield* runner.ensureRunning(Effect.never.pipe(Effect.as("x"))).pipe(Effect.forkChild)
+      yield* Effect.sleep("10 millis")
+      yield* runner.cancel
+      yield* Fiber.await(fiber)
+      expect(yield* Ref.get(count)).toBeGreaterThanOrEqual(1)
+    }),
+  )
+
+  it.effect(
+    "onBusy fires when shell starts",
+    Effect.gen(function* () {
+      const s = yield* Scope.Scope
+      const count = yield* Ref.make(0)
+      const runner = Runner.make<string>(s, {
+        onBusy: Ref.update(count, (n) => n + 1),
+      })
+      yield* runner.startShell((_signal) => Effect.succeed("done"))
+      expect(yield* Ref.get(count)).toBe(1)
+    }),
+  )
+
+  // --- busy flag ---
+
+  it.effect(
+    "busy is true during run",
+    Effect.gen(function* () {
+      const s = yield* Scope.Scope
+      const runner = Runner.make<string>(s)
+      const gate = yield* Deferred.make<void>()
+
+      const fiber = yield* runner.ensureRunning(Deferred.await(gate).pipe(Effect.as("ok"))).pipe(Effect.forkChild)
+      yield* Effect.sleep("10 millis")
+      expect(runner.busy).toBe(true)
+
+      yield* Deferred.succeed(gate, undefined)
+      yield* Fiber.await(fiber)
+      expect(runner.busy).toBe(false)
+    }),
+  )
+
+  it.effect(
+    "busy is true during shell",
+    Effect.gen(function* () {
+      const s = yield* Scope.Scope
+      const runner = Runner.make<string>(s)
+      const gate = yield* Deferred.make<void>()
+
+      const fiber = yield* runner
+        .startShell((_signal) => Deferred.await(gate).pipe(Effect.as("ok")))
+        .pipe(Effect.forkChild)
+      yield* Effect.sleep("10 millis")
+      expect(runner.busy).toBe(true)
+
+      yield* Deferred.succeed(gate, undefined)
+      yield* Fiber.await(fiber)
+      expect(runner.busy).toBe(false)
+    }),
+  )
+})

packages/opencode/test/fixture/tui-runtime.ts

@@ -6,21 +6,14 @@ type PluginSpec = string | [string, Record<string, unknown>]
 
 export function mockTuiRuntime(dir: string, plugin: PluginSpec[]) {
   process.env.OPENCODE_PLUGIN_META_FILE = path.join(dir, "plugin-meta.json")
-  const meta = Object.fromEntries(
-    plugin.map((item) => {
-      const spec = Array.isArray(item) ? item[0] : item
-      return [
-        spec,
-        {
-          scope: "local" as const,
-          source: path.join(dir, "tui.json"),
-        },
-      ]
-    }),
-  )
+  const plugin_records = plugin.map((item) => ({
+    item,
+    scope: "local" as const,
+    source: path.join(dir, "tui.json"),
+  }))
   const get = spyOn(TuiConfig, "get").mockResolvedValue({
     plugin,
-    plugin_meta: meta,
+    plugin_records,
   })
   const wait = spyOn(TuiConfig, "waitForDependencies").mockResolvedValue()
   const cwd = spyOn(process, "cwd").mockImplementation(() => dir)

packages/opencode/test/plugin/loader-shared.test.ts

@@ -331,6 +331,168 @@ describe("plugin.loader.shared", () => {
     }
   })
 
+  test("loads npm server plugin from package server export without leading dot", async () => {
+    await using tmp = await tmpdir({
+      init: async (dir) => {
+        const mod = path.join(dir, "mods", "acme-plugin")
+        const dist = path.join(mod, "dist")
+        const mark = path.join(dir, "server-called.txt")
+        await fs.mkdir(dist, { recursive: true })
+
+        await Bun.write(
+          path.join(mod, "package.json"),
+          JSON.stringify(
+            {
+              name: "acme-plugin",
+              type: "module",
+              exports: {
+                ".": "./index.js",
+                "./server": "dist/server.js",
+              },
+            },
+            null,
+            2,
+          ),
+        )
+        await Bun.write(path.join(mod, "index.js"), 'import "./main-throws.js"\nexport default {}\n')
+        await Bun.write(path.join(mod, "main-throws.js"), 'throw new Error("main loaded")\n')
+        await Bun.write(
+          path.join(dist, "server.js"),
+          [
+            "export default {",
+            "  server: async () => {",
+            `    await Bun.write(${JSON.stringify(mark)}, "called")`,
+            "    return {}",
+            "  },",
+            "}",
+            "",
+          ].join("\n"),
+        )
+
+        await Bun.write(path.join(dir, "opencode.json"), JSON.stringify({ plugin: ["acme-plugin@1.0.0"] }, null, 2))
+
+        return {
+          mod,
+          mark,
+        }
+      },
+    })
+
+    const install = spyOn(BunProc, "install").mockResolvedValue(tmp.extra.mod)
+
+    try {
+      const errors = await errs(tmp.path)
+      expect(errors).toHaveLength(0)
+      expect(await Bun.file(tmp.extra.mark).text()).toBe("called")
+    } finally {
+      install.mockRestore()
+    }
+  })
+
+  test("loads npm server plugin from package main without leading dot", async () => {
+    await using tmp = await tmpdir({
+      init: async (dir) => {
+        const mod = path.join(dir, "mods", "acme-plugin")
+        const dist = path.join(mod, "dist")
+        const mark = path.join(dir, "main-called.txt")
+        await fs.mkdir(dist, { recursive: true })
+
+        await Bun.write(
+          path.join(mod, "package.json"),
+          JSON.stringify(
+            {
+              name: "acme-plugin",
+              type: "module",
+              main: "dist/index.js",
+            },
+            null,
+            2,
+          ),
+        )
+        await Bun.write(
+          path.join(dist, "index.js"),
+          [
+            "export default {",
+            "  server: async () => {",
+            `    await Bun.write(${JSON.stringify(mark)}, "called")`,
+            "    return {}",
+            "  },",
+            "}",
+            "",
+          ].join("\n"),
+        )
+
+        await Bun.write(path.join(dir, "opencode.json"), JSON.stringify({ plugin: ["acme-plugin@1.0.0"] }, null, 2))
+
+        return {
+          mod,
+          mark,
+        }
+      },
+    })
+
+    const install = spyOn(BunProc, "install").mockResolvedValue(tmp.extra.mod)
+
+    try {
+      const errors = await errs(tmp.path)
+      expect(errors).toHaveLength(0)
+      expect(await Bun.file(tmp.extra.mark).text()).toBe("called")
+    } finally {
+      install.mockRestore()
+    }
+  })
+
+  test("does not use npm package exports dot for server entry", async () => {
+    await using tmp = await tmpdir({
+      init: async (dir) => {
+        const mod = path.join(dir, "mods", "acme-plugin")
+        const mark = path.join(dir, "dot-server.txt")
+        await fs.mkdir(mod, { recursive: true })
+
+        await Bun.write(
+          path.join(mod, "package.json"),
+          JSON.stringify({
+            name: "acme-plugin",
+            type: "module",
+            exports: { ".": "./index.js" },
+          }),
+        )
+        await Bun.write(
+          path.join(mod, "index.js"),
+          [
+            "export default {",
+            '  id: "demo.dot.server",',
+            "  server: async () => {",
+            `    await Bun.write(${JSON.stringify(mark)}, "called")`,
+            "    return {}",
+            "  },",
+            "}",
+            "",
+          ].join("\n"),
+        )
+
+        await Bun.write(path.join(dir, "opencode.json"), JSON.stringify({ plugin: ["acme-plugin@1.0.0"] }, null, 2))
+
+        return { mod, mark }
+      },
+    })
+
+    const install = spyOn(BunProc, "install").mockResolvedValue(tmp.extra.mod)
+
+    try {
+      const errors = await errs(tmp.path)
+      const called = await Bun.file(tmp.extra.mark)
+        .text()
+        .then(() => true)
+        .catch(() => false)
+
+      expect(called).toBe(false)
+      expect(errors.some((x) => x.includes('exports["./server"]') && x.includes("package.json main"))).toBe(true)
+    } finally {
+      install.mockRestore()
+    }
+  })
+
   test("rejects npm server export that resolves outside plugin directory", async () => {
     await using tmp = await tmpdir({
       init: async (dir) => {
@@ -576,6 +738,55 @@ describe("plugin.loader.shared", () => {
     })
   })
 
+  test("initializes server plugins in config order", async () => {
+    await using tmp = await tmpdir({
+      init: async (dir) => {
+        const a = path.join(dir, "a-plugin.ts")
+        const b = path.join(dir, "b-plugin.ts")
+        const marker = path.join(dir, "server-order.txt")
+        const aSpec = pathToFileURL(a).href
+        const bSpec = pathToFileURL(b).href
+
+        await Bun.write(
+          a,
+          `import fs from "fs/promises"
+
+export default {
+  id: "demo.order.a",
+  server: async () => {
+    await fs.appendFile(${JSON.stringify(marker)}, "a-start\\n")
+    await Bun.sleep(25)
+    await fs.appendFile(${JSON.stringify(marker)}, "a-end\\n")
+    return {}
+  },
+}
+`,
+        )
+        await Bun.write(
+          b,
+          `import fs from "fs/promises"
+
+export default {
+  id: "demo.order.b",
+  server: async () => {
+    await fs.appendFile(${JSON.stringify(marker)}, "b\\n")
+    return {}
+  },
+}
+`,
+        )
+
+        await Bun.write(path.join(dir, "opencode.json"), JSON.stringify({ plugin: [aSpec, bSpec] }, null, 2))
+
+        return { marker }
+      },
+    })
+
+    await load(tmp.path)
+    const lines = (await fs.readFile(tmp.extra.marker, "utf8")).trim().split("\n")
+    expect(lines).toEqual(["a-start", "a-end", "b"])
+  })
+
   test("skips external plugins in pure mode", async () => {
     await using tmp = await tmpdir({
       init: async (dir) => {

packages/opencode/test/server/session-actions.test.ts

@@ -0,0 +1,83 @@
+import { afterEach, describe, expect, mock, spyOn, test } from "bun:test"
+import { Instance } from "../../src/project/instance"
+import { Server } from "../../src/server/server"
+import { Session } from "../../src/session"
+import { ModelID, ProviderID } from "../../src/provider/schema"
+import { MessageID, PartID, type SessionID } from "../../src/session/schema"
+import { SessionPrompt } from "../../src/session/prompt"
+import { Log } from "../../src/util/log"
+import { tmpdir } from "../fixture/fixture"
+
+Log.init({ print: false })
+
+afterEach(async () => {
+  mock.restore()
+  await Instance.disposeAll()
+})
+
+async function user(sessionID: SessionID, text: string) {
+  const msg = await Session.updateMessage({
+    id: MessageID.ascending(),
+    role: "user",
+    sessionID,
+    agent: "build",
+    model: { providerID: ProviderID.make("test"), modelID: ModelID.make("test") },
+    time: { created: Date.now() },
+  })
+  await Session.updatePart({
+    id: PartID.ascending(),
+    sessionID,
+    messageID: msg.id,
+    type: "text",
+    text,
+  })
+  return msg
+}
+
+describe("session action routes", () => {
+  test("abort route calls SessionPrompt.cancel", async () => {
+    await using tmp = await tmpdir({ git: true })
+    await Instance.provide({
+      directory: tmp.path,
+      fn: async () => {
+        const session = await Session.create({})
+        const cancel = spyOn(SessionPrompt, "cancel").mockResolvedValue()
+        const app = Server.Default()
+
+        const res = await app.request(`/session/${session.id}/abort`, {
+          method: "POST",
+        })
+
+        expect(res.status).toBe(200)
+        expect(await res.json()).toBe(true)
+        expect(cancel).toHaveBeenCalledWith(session.id)
+
+        await Session.remove(session.id)
+      },
+    })
+  })
+
+  test("delete message route returns 400 when session is busy", async () => {
+    await using tmp = await tmpdir({ git: true })
+    await Instance.provide({
+      directory: tmp.path,
+      fn: async () => {
+        const session = await Session.create({})
+        const msg = await user(session.id, "hello")
+        const busy = spyOn(SessionPrompt, "assertNotBusy").mockRejectedValue(new Session.BusyError(session.id))
+        const remove = spyOn(Session, "removeMessage").mockResolvedValue(msg.id)
+        const app = Server.Default()
+
+        const res = await app.request(`/session/${session.id}/message/${msg.id}`, {
+          method: "DELETE",
+        })
+
+        expect(res.status).toBe(400)
+        expect(busy).toHaveBeenCalledWith(session.id)
+        expect(remove).not.toHaveBeenCalled()
+
+        await Session.remove(session.id)
+      },
+    })
+  })
+})

packages/opencode/test/server/session-list.test.ts

@@ -1,26 +1,30 @@
-import { describe, expect, test } from "bun:test"
-import path from "path"
+import { afterEach, describe, expect, test } from "bun:test"
 import { Instance } from "../../src/project/instance"
 import { Session } from "../../src/session"
 import { Log } from "../../src/util/log"
+import { tmpdir } from "../fixture/fixture"
 
-const projectRoot = path.join(__dirname, "../..")
 Log.init({ print: false })
 
+afterEach(async () => {
+  await Instance.disposeAll()
+})
+
 describe("Session.list", () => {
   test("filters by directory", async () => {
+    await using tmp = await tmpdir({ git: true })
     await Instance.provide({
-      directory: projectRoot,
+      directory: tmp.path,
       fn: async () => {
         const first = await Session.create({})
 
-        const otherDir = path.join(projectRoot, "..", "__session_list_other")
+        await using other = await tmpdir({ git: true })
         const second = await Instance.provide({
-          directory: otherDir,
+          directory: other.path,
           fn: async () => Session.create({}),
         })
 
-        const sessions = [...Session.list({ directory: projectRoot })]
+        const sessions = [...Session.list({ directory: tmp.path })]
         const ids = sessions.map((s) => s.id)
 
         expect(ids).toContain(first.id)
@@ -30,8 +34,9 @@ describe("Session.list", () => {
   })
 
   test("filters root sessions", async () => {
+    await using tmp = await tmpdir({ git: true })
     await Instance.provide({
-      directory: projectRoot,
+      directory: tmp.path,
       fn: async () => {
         const root = await Session.create({ title: "root-session" })
         const child = await Session.create({ title: "child-session", parentID: root.id })
@@ -46,8 +51,9 @@ describe("Session.list", () => {
   })
 
   test("filters by start time", async () => {
+    await using tmp = await tmpdir({ git: true })
     await Instance.provide({
-      directory: projectRoot,
+      directory: tmp.path,
       fn: async () => {
         const session = await Session.create({ title: "new-session" })
         const futureStart = Date.now() + 86400000
@@ -59,8 +65,9 @@ describe("Session.list", () => {
   })
 
   test("filters by search term", async () => {
+    await using tmp = await tmpdir({ git: true })
     await Instance.provide({
-      directory: projectRoot,
+      directory: tmp.path,
       fn: async () => {
         await Session.create({ title: "unique-search-term-abc" })
         await Session.create({ title: "other-session-xyz" })
@@ -75,8 +82,9 @@ describe("Session.list", () => {
   })
 
   test("respects limit parameter", async () => {
+    await using tmp = await tmpdir({ git: true })
     await Instance.provide({
-      directory: projectRoot,
+      directory: tmp.path,
       fn: async () => {
         await Session.create({ title: "session-1" })
         await Session.create({ title: "session-2" })

packages/opencode/test/server/session-messages.test.ts

@@ -1,15 +1,30 @@
-import { describe, expect, test } from "bun:test"
-import path from "path"
+import { afterEach, describe, expect, test } from "bun:test"
 import { Instance } from "../../src/project/instance"
 import { Server } from "../../src/server/server"
 import { Session } from "../../src/session"
 import { MessageV2 } from "../../src/session/message-v2"
 import { MessageID, PartID, type SessionID } from "../../src/session/schema"
 import { Log } from "../../src/util/log"
+import { tmpdir } from "../fixture/fixture"
 
-const root = path.join(__dirname, "../..")
 Log.init({ print: false })
 
+afterEach(async () => {
+  await Instance.disposeAll()
+})
+
+async function withoutWatcher<T>(fn: () => Promise<T>) {
+  if (process.platform !== "win32") return fn()
+  const prev = process.env.OPENCODE_EXPERIMENTAL_DISABLE_FILEWATCHER
+  process.env.OPENCODE_EXPERIMENTAL_DISABLE_FILEWATCHER = "true"
+  try {
+    return await fn()
+  } finally {
+    if (prev === undefined) delete process.env.OPENCODE_EXPERIMENTAL_DISABLE_FILEWATCHER
+    else process.env.OPENCODE_EXPERIMENTAL_DISABLE_FILEWATCHER = prev
+  }
+}
+
 async function fill(sessionID: SessionID, count: number, time = (i: number) => Date.now() + i) {
   const ids = [] as MessageID[]
   for (let i = 0; i < count; i++) {
@@ -38,89 +53,101 @@ async function fill(sessionID: SessionID, count: number, time = (i: number) => D
 
 describe("session messages endpoint", () => {
   test("returns cursor headers for older pages", async () => {
-    await Instance.provide({
-      directory: root,
-      fn: async () => {
-        const session = await Session.create({})
-        const ids = await fill(session.id, 5)
-        const app = Server.Default()
+    await using tmp = await tmpdir({ git: true })
+    await withoutWatcher(() =>
+      Instance.provide({
+        directory: tmp.path,
+        fn: async () => {
+          const session = await Session.create({})
+          const ids = await fill(session.id, 5)
+          const app = Server.Default()
 
-        const a = await app.request(`/session/${session.id}/message?limit=2`)
-        expect(a.status).toBe(200)
-        const aBody = (await a.json()) as MessageV2.WithParts[]
-        expect(aBody.map((item) => item.info.id)).toEqual(ids.slice(-2))
-        const cursor = a.headers.get("x-next-cursor")
-        expect(cursor).toBeTruthy()
-        expect(a.headers.get("link")).toContain('rel="next"')
+          const a = await app.request(`/session/${session.id}/message?limit=2`)
+          expect(a.status).toBe(200)
+          const aBody = (await a.json()) as MessageV2.WithParts[]
+          expect(aBody.map((item) => item.info.id)).toEqual(ids.slice(-2))
+          const cursor = a.headers.get("x-next-cursor")
+          expect(cursor).toBeTruthy()
+          expect(a.headers.get("link")).toContain('rel="next"')
 
-        const b = await app.request(`/session/${session.id}/message?limit=2&before=${encodeURIComponent(cursor!)}`)
-        expect(b.status).toBe(200)
-        const bBody = (await b.json()) as MessageV2.WithParts[]
-        expect(bBody.map((item) => item.info.id)).toEqual(ids.slice(-4, -2))
+          const b = await app.request(`/session/${session.id}/message?limit=2&before=${encodeURIComponent(cursor!)}`)
+          expect(b.status).toBe(200)
+          const bBody = (await b.json()) as MessageV2.WithParts[]
+          expect(bBody.map((item) => item.info.id)).toEqual(ids.slice(-4, -2))
 
-        await Session.remove(session.id)
-      },
-    })
+          await Session.remove(session.id)
+        },
+      }),
+    )
   })
 
   test("keeps full-history responses when limit is omitted", async () => {
-    await Instance.provide({
-      directory: root,
-      fn: async () => {
-        const session = await Session.create({})
-        const ids = await fill(session.id, 3)
-        const app = Server.Default()
+    await using tmp = await tmpdir({ git: true })
+    await withoutWatcher(() =>
+      Instance.provide({
+        directory: tmp.path,
+        fn: async () => {
+          const session = await Session.create({})
+          const ids = await fill(session.id, 3)
+          const app = Server.Default()
 
-        const res = await app.request(`/session/${session.id}/message`)
-        expect(res.status).toBe(200)
-        const body = (await res.json()) as MessageV2.WithParts[]
-        expect(body.map((item) => item.info.id)).toEqual(ids)
+          const res = await app.request(`/session/${session.id}/message`)
+          expect(res.status).toBe(200)
+          const body = (await res.json()) as MessageV2.WithParts[]
+          expect(body.map((item) => item.info.id)).toEqual(ids)
 
-        await Session.remove(session.id)
-      },
-    })
+          await Session.remove(session.id)
+        },
+      }),
+    )
   })
 
   test("rejects invalid cursors and missing sessions", async () => {
-    await Instance.provide({
-      directory: root,
-      fn: async () => {
-        const session = await Session.create({})
-        const app = Server.Default()
+    await using tmp = await tmpdir({ git: true })
+    await withoutWatcher(() =>
+      Instance.provide({
+        directory: tmp.path,
+        fn: async () => {
+          const session = await Session.create({})
+          const app = Server.Default()
 
-        const bad = await app.request(`/session/${session.id}/message?limit=2&before=bad`)
-        expect(bad.status).toBe(400)
+          const bad = await app.request(`/session/${session.id}/message?limit=2&before=bad`)
+          expect(bad.status).toBe(400)
 
-        const miss = await app.request(`/session/ses_missing/message?limit=2`)
-        expect(miss.status).toBe(404)
+          const miss = await app.request(`/session/ses_missing/message?limit=2`)
+          expect(miss.status).toBe(404)
 
-        await Session.remove(session.id)
-      },
-    })
+          await Session.remove(session.id)
+        },
+      }),
+    )
   })
 
   test("does not truncate large legacy limit requests", async () => {
-    await Instance.provide({
-      directory: root,
-      fn: async () => {
-        const session = await Session.create({})
-        await fill(session.id, 520)
-        const app = Server.Default()
+    await using tmp = await tmpdir({ git: true })
+    await withoutWatcher(() =>
+      Instance.provide({
+        directory: tmp.path,
+        fn: async () => {
+          const session = await Session.create({})
+          await fill(session.id, 520)
+          const app = Server.Default()
 
-        const res = await app.request(`/session/${session.id}/message?limit=510`)
-        expect(res.status).toBe(200)
-        const body = (await res.json()) as MessageV2.WithParts[]
-        expect(body).toHaveLength(510)
+          const res = await app.request(`/session/${session.id}/message?limit=510`)
+          expect(res.status).toBe(200)
+          const body = (await res.json()) as MessageV2.WithParts[]
+          expect(body).toHaveLength(510)
 
-        await Session.remove(session.id)
-      },
-    })
+          await Session.remove(session.id)
+        },
+      }),
+    )
   })
 })
 
 describe("session.prompt_async error handling", () => {
   test("prompt_async route has error handler for detached prompt call", async () => {
-    const src = await Bun.file(path.join(import.meta.dir, "../../src/server/routes/session.ts")).text()
+    const src = await Bun.file(new URL("../../src/server/routes/session.ts", import.meta.url)).text()
     const start = src.indexOf('"/:sessionID/prompt_async"')
     const end = src.indexOf('"/:sessionID/command"', start)
     expect(start).toBeGreaterThan(-1)

packages/opencode/test/server/session-select.test.ts

@@ -1,17 +1,21 @@
-import { describe, expect, test } from "bun:test"
-import path from "path"
+import { afterEach, describe, expect, test } from "bun:test"
 import { Session } from "../../src/session"
 import { Log } from "../../src/util/log"
 import { Instance } from "../../src/project/instance"
 import { Server } from "../../src/server/server"
+import { tmpdir } from "../fixture/fixture"
 
-const projectRoot = path.join(__dirname, "../..")
 Log.init({ print: false })
 
+afterEach(async () => {
+  await Instance.disposeAll()
+})
+
 describe("tui.selectSession endpoint", () => {
   test("should return 200 when called with valid session", async () => {
+    await using tmp = await tmpdir({ git: true })
     await Instance.provide({
-      directory: projectRoot,
+      directory: tmp.path,
       fn: async () => {
         // #given
         const session = await Session.create({})
@@ -35,8 +39,9 @@ describe("tui.selectSession endpoint", () => {
   })
 
   test("should return 404 when session does not exist", async () => {
+    await using tmp = await tmpdir({ git: true })
     await Instance.provide({
-      directory: projectRoot,
+      directory: tmp.path,
       fn: async () => {
         // #given
         const nonExistentSessionID = "ses_nonexistent123"
@@ -56,8 +61,9 @@ describe("tui.selectSession endpoint", () => {
   })
 
   test("should return 400 when session ID format is invalid", async () => {
+    await using tmp = await tmpdir({ git: true })
     await Instance.provide({
-      directory: projectRoot,
+      directory: tmp.path,
       fn: async () => {
         // #given
         const invalidSessionID = "invalid_session_id"

packages/opencode/test/session/compaction.test.ts

@@ -129,7 +129,7 @@ async function tool(sessionID: SessionID, messageID: MessageID, tool: string, ou
 }
 
 function fake(
-  input: Parameters<(typeof SessionProcessorModule.SessionProcessor)["create"]>[0],
+  input: Parameters<SessionProcessorModule.SessionProcessor.Interface["create"]>[0],
   result: "continue" | "compact",
 ) {
   const msg = input.assistantMessage
@@ -509,6 +509,36 @@ describe("session.compaction.prune", () => {
 })
 
 describe("session.compaction.process", () => {
+  test("throws when parent is not a user message", async () => {
+    await using tmp = await tmpdir()
+    await Instance.provide({
+      directory: tmp.path,
+      fn: async () => {
+        const session = await Session.create({})
+        const msg = await user(session.id, "hello")
+        const reply = await assistant(session.id, msg.id, tmp.path)
+        const rt = runtime("continue")
+        try {
+          const msgs = await Session.messages({ sessionID: session.id })
+          await expect(
+            rt.runPromise(
+              SessionCompaction.Service.use((svc) =>
+                svc.process({
+                  parentID: reply.id,
+                  messages: msgs,
+                  sessionID: session.id,
+                  auto: false,
+                }),
+              ),
+            ),
+          ).rejects.toThrow(`Compaction parent must be a user message: ${reply.id}`)
+        } finally {
+          await rt.dispose()
+        }
+      },
+    })
+  })
+
   test("publishes compacted event on continue", async () => {
     await using tmp = await tmpdir()
     await Instance.provide({
@@ -540,7 +570,6 @@ describe("session.compaction.process", () => {
                 parentID: msg.id,
                 messages: msgs,
                 sessionID: session.id,
-                abort: new AbortController().signal,
                 auto: false,
               }),
             ),
@@ -580,7 +609,6 @@ describe("session.compaction.process", () => {
                 parentID: msg.id,
                 messages: msgs,
                 sessionID: session.id,
-                abort: new AbortController().signal,
                 auto: false,
               }),
             ),
@@ -621,7 +649,6 @@ describe("session.compaction.process", () => {
                 parentID: msg.id,
                 messages: msgs,
                 sessionID: session.id,
-                abort: new AbortController().signal,
                 auto: true,
               }),
             ),
@@ -675,7 +702,6 @@ describe("session.compaction.process", () => {
                 parentID: msg.id,
                 messages: msgs,
                 sessionID: session.id,
-                abort: new AbortController().signal,
                 auto: true,
                 overflow: true,
               }),
@@ -717,7 +743,6 @@ describe("session.compaction.process", () => {
                 parentID: msg.id,
                 messages: msgs,
                 sessionID: session.id,
-                abort: new AbortController().signal,
                 auto: true,
                 overflow: true,
               }),
@@ -792,7 +817,6 @@ describe("session.compaction.process", () => {
                   parentID: msg.id,
                   messages: msgs,
                   sessionID: session.id,
-                  abort: abort.signal,
                   auto: false,
                 }),
               ),
@@ -858,7 +882,6 @@ describe("session.compaction.process", () => {
                   parentID: msg.id,
                   messages: msgs,
                   sessionID: session.id,
-                  abort: abort.signal,
                   auto: false,
                 }),
               ),
@@ -892,6 +915,91 @@ describe("session.compaction.process", () => {
       },
     })
   })
+
+  test("does not allow tool calls while generating the summary", async () => {
+    const stub = llm()
+    stub.push(
+      Stream.make(
+        { type: "start" } satisfies LLM.Event,
+        { type: "tool-input-start", id: "call-1", toolName: "_noop" } satisfies LLM.Event,
+        { type: "tool-call", toolCallId: "call-1", toolName: "_noop", input: {} } satisfies LLM.Event,
+        {
+          type: "finish-step",
+          finishReason: "tool-calls",
+          rawFinishReason: "tool_calls",
+          response: { id: "res", modelId: "test-model", timestamp: new Date() },
+          providerMetadata: undefined,
+          usage: {
+            inputTokens: 1,
+            outputTokens: 1,
+            totalTokens: 2,
+            inputTokenDetails: {
+              noCacheTokens: undefined,
+              cacheReadTokens: undefined,
+              cacheWriteTokens: undefined,
+            },
+            outputTokenDetails: {
+              textTokens: undefined,
+              reasoningTokens: undefined,
+            },
+          },
+        } satisfies LLM.Event,
+        {
+          type: "finish",
+          finishReason: "tool-calls",
+          rawFinishReason: "tool_calls",
+          totalUsage: {
+            inputTokens: 1,
+            outputTokens: 1,
+            totalTokens: 2,
+            inputTokenDetails: {
+              noCacheTokens: undefined,
+              cacheReadTokens: undefined,
+              cacheWriteTokens: undefined,
+            },
+            outputTokenDetails: {
+              textTokens: undefined,
+              reasoningTokens: undefined,
+            },
+          },
+        } satisfies LLM.Event,
+      ),
+    )
+
+    await using tmp = await tmpdir({ git: true })
+    await Instance.provide({
+      directory: tmp.path,
+      fn: async () => {
+        spyOn(ProviderModule.Provider, "getModel").mockResolvedValue(createModel({ context: 100_000, output: 32_000 }))
+
+        const session = await Session.create({})
+        const msg = await user(session.id, "hello")
+        const rt = liveRuntime(stub.layer)
+        try {
+          const msgs = await Session.messages({ sessionID: session.id })
+          await rt.runPromise(
+            SessionCompaction.Service.use((svc) =>
+              svc.process({
+                parentID: msg.id,
+                messages: msgs,
+                sessionID: session.id,
+                auto: false,
+              }),
+            ),
+          )
+
+          const summary = (await Session.messages({ sessionID: session.id })).find(
+            (item) => item.info.role === "assistant" && item.info.summary,
+          )
+
+          expect(summary?.info.role).toBe("assistant")
+          expect(summary?.parts.some((part) => part.type === "tool")).toBe(false)
+        } finally {
+          await rt.dispose()
+        }
+      },
+    })
+  })
 })
 
 describe("util.token.estimate", () => {

packages/opencode/test/session/llm.test.ts

@@ -1,7 +1,9 @@
 import { afterAll, beforeAll, beforeEach, describe, expect, test } from "bun:test"
 import path from "path"
 import { tool, type ModelMessage } from "ai"
+import { Cause, Exit, Stream } from "effect"
 import z from "zod"
+import { makeRuntime } from "../../src/effect/run-service"
 import { LLM } from "../../src/session/llm"
 import { Instance } from "../../src/project/instance"
 import { Provider } from "../../src/provider/provider"
@@ -109,7 +111,11 @@ type Capture = {
 
 const state = {
   server: null as ReturnType<typeof Bun.serve> | null,
-  queue: [] as Array<{ path: string; response: Response; resolve: (value: Capture) => void }>,
+  queue: [] as Array<{
+    path: string
+    response: Response | ((req: Request, capture: Capture) => Response)
+    resolve: (value: Capture) => void
+  }>,
 }
 
 function deferred<T>() {
@@ -126,6 +132,58 @@ function waitRequest(pathname: string, response: Response) {
   return pending.promise
 }
 
+function timeout(ms: number) {
+  return new Promise<never>((_, reject) => {
+    setTimeout(() => reject(new Error(`timed out after ${ms}ms`)), ms)
+  })
+}
+
+function waitStreamingRequest(pathname: string) {
+  const request = deferred<Capture>()
+  const requestAborted = deferred<void>()
+  const responseCanceled = deferred<void>()
+  const encoder = new TextEncoder()
+
+  state.queue.push({
+    path: pathname,
+    resolve: request.resolve,
+    response(req: Request) {
+      req.signal.addEventListener("abort", () => requestAborted.resolve(), { once: true })
+
+      return new Response(
+        new ReadableStream<Uint8Array>({
+          start(controller) {
+            controller.enqueue(
+              encoder.encode(
+                [
+                  `data: ${JSON.stringify({
+                    id: "chatcmpl-abort",
+                    object: "chat.completion.chunk",
+                    choices: [{ delta: { role: "assistant" } }],
+                  })}`,
+                ].join("\n\n") + "\n\n",
+              ),
+            )
+          },
+          cancel() {
+            responseCanceled.resolve()
+          },
+        }),
+        {
+          status: 200,
+          headers: { "Content-Type": "text/event-stream" },
+        },
+      )
+    },
+  })
+
+  return {
+    request: request.promise,
+    requestAborted: requestAborted.promise,
+    responseCanceled: responseCanceled.promise,
+  }
+}
+
 beforeAll(() => {
   state.server = Bun.serve({
     port: 0,
@@ -143,7 +201,9 @@ beforeAll(() => {
         return new Response("not found", { status: 404 })
       }
 
-      return next.response
+      return typeof next.response === "function"
+        ? next.response(req, { url, headers: req.headers, body })
+        : next.response
     },
   })
 })
@@ -325,6 +385,162 @@ describe("session.llm.stream", () => {
     })
   })
 
+  test("raw stream abort signal cancels provider response body promptly", async () => {
+    const server = state.server
+    if (!server) throw new Error("Server not initialized")
+
+    const providerID = "alibaba"
+    const modelID = "qwen-plus"
+    const fixture = await loadFixture(providerID, modelID)
+    const model = fixture.model
+    const pending = waitStreamingRequest("/chat/completions")
+
+    await using tmp = await tmpdir({
+      init: async (dir) => {
+        await Bun.write(
+          path.join(dir, "opencode.json"),
+          JSON.stringify({
+            $schema: "https://opencode.ai/config.json",
+            enabled_providers: [providerID],
+            provider: {
+              [providerID]: {
+                options: {
+                  apiKey: "test-key",
+                  baseURL: `${server.url.origin}/v1`,
+                },
+              },
+            },
+          }),
+        )
+      },
+    })
+
+    await Instance.provide({
+      directory: tmp.path,
+      fn: async () => {
+        const resolved = await Provider.getModel(ProviderID.make(providerID), ModelID.make(model.id))
+        const sessionID = SessionID.make("session-test-raw-abort")
+        const agent = {
+          name: "test",
+          mode: "primary",
+          options: {},
+          permission: [{ permission: "*", pattern: "*", action: "allow" }],
+        } satisfies Agent.Info
+        const user = {
+          id: MessageID.make("user-raw-abort"),
+          sessionID,
+          role: "user",
+          time: { created: Date.now() },
+          agent: agent.name,
+          model: { providerID: ProviderID.make(providerID), modelID: resolved.id },
+        } satisfies MessageV2.User
+
+        const ctrl = new AbortController()
+        const result = await LLM.stream({
+          user,
+          sessionID,
+          model: resolved,
+          agent,
+          system: ["You are a helpful assistant."],
+          abort: ctrl.signal,
+          messages: [{ role: "user", content: "Hello" }],
+          tools: {},
+        })
+
+        const iter = result.fullStream[Symbol.asyncIterator]()
+        await pending.request
+        await iter.next()
+        ctrl.abort()
+
+        await Promise.race([pending.responseCanceled, timeout(500)])
+        await Promise.race([pending.requestAborted, timeout(500)]).catch(() => undefined)
+        await iter.return?.()
+      },
+    })
+  })
+
+  test("service stream cancellation cancels provider response body promptly", async () => {
+    const server = state.server
+    if (!server) throw new Error("Server not initialized")
+
+    const providerID = "alibaba"
+    const modelID = "qwen-plus"
+    const fixture = await loadFixture(providerID, modelID)
+    const model = fixture.model
+    const pending = waitStreamingRequest("/chat/completions")
+
+    await using tmp = await tmpdir({
+      init: async (dir) => {
+        await Bun.write(
+          path.join(dir, "opencode.json"),
+          JSON.stringify({
+            $schema: "https://opencode.ai/config.json",
+            enabled_providers: [providerID],
+            provider: {
+              [providerID]: {
+                options: {
+                  apiKey: "test-key",
+                  baseURL: `${server.url.origin}/v1`,
+                },
+              },
+            },
+          }),
+        )
+      },
+    })
+
+    await Instance.provide({
+      directory: tmp.path,
+      fn: async () => {
+        const resolved = await Provider.getModel(ProviderID.make(providerID), ModelID.make(model.id))
+        const sessionID = SessionID.make("session-test-service-abort")
+        const agent = {
+          name: "test",
+          mode: "primary",
+          options: {},
+          permission: [{ permission: "*", pattern: "*", action: "allow" }],
+        } satisfies Agent.Info
+        const user = {
+          id: MessageID.make("user-service-abort"),
+          sessionID,
+          role: "user",
+          time: { created: Date.now() },
+          agent: agent.name,
+          model: { providerID: ProviderID.make(providerID), modelID: resolved.id },
+        } satisfies MessageV2.User
+
+        const ctrl = new AbortController()
+        const { runPromiseExit } = makeRuntime(LLM.Service, LLM.defaultLayer)
+        const run = runPromiseExit(
+          (svc) =>
+            svc
+              .stream({
+                user,
+                sessionID,
+                model: resolved,
+                agent,
+                system: ["You are a helpful assistant."],
+                messages: [{ role: "user", content: "Hello" }],
+                tools: {},
+              })
+              .pipe(Stream.runDrain),
+          { signal: ctrl.signal },
+        )
+
+        await pending.request
+        ctrl.abort()
+
+        await Promise.race([pending.responseCanceled, timeout(500)])
+        const exit = await run
+        expect(Exit.isFailure(exit)).toBe(true)
+        if (Exit.isFailure(exit)) {
+          expect(Cause.hasInterrupts(exit.cause)).toBe(true)
+        }
+        await Promise.race([pending.requestAborted, timeout(500)]).catch(() => undefined)
+      },
+    })
+  })
+
   test("keeps tools enabled by prompt permissions", async () => {
     const server = state.server
     if (!server) {

packages/opencode/test/session/processor-effect.test.ts

@@ -1,7 +1,7 @@
 import { NodeFileSystem } from "@effect/platform-node"
 import { expect } from "bun:test"
 import { APICallError } from "ai"
-import { Effect, Layer, ServiceMap } from "effect"
+import { Cause, Effect, Exit, Fiber, Layer, ServiceMap } from "effect"
 import * as Stream from "effect/Stream"
 import path from "path"
 import type { Agent } from "../../src/agent/agent"
@@ -10,7 +10,6 @@ import { Bus } from "../../src/bus"
 import { Config } from "../../src/config/config"
 import { Permission } from "../../src/permission"
 import { Plugin } from "../../src/plugin"
-import { Instance } from "../../src/project/instance"
 import type { Provider } from "../../src/provider/provider"
 import { ModelID, ProviderID } from "../../src/provider/schema"
 import { Session } from "../../src/session"
@@ -120,21 +119,8 @@ function fail<E>(err: E, ...items: LLM.Event[]) {
   return stream(...items).pipe(Stream.concat(Stream.fail(err)))
 }
 
-function wait(abort: AbortSignal) {
-  return Effect.promise(
-    () =>
-      new Promise<void>((done) => {
-        abort.addEventListener("abort", () => done(), { once: true })
-      }),
-  )
-}
-
-function hang(input: LLM.StreamInput, ...items: LLM.Event[]) {
-  return stream(...items).pipe(
-    Stream.concat(
-      Stream.unwrap(wait(input.abort).pipe(Effect.as(Stream.fail(new DOMException("Aborted", "AbortError"))))),
-    ),
-  )
+function hang(_input: LLM.StreamInput, ...items: LLM.Event[]) {
+  return stream(...items).pipe(Stream.concat(Stream.fromEffect(Effect.never)))
 }
 
 function model(context: number): Provider.Model {
@@ -291,13 +277,11 @@ it.effect("session.processor effect tests capture llm input cleanly", () => {
         const chat = yield* session.create({})
         const parent = yield* user(chat.id, "hi")
         const msg = yield* assistant(chat.id, parent.id, path.resolve(dir))
-        const abort = new AbortController()
         const mdl = model(100)
         const handle = yield* processors.create({
           assistantMessage: msg,
           sessionID: chat.id,
           model: mdl,
-          abort: abort.signal,
         })
 
         const input = {
@@ -313,7 +297,6 @@ it.effect("session.processor effect tests capture llm input cleanly", () => {
           model: mdl,
           agent: agent(),
           system: [],
-          abort: abort.signal,
           messages: [{ role: "user", content: "hi" }],
           tools: {},
         } satisfies LLM.StreamInput
@@ -359,13 +342,11 @@ it.effect("session.processor effect tests stop after token overflow requests com
         const chat = yield* session.create({})
         const parent = yield* user(chat.id, "compact")
         const msg = yield* assistant(chat.id, parent.id, path.resolve(dir))
-        const abort = new AbortController()
         const mdl = model(20)
         const handle = yield* processors.create({
           assistantMessage: msg,
           sessionID: chat.id,
           model: mdl,
-          abort: abort.signal,
         })
 
         const value = yield* handle.process({
@@ -381,7 +362,6 @@ it.effect("session.processor effect tests stop after token overflow requests com
           model: mdl,
           agent: agent(),
           system: [],
-          abort: abort.signal,
           messages: [{ role: "user", content: "compact" }],
           tools: {},
         })
@@ -433,13 +413,11 @@ it.effect("session.processor effect tests reset reasoning state across retries",
         const chat = yield* session.create({})
         const parent = yield* user(chat.id, "reason")
         const msg = yield* assistant(chat.id, parent.id, path.resolve(dir))
-        const abort = new AbortController()
         const mdl = model(100)
         const handle = yield* processors.create({
           assistantMessage: msg,
           sessionID: chat.id,
           model: mdl,
-          abort: abort.signal,
         })
 
         const value = yield* handle.process({
@@ -455,7 +433,6 @@ it.effect("session.processor effect tests reset reasoning state across retries",
           model: mdl,
           agent: agent(),
           system: [],
-          abort: abort.signal,
           messages: [{ role: "user", content: "reason" }],
           tools: {},
         })
@@ -485,13 +462,11 @@ it.effect("session.processor effect tests do not retry unknown json errors", ()
         const chat = yield* session.create({})
         const parent = yield* user(chat.id, "json")
         const msg = yield* assistant(chat.id, parent.id, path.resolve(dir))
-        const abort = new AbortController()
         const mdl = model(100)
         const handle = yield* processors.create({
           assistantMessage: msg,
           sessionID: chat.id,
           model: mdl,
-          abort: abort.signal,
         })
 
         const value = yield* handle.process({
@@ -507,7 +482,6 @@ it.effect("session.processor effect tests do not retry unknown json errors", ()
           model: mdl,
           agent: agent(),
           system: [],
-          abort: abort.signal,
           messages: [{ role: "user", content: "json" }],
           tools: {},
         })
@@ -535,13 +509,11 @@ it.effect("session.processor effect tests retry recognized structured json error
         const chat = yield* session.create({})
         const parent = yield* user(chat.id, "retry json")
         const msg = yield* assistant(chat.id, parent.id, path.resolve(dir))
-        const abort = new AbortController()
         const mdl = model(100)
         const handle = yield* processors.create({
           assistantMessage: msg,
           sessionID: chat.id,
           model: mdl,
-          abort: abort.signal,
         })
 
         const value = yield* handle.process({
@@ -557,7 +529,6 @@ it.effect("session.processor effect tests retry recognized structured json error
           model: mdl,
           agent: agent(),
           system: [],
-          abort: abort.signal,
           messages: [{ role: "user", content: "retry json" }],
           tools: {},
         })
@@ -601,7 +572,6 @@ it.effect("session.processor effect tests publish retry status updates", () => {
         const chat = yield* session.create({})
         const parent = yield* user(chat.id, "retry")
         const msg = yield* assistant(chat.id, parent.id, path.resolve(dir))
-        const abort = new AbortController()
         const mdl = model(100)
         const states: number[] = []
         const off = yield* bus.subscribeCallback(SessionStatus.Event.Status, (evt) => {
@@ -612,7 +582,6 @@ it.effect("session.processor effect tests publish retry status updates", () => {
           assistantMessage: msg,
           sessionID: chat.id,
           model: mdl,
-          abort: abort.signal,
         })
 
         const value = yield* handle.process({
@@ -628,7 +597,6 @@ it.effect("session.processor effect tests publish retry status updates", () => {
           model: mdl,
           agent: agent(),
           system: [],
-          abort: abort.signal,
           messages: [{ role: "user", content: "retry" }],
           tools: {},
         })
@@ -656,13 +624,11 @@ it.effect("session.processor effect tests compact on structured context overflow
         const chat = yield* session.create({})
         const parent = yield* user(chat.id, "compact json")
         const msg = yield* assistant(chat.id, parent.id, path.resolve(dir))
-        const abort = new AbortController()
         const mdl = model(100)
         const handle = yield* processors.create({
           assistantMessage: msg,
           sessionID: chat.id,
           model: mdl,
-          abort: abort.signal,
         })
 
         const value = yield* handle.process({
@@ -678,7 +644,6 @@ it.effect("session.processor effect tests compact on structured context overflow
           model: mdl,
           agent: agent(),
           system: [],
-          abort: abort.signal,
           messages: [{ role: "user", content: "compact json" }],
           tools: {},
         })
@@ -696,7 +661,6 @@ it.effect("session.processor effect tests mark pending tools as aborted on clean
     (dir) =>
       Effect.gen(function* () {
         const ready = defer<void>()
-        const seen = defer<void>()
         const test = yield* TestLLM
         const processors = yield* SessionProcessor.Service
         const session = yield* Session.Service
@@ -710,17 +674,15 @@ it.effect("session.processor effect tests mark pending tools as aborted on clean
         const chat = yield* session.create({})
         const parent = yield* user(chat.id, "tool abort")
         const msg = yield* assistant(chat.id, parent.id, path.resolve(dir))
-        const abort = new AbortController()
         const mdl = model(100)
         const handle = yield* processors.create({
           assistantMessage: msg,
           sessionID: chat.id,
           model: mdl,
-          abort: abort.signal,
         })
 
-        const run = Effect.runPromise(
-          handle.process({
+        const run = yield* handle
+          .process({
             user: {
               id: parent.id,
               sessionID: chat.id,
@@ -733,20 +695,25 @@ it.effect("session.processor effect tests mark pending tools as aborted on clean
             model: mdl,
             agent: agent(),
             system: [],
-            abort: abort.signal,
             messages: [{ role: "user", content: "tool abort" }],
             tools: {},
-          }),
-        )
+          })
+          .pipe(Effect.forkChild)
 
         yield* Effect.promise(() => ready.promise)
-        abort.abort()
+        yield* Fiber.interrupt(run)
 
-        const value = yield* Effect.promise(() => run)
+        const exit = yield* Fiber.await(run)
+        if (Exit.isFailure(exit) && Cause.hasInterruptsOnly(exit.cause)) {
+          yield* handle.abort()
+        }
         const parts = yield* Effect.promise(() => MessageV2.parts(msg.id))
         const tool = parts.find((part): part is MessageV2.ToolPart => part.type === "tool")
 
-        expect(value).toBe("stop")
+        expect(Exit.isFailure(exit)).toBe(true)
+        if (Exit.isFailure(exit)) {
+          expect(Cause.hasInterruptsOnly(exit.cause)).toBe(true)
+        }
         expect(yield* test.calls).toBe(1)
         expect(tool?.state.status).toBe("error")
         if (tool?.state.status === "error") {
@@ -779,7 +746,6 @@ it.effect("session.processor effect tests record aborted errors and idle state",
         const chat = yield* session.create({})
         const parent = yield* user(chat.id, "abort")
         const msg = yield* assistant(chat.id, parent.id, path.resolve(dir))
-        const abort = new AbortController()
         const mdl = model(100)
         const errs: string[] = []
         const off = yield* bus.subscribeCallback(Session.Event.Error, (evt) => {
@@ -792,11 +758,10 @@ it.effect("session.processor effect tests record aborted errors and idle state",
           assistantMessage: msg,
           sessionID: chat.id,
           model: mdl,
-          abort: abort.signal,
         })
 
-        const run = Effect.runPromise(
-          handle.process({
+        const run = yield* handle
+          .process({
             user: {
               id: parent.id,
               sessionID: chat.id,
@@ -809,22 +774,27 @@ it.effect("session.processor effect tests record aborted errors and idle state",
             model: mdl,
             agent: agent(),
             system: [],
-            abort: abort.signal,
             messages: [{ role: "user", content: "abort" }],
             tools: {},
-          }),
-        )
+          })
+          .pipe(Effect.forkChild)
 
         yield* Effect.promise(() => ready.promise)
-        abort.abort()
+        yield* Fiber.interrupt(run)
 
-        const value = yield* Effect.promise(() => run)
+        const exit = yield* Fiber.await(run)
+        if (Exit.isFailure(exit) && Cause.hasInterruptsOnly(exit.cause)) {
+          yield* handle.abort()
+        }
         yield* Effect.promise(() => seen.promise)
         const stored = yield* Effect.promise(() => MessageV2.get({ sessionID: chat.id, messageID: msg.id }))
         const state = yield* status.get(chat.id)
         off()
 
-        expect(value).toBe("stop")
+        expect(Exit.isFailure(exit)).toBe(true)
+        if (Exit.isFailure(exit)) {
+          expect(Cause.hasInterruptsOnly(exit.cause)).toBe(true)
+        }
         expect(handle.message.error?.name).toBe("MessageAbortedError")
         expect(stored.info.role).toBe("assistant")
         if (stored.info.role === "assistant") {
@@ -836,3 +806,67 @@ it.effect("session.processor effect tests record aborted errors and idle state",
     { git: true },
   )
 })
+
+it.effect("session.processor effect tests mark interruptions aborted without manual abort", () => {
+  return provideTmpdirInstance(
+    (dir) =>
+      Effect.gen(function* () {
+        const ready = defer<void>()
+        const processors = yield* SessionProcessor.Service
+        const session = yield* Session.Service
+        const status = yield* SessionStatus.Service
+        const test = yield* TestLLM
+
+        yield* test.push((input) =>
+          hang(input, start()).pipe(
+            Stream.tap((event) => (event.type === "start" ? Effect.sync(() => ready.resolve()) : Effect.void)),
+          ),
+        )
+
+        const chat = yield* session.create({})
+        const parent = yield* user(chat.id, "interrupt")
+        const msg = yield* assistant(chat.id, parent.id, path.resolve(dir))
+        const mdl = model(100)
+        const handle = yield* processors.create({
+          assistantMessage: msg,
+          sessionID: chat.id,
+          model: mdl,
+        })
+
+        const run = yield* handle
+          .process({
+            user: {
+              id: parent.id,
+              sessionID: chat.id,
+              role: "user",
+              time: parent.time,
+              agent: parent.agent,
+              model: { providerID: ref.providerID, modelID: ref.modelID },
+            } satisfies MessageV2.User,
+            sessionID: chat.id,
+            model: mdl,
+            agent: agent(),
+            system: [],
+            messages: [{ role: "user", content: "interrupt" }],
+            tools: {},
+          })
+          .pipe(Effect.forkChild)
+
+        yield* Effect.promise(() => ready.promise)
+        yield* Fiber.interrupt(run)
+
+        const exit = yield* Fiber.await(run)
+        const stored = yield* Effect.promise(() => MessageV2.get({ sessionID: chat.id, messageID: msg.id }))
+        const state = yield* status.get(chat.id)
+
+        expect(Exit.isFailure(exit)).toBe(true)
+        expect(handle.message.error?.name).toBe("MessageAbortedError")
+        expect(stored.info.role).toBe("assistant")
+        if (stored.info.role === "assistant") {
+          expect(stored.info.error?.name).toBe("MessageAbortedError")
+        }
+        expect(state).toMatchObject({ type: "idle" })
+      }),
+    { git: true },
+  )
+})

packages/opencode/test/session/prompt-concurrency.test.ts

@@ -0,0 +1,247 @@
+import { describe, expect, spyOn, test } from "bun:test"
+import { Instance } from "../../src/project/instance"
+import { Provider } from "../../src/provider/provider"
+import { Session } from "../../src/session"
+import { MessageV2 } from "../../src/session/message-v2"
+import { SessionPrompt } from "../../src/session/prompt"
+import { SessionStatus } from "../../src/session/status"
+import { MessageID, PartID, SessionID } from "../../src/session/schema"
+import { Log } from "../../src/util/log"
+import { tmpdir } from "../fixture/fixture"
+
+Log.init({ print: false })
+
+function deferred() {
+  let resolve!: () => void
+  const promise = new Promise<void>((done) => {
+    resolve = done
+  })
+  return { promise, resolve }
+}
+
+// Helper: seed a session with a user message + finished assistant message
+// so loop() exits immediately without calling any LLM
+async function seed(sessionID: SessionID) {
+  const userMsg: MessageV2.Info = {
+    id: MessageID.ascending(),
+    role: "user",
+    sessionID,
+    time: { created: Date.now() },
+    agent: "build",
+    model: { providerID: "openai" as any, modelID: "gpt-5.2" as any },
+  }
+  await Session.updateMessage(userMsg)
+  await Session.updatePart({
+    id: PartID.ascending(),
+    messageID: userMsg.id,
+    sessionID,
+    type: "text",
+    text: "hello",
+  })
+
+  const assistantMsg: MessageV2.Info = {
+    id: MessageID.ascending(),
+    role: "assistant",
+    parentID: userMsg.id,
+    sessionID,
+    mode: "build",
+    agent: "build",
+    cost: 0,
+    path: { cwd: "/tmp", root: "/tmp" },
+    tokens: { input: 0, output: 0, reasoning: 0, cache: { read: 0, write: 0 } },
+    modelID: "gpt-5.2" as any,
+    providerID: "openai" as any,
+    time: { created: Date.now(), completed: Date.now() },
+    finish: "stop",
+  }
+  await Session.updateMessage(assistantMsg)
+  await Session.updatePart({
+    id: PartID.ascending(),
+    messageID: assistantMsg.id,
+    sessionID,
+    type: "text",
+    text: "hi there",
+  })
+
+  return { userMsg, assistantMsg }
+}
+
+describe("session.prompt concurrency", () => {
+  test("loop returns assistant message and sets status to idle", async () => {
+    await using tmp = await tmpdir({ git: true })
+    await Instance.provide({
+      directory: tmp.path,
+      fn: async () => {
+        const session = await Session.create({})
+        await seed(session.id)
+
+        const result = await SessionPrompt.loop({ sessionID: session.id })
+        expect(result.info.role).toBe("assistant")
+        if (result.info.role === "assistant") expect(result.info.finish).toBe("stop")
+
+        const status = await SessionStatus.get(session.id)
+        expect(status.type).toBe("idle")
+      },
+    })
+  })
+
+  test("concurrent loop callers get the same result", async () => {
+    await using tmp = await tmpdir({ git: true })
+    await Instance.provide({
+      directory: tmp.path,
+      fn: async () => {
+        const session = await Session.create({})
+        await seed(session.id)
+
+        const [a, b] = await Promise.all([
+          SessionPrompt.loop({ sessionID: session.id }),
+          SessionPrompt.loop({ sessionID: session.id }),
+        ])
+
+        expect(a.info.id).toBe(b.info.id)
+        expect(a.info.role).toBe("assistant")
+      },
+    })
+  })
+
+  test("assertNotBusy throws when loop is running", async () => {
+    await using tmp = await tmpdir({ git: true })
+    await Instance.provide({
+      directory: tmp.path,
+      fn: async () => {
+        const session = await Session.create({})
+        const userMsg: MessageV2.Info = {
+          id: MessageID.ascending(),
+          role: "user",
+          sessionID: session.id,
+          time: { created: Date.now() },
+          agent: "build",
+          model: { providerID: "openai" as any, modelID: "gpt-5.2" as any },
+        }
+        await Session.updateMessage(userMsg)
+        await Session.updatePart({
+          id: PartID.ascending(),
+          messageID: userMsg.id,
+          sessionID: session.id,
+          type: "text",
+          text: "hello",
+        })
+
+        const ready = deferred()
+        const gate = deferred()
+        const getModel = spyOn(Provider, "getModel").mockImplementation(async () => {
+          ready.resolve()
+          await gate.promise
+          throw new Error("test stop")
+        })
+
+        try {
+          const loopPromise = SessionPrompt.loop({ sessionID: session.id }).catch(() => undefined)
+          await ready.promise
+
+          await expect(SessionPrompt.assertNotBusy(session.id)).rejects.toBeInstanceOf(Session.BusyError)
+
+          gate.resolve()
+          await loopPromise
+        } finally {
+          gate.resolve()
+          getModel.mockRestore()
+        }
+
+        // After loop completes, assertNotBusy should succeed
+        await SessionPrompt.assertNotBusy(session.id)
+      },
+    })
+  })
+
+  test("cancel sets status to idle", async () => {
+    await using tmp = await tmpdir({ git: true })
+    await Instance.provide({
+      directory: tmp.path,
+      fn: async () => {
+        const session = await Session.create({})
+        // Seed only a user message — loop must call getModel to proceed
+        const userMsg: MessageV2.Info = {
+          id: MessageID.ascending(),
+          role: "user",
+          sessionID: session.id,
+          time: { created: Date.now() },
+          agent: "build",
+          model: { providerID: "openai" as any, modelID: "gpt-5.2" as any },
+        }
+        await Session.updateMessage(userMsg)
+        await Session.updatePart({
+          id: PartID.ascending(),
+          messageID: userMsg.id,
+          sessionID: session.id,
+          type: "text",
+          text: "hello",
+        })
+        // Also seed an assistant message so lastAssistant() fallback can find it
+        const assistantMsg: MessageV2.Info = {
+          id: MessageID.ascending(),
+          role: "assistant",
+          parentID: userMsg.id,
+          sessionID: session.id,
+          mode: "build",
+          agent: "build",
+          cost: 0,
+          path: { cwd: "/tmp", root: "/tmp" },
+          tokens: { input: 0, output: 0, reasoning: 0, cache: { read: 0, write: 0 } },
+          modelID: "gpt-5.2" as any,
+          providerID: "openai" as any,
+          time: { created: Date.now() },
+        }
+        await Session.updateMessage(assistantMsg)
+        await Session.updatePart({
+          id: PartID.ascending(),
+          messageID: assistantMsg.id,
+          sessionID: session.id,
+          type: "text",
+          text: "hi there",
+        })
+
+        const ready = deferred()
+        const gate = deferred()
+        const getModel = spyOn(Provider, "getModel").mockImplementation(async () => {
+          ready.resolve()
+          await gate.promise
+          throw new Error("test stop")
+        })
+
+        try {
+          // Start loop — it will block in getModel (assistant has no finish, so loop continues)
+          const loopPromise = SessionPrompt.loop({ sessionID: session.id })
+
+          await ready.promise
+
+          await SessionPrompt.cancel(session.id)
+
+          const status = await SessionStatus.get(session.id)
+          expect(status.type).toBe("idle")
+
+          // loop should resolve cleanly, not throw "All fibers interrupted"
+          const result = await loopPromise
+          expect(result.info.role).toBe("assistant")
+          expect(result.info.id).toBe(assistantMsg.id)
+        } finally {
+          gate.resolve()
+          getModel.mockRestore()
+        }
+      },
+    })
+  }, 10000)
+
+  test("cancel on idle session just sets idle", async () => {
+    await using tmp = await tmpdir({ git: true })
+    await Instance.provide({
+      directory: tmp.path,
+      fn: async () => {
+        const session = await Session.create({})
+        await SessionPrompt.cancel(session.id)
+        const status = await SessionStatus.get(session.id)
+        expect(status.type).toBe("idle")
+      },
+    })
+  })
+})

packages/opencode/test/storage/storage.test.ts

@@ -0,0 +1,295 @@
+import { describe, expect, test } from "bun:test"
+import fs from "fs/promises"
+import path from "path"
+import { Effect, Layer, ManagedRuntime } from "effect"
+import { AppFileSystem } from "../../src/filesystem"
+import { Global } from "../../src/global"
+import { Storage } from "../../src/storage/storage"
+import { tmpdir } from "../fixture/fixture"
+
+const dir = path.join(Global.Path.data, "storage")
+
+async function withScope<T>(fn: (root: string[]) => Promise<T>) {
+  const root = ["storage_test", crypto.randomUUID()]
+  try {
+    return await fn(root)
+  } finally {
+    await fs.rm(path.join(dir, ...root), { recursive: true, force: true })
+  }
+}
+
+function map(root: string, file: string) {
+  if (file === Global.Path.data) return root
+  if (file.startsWith(Global.Path.data + path.sep)) return path.join(root, path.relative(Global.Path.data, file))
+  return file
+}
+
+function layer(root: string) {
+  return Layer.effect(
+    AppFileSystem.Service,
+    Effect.gen(function* () {
+      const fs = yield* AppFileSystem.Service
+      return AppFileSystem.Service.of({
+        ...fs,
+        isDir: (file) => fs.isDir(map(root, file)),
+        readJson: (file) => fs.readJson(map(root, file)),
+        writeWithDirs: (file, content, mode) => fs.writeWithDirs(map(root, file), content, mode),
+        readFileString: (file) => fs.readFileString(map(root, file)),
+        remove: (file) => fs.remove(map(root, file)),
+        glob: (pattern, options) =>
+          fs.glob(pattern, options?.cwd ? { ...options, cwd: map(root, options.cwd) } : options),
+      })
+    }),
+  ).pipe(Layer.provide(AppFileSystem.defaultLayer))
+}
+
+async function withStorage<T>(
+  root: string,
+  fn: (run: <A, E>(body: Effect.Effect<A, E, Storage.Service>) => Promise<A>) => Promise<T>,
+) {
+  const rt = ManagedRuntime.make(Storage.layer.pipe(Layer.provide(layer(root))))
+  try {
+    return await fn((body) => rt.runPromise(body))
+  } finally {
+    await rt.dispose()
+  }
+}
+
+async function write(file: string, value: unknown) {
+  await fs.mkdir(path.dirname(file), { recursive: true })
+  await Bun.write(file, JSON.stringify(value, null, 2))
+}
+
+async function text(file: string, value: string) {
+  await fs.mkdir(path.dirname(file), { recursive: true })
+  await Bun.write(file, value)
+}
+
+async function exists(file: string) {
+  return fs
+    .stat(file)
+    .then(() => true)
+    .catch(() => false)
+}
+
+describe("Storage", () => {
+  test("round-trips JSON content", async () => {
+    await withScope(async (root) => {
+      const key = [...root, "session_diff", "roundtrip"]
+      const value = [{ file: "a.ts", additions: 2, deletions: 1 }]
+
+      await Storage.write(key, value)
+
+      expect(await Storage.read<typeof value>(key)).toEqual(value)
+    })
+  })
+
+  test("maps missing reads to NotFoundError", async () => {
+    await withScope(async (root) => {
+      await expect(Storage.read([...root, "missing", "value"])).rejects.toMatchObject({ name: "NotFoundError" })
+    })
+  })
+
+  test("update on missing key throws NotFoundError", async () => {
+    await withScope(async (root) => {
+      await expect(
+        Storage.update<{ value: number }>([...root, "missing", "key"], (draft) => {
+          draft.value += 1
+        }),
+      ).rejects.toMatchObject({ name: "NotFoundError" })
+    })
+  })
+
+  test("write overwrites existing value", async () => {
+    await withScope(async (root) => {
+      const key = [...root, "overwrite", "test"]
+      await Storage.write<{ v: number }>(key, { v: 1 })
+      await Storage.write<{ v: number }>(key, { v: 2 })
+
+      expect(await Storage.read<{ v: number }>(key)).toEqual({ v: 2 })
+    })
+  })
+
+  test("remove on missing key is a no-op", async () => {
+    await withScope(async (root) => {
+      await expect(Storage.remove([...root, "nonexistent", "key"])).resolves.toBeUndefined()
+    })
+  })
+
+  test("list on missing prefix returns empty", async () => {
+    await withScope(async (root) => {
+      expect(await Storage.list([...root, "nonexistent"])).toEqual([])
+    })
+  })
+
+  test("serializes concurrent updates for the same key", async () => {
+    await withScope(async (root) => {
+      const key = [...root, "counter", "shared"]
+      await Storage.write(key, { value: 0 })
+
+      await Promise.all(
+        Array.from({ length: 25 }, () =>
+          Storage.update<{ value: number }>(key, (draft) => {
+            draft.value += 1
+          }),
+        ),
+      )
+
+      expect(await Storage.read<{ value: number }>(key)).toEqual({ value: 25 })
+    })
+  })
+
+  test("concurrent reads do not block each other", async () => {
+    await withScope(async (root) => {
+      const key = [...root, "concurrent", "reads"]
+      await Storage.write(key, { ok: true })
+
+      const results = await Promise.all(Array.from({ length: 10 }, () => Storage.read(key)))
+
+      expect(results).toHaveLength(10)
+      for (const r of results) expect(r).toEqual({ ok: true })
+    })
+  })
+
+  test("nested keys create deep paths", async () => {
+    await withScope(async (root) => {
+      const key = [...root, "a", "b", "c", "deep"]
+      await Storage.write<{ nested: boolean }>(key, { nested: true })
+
+      expect(await Storage.read<{ nested: boolean }>(key)).toEqual({ nested: true })
+      expect(await Storage.list([...root, "a"])).toEqual([key])
+    })
+  })
+
+  test("lists and removes stored entries", async () => {
+    await withScope(async (root) => {
+      const a = [...root, "list", "a"]
+      const b = [...root, "list", "b"]
+      const prefix = [...root, "list"]
+
+      await Storage.write(b, { value: 2 })
+      await Storage.write(a, { value: 1 })
+
+      expect(await Storage.list(prefix)).toEqual([a, b])
+
+      await Storage.remove(a)
+
+      expect(await Storage.list(prefix)).toEqual([b])
+      await expect(Storage.read(a)).rejects.toMatchObject({ name: "NotFoundError" })
+    })
+  })
+
+  test("migration 2 runs when marker contents are invalid", async () => {
+    await using tmp = await tmpdir()
+    const storage = path.join(tmp.path, "storage")
+    const diffs = [
+      { additions: 2, deletions: 1 },
+      { additions: 3, deletions: 4 },
+    ]
+
+    await text(path.join(storage, "migration"), "wat")
+    await write(path.join(storage, "session", "proj_test", "ses_test.json"), {
+      id: "ses_test",
+      projectID: "proj_test",
+      title: "legacy",
+      summary: { diffs },
+    })
+
+    await withStorage(tmp.path, async (run) => {
+      expect(await run(Storage.Service.use((svc) => svc.list(["session_diff"])))).toEqual([
+        ["session_diff", "ses_test"],
+      ])
+      expect(await run(Storage.Service.use((svc) => svc.read<typeof diffs>(["session_diff", "ses_test"])))).toEqual(
+        diffs,
+      )
+      expect(
+        await run(
+          Storage.Service.use((svc) =>
+            svc.read<{
+              id: string
+              projectID: string
+              title: string
+              summary: {
+                additions: number
+                deletions: number
+              }
+            }>(["session", "proj_test", "ses_test"]),
+          ),
+        ),
+      ).toEqual({
+        id: "ses_test",
+        projectID: "proj_test",
+        title: "legacy",
+        summary: {
+          additions: 5,
+          deletions: 5,
+        },
+      })
+    })
+
+    expect(await Bun.file(path.join(storage, "migration")).text()).toBe("2")
+  })
+
+  test("migration 1 tolerates malformed legacy records", async () => {
+    await using tmp = await tmpdir({ git: true })
+    const storage = path.join(tmp.path, "storage")
+    const legacy = path.join(tmp.path, "project", "legacy")
+
+    await write(path.join(legacy, "storage", "session", "message", "probe", "0.json"), [])
+    await write(path.join(legacy, "storage", "session", "message", "probe", "1.json"), {
+      path: { root: tmp.path },
+    })
+    await write(path.join(legacy, "storage", "session", "info", "ses_legacy.json"), {
+      id: "ses_legacy",
+      title: "legacy",
+    })
+    await write(path.join(legacy, "storage", "session", "message", "ses_legacy", "msg_legacy.json"), {
+      role: "user",
+      text: "hello",
+    })
+
+    await withStorage(tmp.path, async (run) => {
+      const projects = await run(Storage.Service.use((svc) => svc.list(["project"])))
+      expect(projects).toHaveLength(1)
+      const project = projects[0]![1]
+
+      expect(await run(Storage.Service.use((svc) => svc.list(["session", project])))).toEqual([
+        ["session", project, "ses_legacy"],
+      ])
+      expect(
+        await run(
+          Storage.Service.use((svc) => svc.read<{ id: string; title: string }>(["session", project, "ses_legacy"])),
+        ),
+      ).toEqual({
+        id: "ses_legacy",
+        title: "legacy",
+      })
+      expect(
+        await run(
+          Storage.Service.use((svc) =>
+            svc.read<{ role: string; text: string }>(["message", "ses_legacy", "msg_legacy"]),
+          ),
+        ),
+      ).toEqual({
+        role: "user",
+        text: "hello",
+      })
+    })
+
+    expect(await Bun.file(path.join(storage, "migration")).text()).toBe("2")
+  })
+
+  test("failed migrations do not advance the marker", async () => {
+    await using tmp = await tmpdir()
+    const storage = path.join(tmp.path, "storage")
+    const legacy = path.join(tmp.path, "project", "legacy")
+
+    await text(path.join(legacy, "storage", "session", "message", "probe", "0.json"), "{")
+
+    await withStorage(tmp.path, async (run) => {
+      expect(await run(Storage.Service.use((svc) => svc.list(["project"])))).toEqual([])
+    })
+
+    expect(await exists(path.join(storage, "migration"))).toBe(false)
+  })
+})

packages/plugin/package.json

@@ -1,7 +1,7 @@
 {
   "$schema": "https://json.schemastore.org/package.json",
   "name": "@opencode-ai/plugin",
-  "version": "1.3.7",
+  "version": "1.3.10",
   "type": "module",
   "license": "MIT",
   "scripts": {

packages/plugin/src/tui.ts

@@ -296,6 +296,7 @@ export type TuiSlotMap = {
     workspace_id?: string
   }
   home_bottom: {}
+  home_footer: {}
   sidebar_title: {
     session_id: string
     title: string

packages/sdk/js/package.json

@@ -1,7 +1,7 @@
 {
   "$schema": "https://json.schemastore.org/package.json",
   "name": "@opencode-ai/sdk",
-  "version": "1.3.7",
+  "version": "1.3.10",
   "type": "module",
   "license": "MIT",
   "scripts": {

packages/slack/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@opencode-ai/slack",
-  "version": "1.3.7",
+  "version": "1.3.10",
   "type": "module",
   "license": "MIT",
   "scripts": {

packages/ui/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@opencode-ai/ui",
-  "version": "1.3.7",
+  "version": "1.3.10",
   "type": "module",
   "license": "MIT",
   "exports": {

packages/util/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@opencode-ai/util",
-  "version": "1.3.7",
+  "version": "1.3.10",
   "private": true,
   "type": "module",
   "license": "MIT",

packages/web/package.json

@@ -2,7 +2,7 @@
   "name": "@opencode-ai/web",
   "type": "module",
   "license": "MIT",
-  "version": "1.3.7",
+  "version": "1.3.10",
   "scripts": {
     "dev": "astro dev",
     "dev:remote": "VITE_API_URL=https://api.opencode.ai astro dev",

packages/web/src/content/docs/ar/zen.mdx

@@ -20,24 +20,24 @@ OpenCode Zen متاح حاليا بنسخة تجريبية.
 
 ## الخلفية
 
-توجد نماذج كثيرة جدا، لكن عددا قليلا فقط منها يعمل بشكل جيد كعوامل للبرمجة.
-بالإضافة إلى ذلك، فإن معظم المزوّدين يختلفون كثيرا في طريقة الإعداد؛ لذلك قد
-تحصل على أداء وجودة مختلفين بشكل كبير.
+يوجد عدد كبير من النماذج، لكن عددا قليلا فقط منها يعمل جيدا كعوامل للبرمجة.
+إضافة إلى ذلك، يتم إعداد معظم المزوّدين بطرق مختلفة جدا؛ لذلك ستحصل على أداء
+وجودة مختلفين بشكل كبير.
 
 :::tip
 اختبرنا مجموعة منتقاة من النماذج والمزوّدين الذين يعملون جيدا مع OpenCode.
 :::
 
-لذلك إذا كنت تستخدم نموذجا عبر خدمة مثل OpenRouter، فلن تكون واثقا أبدا من أنك
-تحصل على أفضل نسخة من النموذج الذي تريده.
+لذلك إذا كنت تستخدم نموذجا عبر خدمة مثل OpenRouter، فلا يمكنك أبدا التأكد مما
+إذا كنت تحصل على أفضل نسخة من النموذج الذي تريده.
 
-لمعالجة ذلك، قمنا بعدة أمور:
+ولحل ذلك، قمنا بعدة أمور:
 
-1. اختبرنا مجموعة منتقاة من النماذج وتحدثنا مع فرقها حول افضل طريقة لتشغيلها.
-2. ثم عملنا مع عدد من المزوّدين للتأكد من تقديمها بشكل صحيح.
-3. أخيرا، قمنا بقياس أداء توليفة النموذج/المزوّد وخرجنا بقائمة نوصي بها بثقة.
+1. اختبرنا مجموعة منتقاة من النماذج وتحدثنا مع فرقها حول أفضل طريقة لتشغيلها.
+2. ثم عملنا مع عدد من المزوّدين للتأكد من تقديم هذه النماذج بالشكل الصحيح.
+3. وأخيرا، أجرينا قياسا معياريا لتوليفة النموذج/المزوّد ووضعنا قائمة نشعر بثقة في التوصية بها.
 
-OpenCode Zen هو بوابة للذكاء الاصطناعي تتيح لك الوصول إلى هذه النماذج.
+OpenCode Zen هي بوابة AI تتيح لك الوصول إلى هذه النماذج.
 
 ---
 
@@ -47,9 +47,9 @@ OpenCode Zen هو بوابة للذكاء الاصطناعي تتيح لك ال
 
 1. تسجّل الدخول إلى **<a href={console}>OpenCode Zen</a>**، وتضيف تفاصيل الفوترة، ثم تنسخ مفتاح API.
 2. تشغّل الأمر `/connect` في واجهة TUI، وتختار OpenCode Zen، ثم تلصق مفتاح API.
-3. شغّل `/models` في واجهة TUI لعرض قائمة النماذج التي نوصي بها.
+3. تشغّل `/models` في واجهة TUI لرؤية قائمة النماذج التي نوصي بها.
 
-يتم احتساب الرسوم لكل طلب، ويمكنك إضافة رصيد إلى حسابك.
+تتم محاسبتك على كل طلب ويمكنك إضافة رصيد إلى حسابك.
 
 ---
 
@@ -57,44 +57,44 @@ OpenCode Zen هو بوابة للذكاء الاصطناعي تتيح لك ال
 
 يمكنك أيضا الوصول إلى نماذجنا عبر نقاط نهاية API التالية.
 
-| النموذج            | معرّف النموذج      | نقطة النهاية                                       | حزمة AI SDK                 |
-| ------------------ | ------------------ | -------------------------------------------------- | --------------------------- |
-| GPT 5.4            | gpt-5.4            | `https://opencode.ai/zen/v1/responses`             | `@ai-sdk/openai`            |
-| GPT 5.3 Codex      | gpt-5.3-codex      | `https://opencode.ai/zen/v1/responses`             | `@ai-sdk/openai`            |
-| GPT 5.2            | gpt-5.2            | `https://opencode.ai/zen/v1/responses`             | `@ai-sdk/openai`            |
-| GPT 5.2 Codex      | gpt-5.2-codex      | `https://opencode.ai/zen/v1/responses`             | `@ai-sdk/openai`            |
-| GPT 5.1            | gpt-5.1            | `https://opencode.ai/zen/v1/responses`             | `@ai-sdk/openai`            |
-| GPT 5.1 Codex      | gpt-5.1-codex      | `https://opencode.ai/zen/v1/responses`             | `@ai-sdk/openai`            |
-| GPT 5.1 Codex Max  | gpt-5.1-codex-max  | `https://opencode.ai/zen/v1/responses`             | `@ai-sdk/openai`            |
-| GPT 5.1 Codex Mini | gpt-5.1-codex-mini | `https://opencode.ai/zen/v1/responses`             | `@ai-sdk/openai`            |
-| GPT 5              | gpt-5              | `https://opencode.ai/zen/v1/responses`             | `@ai-sdk/openai`            |
-| GPT 5 Codex        | gpt-5-codex        | `https://opencode.ai/zen/v1/responses`             | `@ai-sdk/openai`            |
-| GPT 5 Nano         | gpt-5-nano         | `https://opencode.ai/zen/v1/responses`             | `@ai-sdk/openai`            |
-| Claude Opus 4.6    | claude-opus-4-6    | `https://opencode.ai/zen/v1/messages`              | `@ai-sdk/anthropic`         |
-| Claude Opus 4.5    | claude-opus-4-5    | `https://opencode.ai/zen/v1/messages`              | `@ai-sdk/anthropic`         |
-| Claude Opus 4.1    | claude-opus-4-1    | `https://opencode.ai/zen/v1/messages`              | `@ai-sdk/anthropic`         |
-| Claude Sonnet 4.6  | claude-sonnet-4-6  | `https://opencode.ai/zen/v1/messages`              | `@ai-sdk/anthropic`         |
-| Claude Sonnet 4.5  | claude-sonnet-4-5  | `https://opencode.ai/zen/v1/messages`              | `@ai-sdk/anthropic`         |
-| Claude Sonnet 4    | claude-sonnet-4    | `https://opencode.ai/zen/v1/messages`              | `@ai-sdk/anthropic`         |
-| Claude Haiku 4.5   | claude-haiku-4-5   | `https://opencode.ai/zen/v1/messages`              | `@ai-sdk/anthropic`         |
-| Claude Haiku 3.5   | claude-3-5-haiku   | `https://opencode.ai/zen/v1/messages`              | `@ai-sdk/anthropic`         |
-| Gemini 3.1 Pro     | gemini-3.1-pro     | `https://opencode.ai/zen/v1/models/gemini-3.1-pro` | `@ai-sdk/google`            |
-| Gemini 3 Pro       | gemini-3-pro       | `https://opencode.ai/zen/v1/models/gemini-3-pro`   | `@ai-sdk/google`            |
-| Gemini 3 Flash     | gemini-3-flash     | `https://opencode.ai/zen/v1/models/gemini-3-flash` | `@ai-sdk/google`            |
-| MiniMax M2.5       | minimax-m2.5       | `https://opencode.ai/zen/v1/chat/completions`      | `@ai-sdk/openai-compatible` |
-| MiniMax M2.5 Free  | minimax-m2.5-free  | `https://opencode.ai/zen/v1/chat/completions`      | `@ai-sdk/openai-compatible` |
-| MiniMax M2.1       | minimax-m2.1       | `https://opencode.ai/zen/v1/chat/completions`      | `@ai-sdk/openai-compatible` |
-| GLM 5              | glm-5              | `https://opencode.ai/zen/v1/chat/completions`      | `@ai-sdk/openai-compatible` |
-| GLM 4.7            | glm-4.7            | `https://opencode.ai/zen/v1/chat/completions`      | `@ai-sdk/openai-compatible` |
-| GLM 4.6            | glm-4.6            | `https://opencode.ai/zen/v1/chat/completions`      | `@ai-sdk/openai-compatible` |
-| Kimi K2.5          | kimi-k2.5          | `https://opencode.ai/zen/v1/chat/completions`      | `@ai-sdk/openai-compatible` |
-| Kimi K2 Thinking   | kimi-k2-thinking   | `https://opencode.ai/zen/v1/chat/completions`      | `@ai-sdk/openai-compatible` |
-| Kimi K2            | kimi-k2            | `https://opencode.ai/zen/v1/chat/completions`      | `@ai-sdk/openai-compatible` |
-| Qwen3 Coder 480B   | qwen3-coder        | `https://opencode.ai/zen/v1/chat/completions`      | `@ai-sdk/openai-compatible` |
-| Big Pickle         | big-pickle         | `https://opencode.ai/zen/v1/chat/completions`      | `@ai-sdk/openai-compatible` |
+| النموذج               | معرّف النموذج         | نقطة النهاية                                       | حزمة AI SDK                 |
+| --------------------- | --------------------- | -------------------------------------------------- | --------------------------- |
+| GPT 5.4               | gpt-5.4               | `https://opencode.ai/zen/v1/responses`             | `@ai-sdk/openai`            |
+| GPT 5.4 Pro           | gpt-5.4-pro           | `https://opencode.ai/zen/v1/responses`             | `@ai-sdk/openai`            |
+| GPT 5.4 Mini          | gpt-5.4-mini          | `https://opencode.ai/zen/v1/responses`             | `@ai-sdk/openai`            |
+| GPT 5.4 Nano          | gpt-5.4-nano          | `https://opencode.ai/zen/v1/responses`             | `@ai-sdk/openai`            |
+| GPT 5.3 Codex         | gpt-5.3-codex         | `https://opencode.ai/zen/v1/responses`             | `@ai-sdk/openai`            |
+| GPT 5.3 Codex Spark   | gpt-5.3-codex-spark   | `https://opencode.ai/zen/v1/responses`             | `@ai-sdk/openai`            |
+| GPT 5.2               | gpt-5.2               | `https://opencode.ai/zen/v1/responses`             | `@ai-sdk/openai`            |
+| GPT 5.2 Codex         | gpt-5.2-codex         | `https://opencode.ai/zen/v1/responses`             | `@ai-sdk/openai`            |
+| GPT 5.1               | gpt-5.1               | `https://opencode.ai/zen/v1/responses`             | `@ai-sdk/openai`            |
+| GPT 5.1 Codex         | gpt-5.1-codex         | `https://opencode.ai/zen/v1/responses`             | `@ai-sdk/openai`            |
+| GPT 5.1 Codex Max     | gpt-5.1-codex-max     | `https://opencode.ai/zen/v1/responses`             | `@ai-sdk/openai`            |
+| GPT 5.1 Codex Mini    | gpt-5.1-codex-mini    | `https://opencode.ai/zen/v1/responses`             | `@ai-sdk/openai`            |
+| GPT 5                 | gpt-5                 | `https://opencode.ai/zen/v1/responses`             | `@ai-sdk/openai`            |
+| GPT 5 Codex           | gpt-5-codex           | `https://opencode.ai/zen/v1/responses`             | `@ai-sdk/openai`            |
+| GPT 5 Nano            | gpt-5-nano            | `https://opencode.ai/zen/v1/responses`             | `@ai-sdk/openai`            |
+| Claude Opus 4.6       | claude-opus-4-6       | `https://opencode.ai/zen/v1/messages`              | `@ai-sdk/anthropic`         |
+| Claude Opus 4.5       | claude-opus-4-5       | `https://opencode.ai/zen/v1/messages`              | `@ai-sdk/anthropic`         |
+| Claude Opus 4.1       | claude-opus-4-1       | `https://opencode.ai/zen/v1/messages`              | `@ai-sdk/anthropic`         |
+| Claude Sonnet 4.6     | claude-sonnet-4-6     | `https://opencode.ai/zen/v1/messages`              | `@ai-sdk/anthropic`         |
+| Claude Sonnet 4.5     | claude-sonnet-4-5     | `https://opencode.ai/zen/v1/messages`              | `@ai-sdk/anthropic`         |
+| Claude Sonnet 4       | claude-sonnet-4       | `https://opencode.ai/zen/v1/messages`              | `@ai-sdk/anthropic`         |
+| Claude Haiku 4.5      | claude-haiku-4-5      | `https://opencode.ai/zen/v1/messages`              | `@ai-sdk/anthropic`         |
+| Claude Haiku 3.5      | claude-3-5-haiku      | `https://opencode.ai/zen/v1/messages`              | `@ai-sdk/anthropic`         |
+| Gemini 3.1 Pro        | gemini-3.1-pro        | `https://opencode.ai/zen/v1/models/gemini-3.1-pro` | `@ai-sdk/google`            |
+| Gemini 3 Flash        | gemini-3-flash        | `https://opencode.ai/zen/v1/models/gemini-3-flash` | `@ai-sdk/google`            |
+| MiniMax M2.5          | minimax-m2.5          | `https://opencode.ai/zen/v1/chat/completions`      | `@ai-sdk/openai-compatible` |
+| MiniMax M2.5 Free     | minimax-m2.5-free     | `https://opencode.ai/zen/v1/chat/completions`      | `@ai-sdk/openai-compatible` |
+| GLM 5                 | glm-5                 | `https://opencode.ai/zen/v1/chat/completions`      | `@ai-sdk/openai-compatible` |
+| Kimi K2.5             | kimi-k2.5             | `https://opencode.ai/zen/v1/chat/completions`      | `@ai-sdk/openai-compatible` |
+| Big Pickle            | big-pickle            | `https://opencode.ai/zen/v1/chat/completions`      | `@ai-sdk/openai-compatible` |
+| MiMo V2 Pro Free      | mimo-v2-pro-free      | `https://opencode.ai/zen/v1/chat/completions`      | `@ai-sdk/openai-compatible` |
+| MiMo V2 Omni Free     | mimo-v2-omni-free     | `https://opencode.ai/zen/v1/chat/completions`      | `@ai-sdk/openai-compatible` |
+| Qwen3.6 Plus Free     | qwen3.6-plus-free     | `https://opencode.ai/zen/v1/chat/completions`      | `@ai-sdk/openai-compatible` |
+| Nemotron 3 Super Free | nemotron-3-super-free | `https://opencode.ai/zen/v1/chat/completions`      | `@ai-sdk/openai-compatible` |
 
-يستخدم [معرّف النموذج](/docs/config/#models) في إعدادات OpenCode الصيغة `opencode/<model-id>`.
-على سبيل المثال، بالنسبة إلى GPT 5.2 Codex ستستخدم `opencode/gpt-5.2-codex` في إعداداتك.
+يستخدم [معرّف النموذج](/docs/config/#models) في إعدادات OpenCode الصيغة `opencode/<model-id>`. على سبيل المثال، بالنسبة إلى GPT 5.3 Codex، ستستخدم `opencode/gpt-5.3-codex` في إعداداتك.
 
 ---
 
@@ -112,57 +112,60 @@ https://opencode.ai/zen/v1/models
 
 ندعم نموذج الدفع حسب الاستخدام. فيما يلي الأسعار **لكل 1M tokens**.
 
-| النموذج                           | الإدخال | الإخراج | قراءة مخزنة | كتابة مخزنة |
-| --------------------------------- | ------- | ------- | ----------- | ----------- |
-| Big Pickle                        | Free    | Free    | Free        | -           |
-| MiniMax M2.5 Free                 | Free    | Free    | Free        | -           |
-| MiniMax M2.5                      | $0.30   | $1.20   | $0.06       | -           |
-| MiniMax M2.1                      | $0.30   | $1.20   | $0.10       | -           |
-| GLM 5                             | $1.00   | $3.20   | $0.20       | -           |
-| GLM 4.7                           | $0.60   | $2.20   | $0.10       | -           |
-| GLM 4.6                           | $0.60   | $2.20   | $0.10       | -           |
-| Kimi K2.5                         | $0.60   | $3.00   | $0.08       | -           |
-| Kimi K2 Thinking                  | $0.40   | $2.50   | -           | -           |
-| Kimi K2                           | $0.40   | $2.50   | -           | -           |
-| Qwen3 Coder 480B                  | $0.45   | $1.50   | -           | -           |
-| Claude Opus 4.6 (≤ 200K tokens)   | $5.00   | $25.00  | $0.50       | $6.25       |
-| Claude Opus 4.6 (> 200K tokens)   | $10.00  | $37.50  | $1.00       | $12.50      |
-| Claude Opus 4.5                   | $5.00   | $25.00  | $0.50       | $6.25       |
-| Claude Opus 4.1                   | $15.00  | $75.00  | $1.50       | $18.75      |
-| Claude Sonnet 4.6 (≤ 200K tokens) | $3.00   | $15.00  | $0.30       | $3.75       |
-| Claude Sonnet 4.6 (> 200K tokens) | $6.00   | $22.50  | $0.60       | $7.50       |
-| Claude Sonnet 4.5 (≤ 200K tokens) | $3.00   | $15.00  | $0.30       | $3.75       |
-| Claude Sonnet 4.5 (> 200K tokens) | $6.00   | $22.50  | $0.60       | $7.50       |
-| Claude Sonnet 4 (≤ 200K tokens)   | $3.00   | $15.00  | $0.30       | $3.75       |
-| Claude Sonnet 4 (> 200K tokens)   | $6.00   | $22.50  | $0.60       | $7.50       |
-| Claude Haiku 4.5                  | $1.00   | $5.00   | $0.10       | $1.25       |
-| Claude Haiku 3.5                  | $0.80   | $4.00   | $0.08       | $1.00       |
-| Gemini 3.1 Pro (≤ 200K tokens)    | $2.00   | $12.00  | $0.20       | -           |
-| Gemini 3.1 Pro (> 200K tokens)    | $4.00   | $18.00  | $0.40       | -           |
-| Gemini 3 Pro (≤ 200K tokens)      | $2.00   | $12.00  | $0.20       | -           |
-| Gemini 3 Pro (> 200K tokens)      | $4.00   | $18.00  | $0.40       | -           |
-| Gemini 3 Flash                    | $0.50   | $3.00   | $0.05       | -           |
-| GPT 5.4                           | $2.50   | $15.00  | $0.25       | -           |
-| GPT 5.3 Codex                     | $1.75   | $14.00  | $0.175      | -           |
-| GPT 5.2                           | $1.75   | $14.00  | $0.175      | -           |
-| GPT 5.2 Codex                     | $1.75   | $14.00  | $0.175      | -           |
-| GPT 5.1                           | $1.07   | $8.50   | $0.107      | -           |
-| GPT 5.1 Codex                     | $1.07   | $8.50   | $0.107      | -           |
-| GPT 5.1 Codex Max                 | $1.25   | $10.00  | $0.125      | -           |
-| GPT 5.1 Codex Mini                | $0.25   | $2.00   | $0.025      | -           |
-| GPT 5                             | $1.07   | $8.50   | $0.107      | -           |
-| GPT 5 Codex                       | $1.07   | $8.50   | $0.107      | -           |
-| GPT 5 Nano                        | Free    | Free    | Free        | -           |
+| النموذج                           | الإدخال | الإخراج | القراءة المخزنة | الكتابة المخزنة |
+| --------------------------------- | ------- | ------- | --------------- | --------------- |
+| Big Pickle                        | Free    | Free    | Free            | -               |
+| MiMo V2 Pro Free                  | Free    | Free    | Free            | -               |
+| MiMo V2 Omni Free                 | Free    | Free    | Free            | -               |
+| Qwen3.6 Plus Free                 | Free    | Free    | Free            | -               |
+| Nemotron 3 Super Free             | Free    | Free    | Free            | -               |
+| MiniMax M2.5 Free                 | Free    | Free    | Free            | -               |
+| MiniMax M2.5                      | $0.30   | $1.20   | $0.06           | $0.375          |
+| GLM 5                             | $1.00   | $3.20   | $0.20           | -               |
+| Kimi K2.5                         | $0.60   | $3.00   | $0.10           | -               |
+| Qwen3 Coder 480B                  | $0.45   | $1.50   | -               | -               |
+| Claude Opus 4.6                   | $5.00   | $25.00  | $0.50           | $6.25           |
+| Claude Opus 4.5                   | $5.00   | $25.00  | $0.50           | $6.25           |
+| Claude Opus 4.1                   | $15.00  | $75.00  | $1.50           | $18.75          |
+| Claude Sonnet 4.6                 | $3.00   | $15.00  | $0.30           | $3.75           |
+| Claude Sonnet 4.5 (≤ 200K tokens) | $3.00   | $15.00  | $0.30           | $3.75           |
+| Claude Sonnet 4.5 (> 200K tokens) | $6.00   | $22.50  | $0.60           | $7.50           |
+| Claude Sonnet 4 (≤ 200K tokens)   | $3.00   | $15.00  | $0.30           | $3.75           |
+| Claude Sonnet 4 (> 200K tokens)   | $6.00   | $22.50  | $0.60           | $7.50           |
+| Claude Haiku 4.5                  | $1.00   | $5.00   | $0.10           | $1.25           |
+| Claude Haiku 3.5                  | $0.80   | $4.00   | $0.08           | $1.00           |
+| Gemini 3.1 Pro (≤ 200K tokens)    | $2.00   | $12.00  | $0.20           | -               |
+| Gemini 3.1 Pro (> 200K tokens)    | $4.00   | $18.00  | $0.40           | -               |
+| Gemini 3 Flash                    | $0.50   | $3.00   | $0.05           | -               |
+| GPT 5.4                           | $2.50   | $15.00  | $0.25           | -               |
+| GPT 5.4 Pro                       | $30.00  | $180.00 | $30.00          | -               |
+| GPT 5.4 Mini                      | $0.75   | $4.50   | $0.075          | -               |
+| GPT 5.4 Nano                      | $0.20   | $1.25   | $0.02           | -               |
+| GPT 5.3 Codex Spark               | $1.75   | $14.00  | $0.175          | -               |
+| GPT 5.3 Codex                     | $1.75   | $14.00  | $0.175          | -               |
+| GPT 5.2                           | $1.75   | $14.00  | $0.175          | -               |
+| GPT 5.2 Codex                     | $1.75   | $14.00  | $0.175          | -               |
+| GPT 5.1                           | $1.07   | $8.50   | $0.107          | -               |
+| GPT 5.1 Codex                     | $1.07   | $8.50   | $0.107          | -               |
+| GPT 5.1 Codex Max                 | $1.25   | $10.00  | $0.125          | -               |
+| GPT 5.1 Codex Mini                | $0.25   | $2.00   | $0.025          | -               |
+| GPT 5                             | $1.07   | $8.50   | $0.107          | -               |
+| GPT 5 Codex                       | $1.07   | $8.50   | $0.107          | -               |
+| GPT 5 Nano                        | Free    | Free    | Free            | -               |
 
 قد تلاحظ _Claude Haiku 3.5_ في سجل الاستخدام. هذا [نموذج منخفض التكلفة](/docs/config/#models) يُستخدم لتوليد عناوين جلساتك.
 
 :::note
-يتم تمرير رسوم بطاقات الائتمان بالتكلفة الفعلية (4.4% + $0.30 لكل معاملة)؛ ولا نفرض أي رسوم إضافية غير ذلك.
+يتم تمرير رسوم بطاقات الائتمان بالتكلفة الفعلية (4.4% + $0.30 لكل معاملة)؛ ولا نفرض أي رسوم إضافية فوق ذلك.
 :::
 
 النماذج المجانية:
 
 - MiniMax M2.5 Free متاح على OpenCode لفترة محدودة. يستخدم الفريق هذه الفترة لجمع الملاحظات وتحسين النموذج.
+- MiMo V2 Pro Free متاح على OpenCode لفترة محدودة. يستخدم الفريق هذه الفترة لجمع الملاحظات وتحسين النموذج.
+- MiMo V2 Omni Free متاح على OpenCode لفترة محدودة. يستخدم الفريق هذه الفترة لجمع الملاحظات وتحسين النموذج.
+- Qwen3.6 Plus Free متاح على OpenCode لفترة محدودة. يستخدم الفريق هذه الفترة لجمع الملاحظات وتحسين النموذج.
+- Nemotron 3 Super Free متاح على OpenCode لفترة محدودة. يستخدم الفريق هذه الفترة لجمع الملاحظات وتحسين النموذج.
 - Big Pickle نموذج خفي ومتاح مجانا على OpenCode لفترة محدودة. يستخدم الفريق هذه الفترة لجمع الملاحظات وتحسين النموذج.
 
 <a href={email}>تواصل معنا</a> إذا كانت لديك أي أسئلة.
@@ -181,32 +184,36 @@ https://opencode.ai/zen/v1/models
 
 يمكنك أيضا تعيين حد شهري للاستخدام لمساحة العمل بالكامل ولكل عضو في فريقك.
 
-على سبيل المثال، لنفترض أنك ضبطت حد الاستخدام الشهري على $20، فلن يتجاوز Zen مبلغ $20 خلال شهر.
-لكن إذا كانت إعادة الشحن التلقائي مفعّلة، فقد ينتهي الأمر بخصم أكثر من $20 إذا انخفض رصيدك عن $5.
+على سبيل المثال، لنفترض أنك ضبطت حد الاستخدام الشهري على $20، فلن يستخدم Zen أكثر من $20 خلال شهر. لكن إذا كانت إعادة الشحن التلقائي مفعّلة، فقد ينتهي الأمر بخصم أكثر من $20 إذا انخفض رصيدك عن $5.
 
 ---
 
-### نماذج مهملة
+### النماذج المهملة
 
 | النموذج          | تاريخ الإيقاف |
 | ---------------- | ------------- |
-| Qwen3 Coder 480B | 6 فبراير 2026 |
-| Kimi K2 Thinking | 6 مارس 2026   |
-| Kimi K2          | 6 مارس 2026   |
 | MiniMax M2.1     | 15 مارس 2026  |
 | GLM 4.7          | 15 مارس 2026  |
 | GLM 4.6          | 15 مارس 2026  |
+| Gemini 3 Pro     | 9 مارس 2026   |
+| Kimi K2 Thinking | 6 مارس 2026   |
+| Kimi K2          | 6 مارس 2026   |
+| Qwen3 Coder 480B | 6 فبراير 2026 |
 
 ---
 
 ## الخصوصية
 
-تتم استضافة جميع نماذجنا في الولايات المتحدة. يلتزم مزوّدونا بسياسة عدم الاحتفاظ بالبيانات (zero-retention) ولا يستخدمون بياناتك لتدريب النماذج، مع الاستثناءات التالية:
+تتم استضافة جميع نماذجنا في الولايات المتحدة. يلتزم مزوّدونا بسياسة عدم الاحتفاظ بالبيانات ولا يستخدمون بياناتك لتدريب النماذج، مع الاستثناءات التالية:
 
-- Big Pickle: خلال فترة إتاحته المجانية، قد تُستخدم البيانات المجمعة لتحسين النموذج.
-- MiniMax M2.5 Free: خلال فترة إتاحته المجانية، قد تُستخدم البيانات المجمعة لتحسين النموذج.
-- OpenAI APIs: يتم الاحتفاظ بالطلبات لمدة 30 يوما وفقا لـ [سياسات بيانات OpenAI](https://platform.openai.com/docs/guides/your-data).
-- Anthropic APIs: يتم الاحتفاظ بالطلبات لمدة 30 يوما وفقا لـ [سياسات بيانات Anthropic](https://docs.anthropic.com/en/docs/claude-code/data-usage).
+- Big Pickle: خلال فترته المجانية، قد تُستخدم البيانات المجمعة لتحسين النموذج.
+- MiniMax M2.5 Free: خلال فترته المجانية، قد تُستخدم البيانات المجمعة لتحسين النموذج.
+- MiMo V2 Pro Free: خلال فترته المجانية، قد تُستخدم البيانات المجمعة لتحسين النموذج.
+- MiMo V2 Omni Free: خلال فترته المجانية، قد تُستخدم البيانات المجمعة لتحسين النموذج.
+- Qwen3.6 Plus Free: خلال فترته المجانية، قد تُستخدم البيانات المجمعة لتحسين النموذج.
+- Nemotron 3 Super Free: خلال فترته المجانية، قد تُستخدم البيانات المجمعة لتحسين النموذج.
+- OpenAI APIs: يتم الاحتفاظ بالطلبات لمدة 30 يوما وفقا لـ [OpenAI's Data Policies](https://platform.openai.com/docs/guides/your-data).
+- Anthropic APIs: يتم الاحتفاظ بالطلبات لمدة 30 يوما وفقا لـ [Anthropic's Data Policies](https://docs.anthropic.com/en/docs/claude-code/data-usage).
 
 ---
 
@@ -230,21 +237,21 @@ https://opencode.ai/zen/v1/models
 - **Admin**: إدارة النماذج والأعضاء ومفاتيح API والفوترة
 - **Member**: إدارة مفاتيح API الخاصة به فقط
 
-يمكن للمسؤولين أيضا تعيين حدود إنفاق شهرية لكل عضو للسيطرة على التكاليف.
+يمكن للمسؤولين أيضا تعيين حدود إنفاق شهرية لكل عضو لإبقاء التكاليف تحت السيطرة.
 
 ---
 
-### صلاحيات الوصول للنماذج
+### الوصول إلى النماذج
 
 يمكن للمسؤولين تفعيل نماذج محددة لمساحة العمل أو تعطيلها. ستعيد الطلبات المرسلة إلى نموذج معطّل خطأ.
 
-يفيد ذلك في الحالات التي تريد فيها تعطيل استخدام نموذج يقوم بجمع البيانات.
+يفيد ذلك في الحالات التي تريد فيها تعطيل استخدام نموذج يجمع البيانات.
 
 ---
 
 ### استخدم مفتاحك الخاص
 
-يمكنك استخدام مفاتيح API الخاصة بك لدى OpenAI أو Anthropic مع الاستمرار في الوصول إلى نماذج أخرى ضمن Zen.
+يمكنك استخدام مفاتيح API الخاصة بك لدى OpenAI أو Anthropic مع الاستمرار في الوصول إلى نماذج أخرى في Zen.
 
 عند استخدام مفاتيحك الخاصة، تتم فوترة tokens مباشرة من المزوّد وليس من Zen.
 
@@ -257,6 +264,6 @@ https://opencode.ai/zen/v1/models
 أنشأنا OpenCode Zen من أجل:
 
 1. **قياس الأداء** لأفضل النماذج/المزوّدين لعوامل البرمجة.
-2. إتاحة خيارات **عالية الجودة** دون خفض الأداء أو توجيه الطلبات إلى مزوّدين أرخص.
-3. تمرير أي **انخفاض في الأسعار** عبر البيع بالتكلفة؛ بحيث تكون الزيادة الوحيدة لتغطية رسوم المعالجة.
-4. عدم فرض **أي ارتباط حصري (lock-in)** عبر تمكينك من استخدامه مع أي عامل برمجة آخر، مع إتاحة استخدام أي مزوّد آخر مع OpenCode أيضا.
+2. إتاحة الخيارات **الأعلى جودة** دون خفض الأداء أو توجيه الطلبات إلى مزوّدين أرخص.
+3. تمرير أي **انخفاضات في الأسعار** عبر البيع بالتكلفة؛ بحيث تكون الزيادة الوحيدة لتغطية رسوم المعالجة.
+4. **عدم فرض أي ارتباط حصري** عبر تمكينك من استخدامه مع أي عامل برمجة آخر. ومع إبقاء إمكانية استخدام أي مزوّد آخر مع OpenCode أيضا دائما.

packages/web/src/content/docs/bs/zen.mdx

@@ -7,31 +7,39 @@ import config from "../../../../config.mjs"
 export const console = config.console
 export const email = `mailto:${config.email}`
 
-OpenCode Zen je lista testiranih i provjerenih modela koje obezbjeduje OpenCode tim.
+OpenCode Zen je lista testiranih i provjerenih modela koje pruža OpenCode tim.
 
 :::note
 OpenCode Zen je trenutno u beta fazi.
 :::
 
-Zen radi kao i svaki drugi provajder u OpenCode. Prijavite se u OpenCode Zen i uzmete API kljuc. Ovo je **potpuno opcionalno** i ne morate ga koristiti da biste koristili OpenCode.
+Zen radi kao i svaki drugi provajder u OpenCode. Prijavite se na OpenCode Zen i
+preuzmete svoj API ključ. To je **potpuno opcionalno** i ne morate ga koristiti
+da biste koristili OpenCode.
 
 ---
 
 ## Pozadina
 
-Postoji veliki broj modela, ali samo mali dio radi dobro kao coding agent. Dodatno, vecina provajdera je drugacije konfigurisana, pa su performanse i kvalitet cesto neujednaceni.
+Postoji veliki broj modela, ali samo mali broj tih modela dobro funkcioniše kao
+coding agent. Osim toga, većina provajdera je konfigurisana veoma različito, pa
+zbog toga dobijate veoma različite performanse i kvalitet.
 
 :::tip
 Testirali smo odabranu grupu modela i provajdera koji dobro rade s OpenCode.
 :::
 
-Ako model koristite preko servisa poput OpenRouter-a, cesto ne mozete biti sigurni da dobijate najbolju verziju zeljenog modela.
+Zato, ako model koristite preko nečega poput OpenRouter, nikada ne možete biti
+sigurni da dobijate najbolju verziju modela koji želite.
 
-Da to rijesimo, uradili smo nekoliko stvari:
+Da bismo to riješili, uradili smo nekoliko stvari:
 
-1. Testirali smo odabrane modele i razgovarali sa njihovim timovima kako ih najbolje pokretati.
-2. Zatim smo saradjivali s nekoliko provajdera da potvrdimo da se modeli isporucuju ispravno.
-3. Na kraju smo benchmarkirali kombinacije model/provajder i sastavili listu koju preporucujemo.
+1. Testirali smo odabranu grupu modela i razgovarali s njihovim timovima o tome
+   kako ih najbolje pokretati.
+2. Zatim smo radili s nekoliko provajdera kako bismo bili sigurni da se ti
+   modeli isporučuju ispravno.
+3. Na kraju smo benchmarkirali kombinacije model/provajder i sastavili listu
+   koju s punim povjerenjem preporučujemo.
 
 OpenCode Zen je AI gateway koji vam daje pristup tim modelima.
 
@@ -41,61 +49,65 @@ OpenCode Zen je AI gateway koji vam daje pristup tim modelima.
 
 OpenCode Zen radi kao i svaki drugi provajder u OpenCode.
 
-1. Prijavite se na **<a href={console}>OpenCode Zen</a>**, dodajte billing podatke i kopirajte API kljuc.
-2. U TUI-ju pokrenite `/connect`, izaberite OpenCode Zen i zalijepite API kljuc.
-3. Pokrenite `/models` u TUI-ju da vidite listu preporucenih modela.
+1. Prijavite se na **<a href={console}>OpenCode Zen</a>**, dodajte podatke za
+   naplatu i kopirajte svoj API ključ.
+2. Pokrenite komandu `/connect` u TUI, izaberite OpenCode Zen i zalijepite svoj API ključ.
+3. Pokrenite `/models` u TUI da vidite listu modela koje preporučujemo.
 
-Naplata je po zahtjevu i mozete dodavati kredit na racun.
+Naplata se vrši po zahtjevu i možete dodavati kredit na svoj račun.
 
 ---
 
 ## Endpoints
 
-Nasim modelima mozete pristupiti i preko sljedecih API endpointa.
+Našim modelima možete pristupiti i preko sljedećih API endpointa.
 
-| Model              | Model ID           | Endpoint                                           | AI SDK Package              |
-| ------------------ | ------------------ | -------------------------------------------------- | --------------------------- |
-| GPT 5.4            | gpt-5.4            | `https://opencode.ai/zen/v1/responses`             | `@ai-sdk/openai`            |
-| GPT 5.3 Codex      | gpt-5.3-codex      | `https://opencode.ai/zen/v1/responses`             | `@ai-sdk/openai`            |
-| GPT 5.2            | gpt-5.2            | `https://opencode.ai/zen/v1/responses`             | `@ai-sdk/openai`            |
-| GPT 5.2 Codex      | gpt-5.2-codex      | `https://opencode.ai/zen/v1/responses`             | `@ai-sdk/openai`            |
-| GPT 5.1            | gpt-5.1            | `https://opencode.ai/zen/v1/responses`             | `@ai-sdk/openai`            |
-| GPT 5.1 Codex      | gpt-5.1-codex      | `https://opencode.ai/zen/v1/responses`             | `@ai-sdk/openai`            |
-| GPT 5.1 Codex Max  | gpt-5.1-codex-max  | `https://opencode.ai/zen/v1/responses`             | `@ai-sdk/openai`            |
-| GPT 5.1 Codex Mini | gpt-5.1-codex-mini | `https://opencode.ai/zen/v1/responses`             | `@ai-sdk/openai`            |
-| GPT 5              | gpt-5              | `https://opencode.ai/zen/v1/responses`             | `@ai-sdk/openai`            |
-| GPT 5 Codex        | gpt-5-codex        | `https://opencode.ai/zen/v1/responses`             | `@ai-sdk/openai`            |
-| GPT 5 Nano         | gpt-5-nano         | `https://opencode.ai/zen/v1/responses`             | `@ai-sdk/openai`            |
-| Claude Opus 4.6    | claude-opus-4-6    | `https://opencode.ai/zen/v1/messages`              | `@ai-sdk/anthropic`         |
-| Claude Opus 4.5    | claude-opus-4-5    | `https://opencode.ai/zen/v1/messages`              | `@ai-sdk/anthropic`         |
-| Claude Opus 4.1    | claude-opus-4-1    | `https://opencode.ai/zen/v1/messages`              | `@ai-sdk/anthropic`         |
-| Claude Sonnet 4.6  | claude-sonnet-4-6  | `https://opencode.ai/zen/v1/messages`              | `@ai-sdk/anthropic`         |
-| Claude Sonnet 4.5  | claude-sonnet-4-5  | `https://opencode.ai/zen/v1/messages`              | `@ai-sdk/anthropic`         |
-| Claude Sonnet 4    | claude-sonnet-4    | `https://opencode.ai/zen/v1/messages`              | `@ai-sdk/anthropic`         |
-| Claude Haiku 4.5   | claude-haiku-4-5   | `https://opencode.ai/zen/v1/messages`              | `@ai-sdk/anthropic`         |
-| Claude Haiku 3.5   | claude-3-5-haiku   | `https://opencode.ai/zen/v1/messages`              | `@ai-sdk/anthropic`         |
-| Gemini 3.1 Pro     | gemini-3.1-pro     | `https://opencode.ai/zen/v1/models/gemini-3.1-pro` | `@ai-sdk/google`            |
-| Gemini 3 Pro       | gemini-3-pro       | `https://opencode.ai/zen/v1/models/gemini-3-pro`   | `@ai-sdk/google`            |
-| Gemini 3 Flash     | gemini-3-flash     | `https://opencode.ai/zen/v1/models/gemini-3-flash` | `@ai-sdk/google`            |
-| MiniMax M2.5       | minimax-m2.5       | `https://opencode.ai/zen/v1/chat/completions`      | `@ai-sdk/openai-compatible` |
-| MiniMax M2.5 Free  | minimax-m2.5-free  | `https://opencode.ai/zen/v1/chat/completions`      | `@ai-sdk/openai-compatible` |
-| MiniMax M2.1       | minimax-m2.1       | `https://opencode.ai/zen/v1/chat/completions`      | `@ai-sdk/openai-compatible` |
-| GLM 5              | glm-5              | `https://opencode.ai/zen/v1/chat/completions`      | `@ai-sdk/openai-compatible` |
-| GLM 4.7            | glm-4.7            | `https://opencode.ai/zen/v1/chat/completions`      | `@ai-sdk/openai-compatible` |
-| GLM 4.6            | glm-4.6            | `https://opencode.ai/zen/v1/chat/completions`      | `@ai-sdk/openai-compatible` |
-| Kimi K2.5          | kimi-k2.5          | `https://opencode.ai/zen/v1/chat/completions`      | `@ai-sdk/openai-compatible` |
-| Kimi K2 Thinking   | kimi-k2-thinking   | `https://opencode.ai/zen/v1/chat/completions`      | `@ai-sdk/openai-compatible` |
-| Kimi K2            | kimi-k2            | `https://opencode.ai/zen/v1/chat/completions`      | `@ai-sdk/openai-compatible` |
-| Qwen3 Coder 480B   | qwen3-coder        | `https://opencode.ai/zen/v1/chat/completions`      | `@ai-sdk/openai-compatible` |
-| Big Pickle         | big-pickle         | `https://opencode.ai/zen/v1/chat/completions`      | `@ai-sdk/openai-compatible` |
+| Model                 | Model ID              | Endpoint                                           | AI SDK Package              |
+| --------------------- | --------------------- | -------------------------------------------------- | --------------------------- |
+| GPT 5.4               | gpt-5.4               | `https://opencode.ai/zen/v1/responses`             | `@ai-sdk/openai`            |
+| GPT 5.4 Pro           | gpt-5.4-pro           | `https://opencode.ai/zen/v1/responses`             | `@ai-sdk/openai`            |
+| GPT 5.4 Mini          | gpt-5.4-mini          | `https://opencode.ai/zen/v1/responses`             | `@ai-sdk/openai`            |
+| GPT 5.4 Nano          | gpt-5.4-nano          | `https://opencode.ai/zen/v1/responses`             | `@ai-sdk/openai`            |
+| GPT 5.3 Codex         | gpt-5.3-codex         | `https://opencode.ai/zen/v1/responses`             | `@ai-sdk/openai`            |
+| GPT 5.3 Codex Spark   | gpt-5.3-codex-spark   | `https://opencode.ai/zen/v1/responses`             | `@ai-sdk/openai`            |
+| GPT 5.2               | gpt-5.2               | `https://opencode.ai/zen/v1/responses`             | `@ai-sdk/openai`            |
+| GPT 5.2 Codex         | gpt-5.2-codex         | `https://opencode.ai/zen/v1/responses`             | `@ai-sdk/openai`            |
+| GPT 5.1               | gpt-5.1               | `https://opencode.ai/zen/v1/responses`             | `@ai-sdk/openai`            |
+| GPT 5.1 Codex         | gpt-5.1-codex         | `https://opencode.ai/zen/v1/responses`             | `@ai-sdk/openai`            |
+| GPT 5.1 Codex Max     | gpt-5.1-codex-max     | `https://opencode.ai/zen/v1/responses`             | `@ai-sdk/openai`            |
+| GPT 5.1 Codex Mini    | gpt-5.1-codex-mini    | `https://opencode.ai/zen/v1/responses`             | `@ai-sdk/openai`            |
+| GPT 5                 | gpt-5                 | `https://opencode.ai/zen/v1/responses`             | `@ai-sdk/openai`            |
+| GPT 5 Codex           | gpt-5-codex           | `https://opencode.ai/zen/v1/responses`             | `@ai-sdk/openai`            |
+| GPT 5 Nano            | gpt-5-nano            | `https://opencode.ai/zen/v1/responses`             | `@ai-sdk/openai`            |
+| Claude Opus 4.6       | claude-opus-4-6       | `https://opencode.ai/zen/v1/messages`              | `@ai-sdk/anthropic`         |
+| Claude Opus 4.5       | claude-opus-4-5       | `https://opencode.ai/zen/v1/messages`              | `@ai-sdk/anthropic`         |
+| Claude Opus 4.1       | claude-opus-4-1       | `https://opencode.ai/zen/v1/messages`              | `@ai-sdk/anthropic`         |
+| Claude Sonnet 4.6     | claude-sonnet-4-6     | `https://opencode.ai/zen/v1/messages`              | `@ai-sdk/anthropic`         |
+| Claude Sonnet 4.5     | claude-sonnet-4-5     | `https://opencode.ai/zen/v1/messages`              | `@ai-sdk/anthropic`         |
+| Claude Sonnet 4       | claude-sonnet-4       | `https://opencode.ai/zen/v1/messages`              | `@ai-sdk/anthropic`         |
+| Claude Haiku 4.5      | claude-haiku-4-5      | `https://opencode.ai/zen/v1/messages`              | `@ai-sdk/anthropic`         |
+| Claude Haiku 3.5      | claude-3-5-haiku      | `https://opencode.ai/zen/v1/messages`              | `@ai-sdk/anthropic`         |
+| Gemini 3.1 Pro        | gemini-3.1-pro        | `https://opencode.ai/zen/v1/models/gemini-3.1-pro` | `@ai-sdk/google`            |
+| Gemini 3 Flash        | gemini-3-flash        | `https://opencode.ai/zen/v1/models/gemini-3-flash` | `@ai-sdk/google`            |
+| MiniMax M2.5          | minimax-m2.5          | `https://opencode.ai/zen/v1/chat/completions`      | `@ai-sdk/openai-compatible` |
+| MiniMax M2.5 Free     | minimax-m2.5-free     | `https://opencode.ai/zen/v1/chat/completions`      | `@ai-sdk/openai-compatible` |
+| GLM 5                 | glm-5                 | `https://opencode.ai/zen/v1/chat/completions`      | `@ai-sdk/openai-compatible` |
+| Kimi K2.5             | kimi-k2.5             | `https://opencode.ai/zen/v1/chat/completions`      | `@ai-sdk/openai-compatible` |
+| Big Pickle            | big-pickle            | `https://opencode.ai/zen/v1/chat/completions`      | `@ai-sdk/openai-compatible` |
+| MiMo V2 Pro Free      | mimo-v2-pro-free      | `https://opencode.ai/zen/v1/chat/completions`      | `@ai-sdk/openai-compatible` |
+| MiMo V2 Omni Free     | mimo-v2-omni-free     | `https://opencode.ai/zen/v1/chat/completions`      | `@ai-sdk/openai-compatible` |
+| Qwen3.6 Plus Free     | qwen3.6-plus-free     | `https://opencode.ai/zen/v1/chat/completions`      | `@ai-sdk/openai-compatible` |
+| Nemotron 3 Super Free | nemotron-3-super-free | `https://opencode.ai/zen/v1/chat/completions`      | `@ai-sdk/openai-compatible` |
 
-[model id](/docs/config/#models) u OpenCode konfiguraciji koristi format `opencode/<model-id>`. Na primjer, za GPT 5.2 Codex u konfiguraciji koristite `opencode/gpt-5.2-codex`.
+[model id](/docs/config/#models) u vašoj OpenCode konfiguraciji koristi format
+`opencode/<model-id>`. Na primjer, za GPT 5.3 Codex u konfiguraciji biste
+koristili `opencode/gpt-5.3-codex`.
 
 ---
 
 ### Modeli
 
-Pun spisak dostupnih modela i metapodataka mozete preuzeti na:
+Pun spisak dostupnih modela i njihovih metapodataka možete preuzeti na:
 
 ```
 https://opencode.ai/zen/v1/models
@@ -105,143 +117,163 @@ https://opencode.ai/zen/v1/models
 
 ## Cijene
 
-Podrzavamo pay-as-you-go model. Ispod su cijene **po 1M tokena**.
+Podržavamo pay-as-you-go model. Ispod su cijene **po 1M tokena**.
 
-| Model                             | Input  | Output | Cached Read | Cached Write |
-| --------------------------------- | ------ | ------ | ----------- | ------------ |
-| Big Pickle                        | Free   | Free   | Free        | -            |
-| MiniMax M2.5 Free                 | Free   | Free   | Free        | -            |
-| MiniMax M2.5                      | $0.30  | $1.20  | $0.06       | -            |
-| MiniMax M2.1                      | $0.30  | $1.20  | $0.10       | -            |
-| GLM 5                             | $1.00  | $3.20  | $0.20       | -            |
-| GLM 4.7                           | $0.60  | $2.20  | $0.10       | -            |
-| GLM 4.6                           | $0.60  | $2.20  | $0.10       | -            |
-| Kimi K2.5                         | $0.60  | $3.00  | $0.08       | -            |
-| Kimi K2 Thinking                  | $0.40  | $2.50  | -           | -            |
-| Kimi K2                           | $0.40  | $2.50  | -           | -            |
-| Qwen3 Coder 480B                  | $0.45  | $1.50  | -           | -            |
-| Claude Opus 4.6 (≤ 200K tokens)   | $5.00  | $25.00 | $0.50       | $6.25        |
-| Claude Opus 4.6 (> 200K tokens)   | $10.00 | $37.50 | $1.00       | $12.50       |
-| Claude Opus 4.5                   | $5.00  | $25.00 | $0.50       | $6.25        |
-| Claude Opus 4.1                   | $15.00 | $75.00 | $1.50       | $18.75       |
-| Claude Sonnet 4.6 (≤ 200K tokens) | $3.00  | $15.00 | $0.30       | $3.75        |
-| Claude Sonnet 4.6 (> 200K tokens) | $6.00  | $22.50 | $0.60       | $7.50        |
-| Claude Sonnet 4.5 (≤ 200K tokens) | $3.00  | $15.00 | $0.30       | $3.75        |
-| Claude Sonnet 4.5 (> 200K tokens) | $6.00  | $22.50 | $0.60       | $7.50        |
-| Claude Sonnet 4 (≤ 200K tokens)   | $3.00  | $15.00 | $0.30       | $3.75        |
-| Claude Sonnet 4 (> 200K tokens)   | $6.00  | $22.50 | $0.60       | $7.50        |
-| Claude Haiku 4.5                  | $1.00  | $5.00  | $0.10       | $1.25        |
-| Claude Haiku 3.5                  | $0.80  | $4.00  | $0.08       | $1.00        |
-| Gemini 3.1 Pro (≤ 200K tokens)    | $2.00  | $12.00 | $0.20       | -            |
-| Gemini 3.1 Pro (> 200K tokens)    | $4.00  | $18.00 | $0.40       | -            |
-| Gemini 3 Pro (≤ 200K tokens)      | $2.00  | $12.00 | $0.20       | -            |
-| Gemini 3 Pro (> 200K tokens)      | $4.00  | $18.00 | $0.40       | -            |
-| Gemini 3 Flash                    | $0.50  | $3.00  | $0.05       | -            |
-| GPT 5.4                           | $2.50  | $15.00 | $0.25       | -            |
-| GPT 5.3 Codex                     | $1.75  | $14.00 | $0.175      | -            |
-| GPT 5.2                           | $1.75  | $14.00 | $0.175      | -            |
-| GPT 5.2 Codex                     | $1.75  | $14.00 | $0.175      | -            |
-| GPT 5.1                           | $1.07  | $8.50  | $0.107      | -            |
-| GPT 5.1 Codex                     | $1.07  | $8.50  | $0.107      | -            |
-| GPT 5.1 Codex Max                 | $1.25  | $10.00 | $0.125      | -            |
-| GPT 5.1 Codex Mini                | $0.25  | $2.00  | $0.025      | -            |
-| GPT 5                             | $1.07  | $8.50  | $0.107      | -            |
-| GPT 5 Codex                       | $1.07  | $8.50  | $0.107      | -            |
-| GPT 5 Nano                        | Free   | Free   | Free        | -            |
+| Model                             | Input  | Output  | Cached Read | Cached Write |
+| --------------------------------- | ------ | ------- | ----------- | ------------ |
+| Big Pickle                        | Free   | Free    | Free        | -            |
+| MiMo V2 Pro Free                  | Free   | Free    | Free        | -            |
+| MiMo V2 Omni Free                 | Free   | Free    | Free        | -            |
+| Qwen3.6 Plus Free                 | Free   | Free    | Free        | -            |
+| Nemotron 3 Super Free             | Free   | Free    | Free        | -            |
+| MiniMax M2.5 Free                 | Free   | Free    | Free        | -            |
+| MiniMax M2.5                      | $0.30  | $1.20   | $0.06       | $0.375       |
+| GLM 5                             | $1.00  | $3.20   | $0.20       | -            |
+| Kimi K2.5                         | $0.60  | $3.00   | $0.10       | -            |
+| Qwen3 Coder 480B                  | $0.45  | $1.50   | -           | -            |
+| Claude Opus 4.6                   | $5.00  | $25.00  | $0.50       | $6.25        |
+| Claude Opus 4.5                   | $5.00  | $25.00  | $0.50       | $6.25        |
+| Claude Opus 4.1                   | $15.00 | $75.00  | $1.50       | $18.75       |
+| Claude Sonnet 4.6                 | $3.00  | $15.00  | $0.30       | $3.75        |
+| Claude Sonnet 4.5 (≤ 200K tokens) | $3.00  | $15.00  | $0.30       | $3.75        |
+| Claude Sonnet 4.5 (> 200K tokens) | $6.00  | $22.50  | $0.60       | $7.50        |
+| Claude Sonnet 4 (≤ 200K tokens)   | $3.00  | $15.00  | $0.30       | $3.75        |
+| Claude Sonnet 4 (> 200K tokens)   | $6.00  | $22.50  | $0.60       | $7.50        |
+| Claude Haiku 4.5                  | $1.00  | $5.00   | $0.10       | $1.25        |
+| Claude Haiku 3.5                  | $0.80  | $4.00   | $0.08       | $1.00        |
+| Gemini 3.1 Pro (≤ 200K tokens)    | $2.00  | $12.00  | $0.20       | -            |
+| Gemini 3.1 Pro (> 200K tokens)    | $4.00  | $18.00  | $0.40       | -            |
+| Gemini 3 Flash                    | $0.50  | $3.00   | $0.05       | -            |
+| GPT 5.4                           | $2.50  | $15.00  | $0.25       | -            |
+| GPT 5.4 Pro                       | $30.00 | $180.00 | $30.00      | -            |
+| GPT 5.4 Mini                      | $0.75  | $4.50   | $0.075      | -            |
+| GPT 5.4 Nano                      | $0.20  | $1.25   | $0.02       | -            |
+| GPT 5.3 Codex Spark               | $1.75  | $14.00  | $0.175      | -            |
+| GPT 5.3 Codex                     | $1.75  | $14.00  | $0.175      | -            |
+| GPT 5.2                           | $1.75  | $14.00  | $0.175      | -            |
+| GPT 5.2 Codex                     | $1.75  | $14.00  | $0.175      | -            |
+| GPT 5.1                           | $1.07  | $8.50   | $0.107      | -            |
+| GPT 5.1 Codex                     | $1.07  | $8.50   | $0.107      | -            |
+| GPT 5.1 Codex Max                 | $1.25  | $10.00  | $0.125      | -            |
+| GPT 5.1 Codex Mini                | $0.25  | $2.00   | $0.025      | -            |
+| GPT 5                             | $1.07  | $8.50   | $0.107      | -            |
+| GPT 5 Codex                       | $1.07  | $8.50   | $0.107      | -            |
+| GPT 5 Nano                        | Free   | Free    | Free        | -            |
 
-U historiji koristenja mozete primijetiti _Claude Haiku 3.5_. To je [low cost model](/docs/config/#models) koji se koristi za generisanje naslova sesija.
+Možda ćete primijetiti _Claude Haiku 3.5_ u historiji korištenja. To je [low cost model](/docs/config/#models) koji se koristi za generisanje naslova vaših sesija.
 
 :::note
-Naknade kartica se prenose po stvarnom trosku (4.4% + $0.30 po transakciji) i ne naplacujemo nista preko toga.
+Naknade za kreditne kartice prosljeđujemo po stvarnom trošku (4.4% + $0.30 po transakciji); ne naplaćujemo ništa preko toga.
 :::
 
 Besplatni modeli:
 
-- MiniMax M2.5 Free je dostupan na OpenCode ograniceno vrijeme. Tim koristi taj period za prikupljanje povratnih informacija i poboljsanje modela.
-- Big Pickle je stealth model koji je besplatan na OpenCode ograniceno vrijeme. Tim koristi taj period za prikupljanje povratnih informacija i poboljsanje modela.
+- MiniMax M2.5 Free je dostupan na OpenCode ograničeno vrijeme. Tim koristi ovo vrijeme da prikupi povratne informacije i poboljša model.
+- MiMo V2 Pro Free je dostupan na OpenCode ograničeno vrijeme. Tim koristi ovo vrijeme da prikupi povratne informacije i poboljša model.
+- MiMo V2 Omni Free je dostupan na OpenCode ograničeno vrijeme. Tim koristi ovo vrijeme da prikupi povratne informacije i poboljša model.
+- Qwen3.6 Plus Free je dostupan na OpenCode ograničeno vrijeme. Tim koristi ovo vrijeme da prikupi povratne informacije i poboljša model.
+- Nemotron 3 Super Free je dostupan na OpenCode ograničeno vrijeme. Tim koristi ovo vrijeme da prikupi povratne informacije i poboljša model.
+- Big Pickle je stealth model koji je besplatan na OpenCode ograničeno vrijeme. Tim koristi ovo vrijeme da prikupi povratne informacije i poboljša model.
 
-Ako imate pitanja, <a href={email}>kontaktirajte nas</a>.
+<a href={email}>Kontaktirajte nas</a> ako imate bilo kakvih pitanja.
 
 ---
 
-### Automatska dopuna
+### Auto-reload
 
-Ako vam stanje padne ispod $5, Zen ce automatski dopuniti $20.
+Ako vam stanje padne ispod $5, Zen će automatski dopuniti $20.
 
-Iznos auto-reloada mozete promijeniti. Auto-reload mozete i potpuno iskljuciti.
+Možete promijeniti iznos auto-reload dopune. Auto-reload možete i potpuno
+isključiti.
 
 ---
 
 ### Mjesečni limiti
 
-Mozete postaviti mjesecni limit potrosnje za cijeli workspace i za svakog clana tima.
+Možete postaviti mjesečni limit korištenja za cijeli workspace i za svakog člana
+vašeg tima.
 
-Na primjer, ako postavite mjesecni limit na $20, Zen nece potrositi vise od $20 u mjesecu. Ali ako je auto-reload ukljucen, ukupna naplata moze preci $20 ako stanje padne ispod $5.
+Na primjer, recimo da postavite mjesečni limit korištenja na $20 — Zen neće
+potrošiti više od $20 u toku mjeseca. Ali ako imate uključen auto-reload, Zen
+vam ipak može naplatiti više od $20 ako vam stanje padne ispod $5.
 
 ---
 
 ### Zastarjeli modeli
 
-| Model            | Datum ukidanja |
-| ---------------- | -------------- |
-| Qwen3 Coder 480B | 6. feb. 2026.  |
-| Kimi K2 Thinking | 6. mart 2026.  |
-| Kimi K2          | 6. mart 2026.  |
-| MiniMax M2.1     | 15. mart 2026. |
-| GLM 4.7          | 15. mart 2026. |
-| GLM 4.6          | 15. mart 2026. |
+| Model            | Datum zastarijevanja |
+| ---------------- | -------------------- |
+| MiniMax M2.1     | March 15, 2026       |
+| GLM 4.7          | March 15, 2026       |
+| GLM 4.6          | March 15, 2026       |
+| Gemini 3 Pro     | March 9, 2026        |
+| Kimi K2 Thinking | March 6, 2026        |
+| Kimi K2          | March 6, 2026        |
+| Qwen3 Coder 480B | Feb 6, 2026          |
 
 ---
 
 ## Privatnost
 
-Svi nasi modeli su hostovani u SAD-u. Provajderi prate zero-retention politiku i ne koriste vase podatke za treniranje modela, uz sljedece izuzetke:
+Svi naši modeli su hostovani u US. Naši provajderi prate zero-retention politiku
+i ne koriste vaše podatke za treniranje modela, uz sljedeće izuzetke:
 
-- Big Pickle: Tokom besplatnog perioda, prikupljeni podaci mogu se koristiti za poboljsanje modela.
-- MiniMax M2.5 Free: Tokom besplatnog perioda, prikupljeni podaci mogu se koristiti za poboljsanje modela.
-- OpenAI API-ji: Zahtjevi se cuvaju 30 dana prema [OpenAI's Data Policies](https://platform.openai.com/docs/guides/your-data).
-- Anthropic API-ji: Zahtjevi se cuvaju 30 dana prema [Anthropic's Data Policies](https://docs.anthropic.com/en/docs/claude-code/data-usage).
+- Big Pickle: Tokom besplatnog perioda, prikupljeni podaci mogu se koristiti za poboljšanje modela.
+- MiniMax M2.5 Free: Tokom besplatnog perioda, prikupljeni podaci mogu se koristiti za poboljšanje modela.
+- MiMo V2 Pro Free: Tokom besplatnog perioda, prikupljeni podaci mogu se koristiti za poboljšanje modela.
+- MiMo V2 Omni Free: Tokom besplatnog perioda, prikupljeni podaci mogu se koristiti za poboljšanje modela.
+- Qwen3.6 Plus Free: Tokom besplatnog perioda, prikupljeni podaci mogu se koristiti za poboljšanje modela.
+- Nemotron 3 Super Free: Tokom besplatnog perioda, prikupljeni podaci mogu se koristiti za poboljšanje modela.
+- OpenAI APIs: Requests are retained for 30 days in accordance with [OpenAI's Data Policies](https://platform.openai.com/docs/guides/your-data).
+- Anthropic APIs: Requests are retained for 30 days in accordance with [Anthropic's Data Policies](https://docs.anthropic.com/en/docs/claude-code/data-usage).
 
 ---
 
 ## Za timove
 
-Zen odlicno radi i za timove. Mozete pozvati clanove tima, dodijeliti uloge, birati modele koje tim koristi i jos mnogo toga.
+Zen odlično radi i za timove. Možete pozvati saigrače, dodijeliti uloge, birati
+modele koje vaš tim koristi i još mnogo toga.
 
 :::note
 Workspaces su trenutno besplatni za timove kao dio beta faze.
 :::
 
-Upravljanje workspace-om je trenutno besplatno za timove tokom beta faze. Vise detalja o cijenama podijelit cemo uskoro.
+Upravljanje vašim workspace-om je trenutno besplatno za timove kao dio beta
+faze. Uskoro ćemo podijeliti više detalja o cijenama.
 
 ---
 
 ### Uloge
 
-Mozete pozvati clanove tima u workspace i dodijeliti uloge:
+Možete pozvati saigrače u svoj workspace i dodijeliti im uloge:
 
-- **Admin**: upravlja modelima, clanovima, API kljucevima i billingom
-- **Member**: upravlja samo svojim API kljucevima
+- **Admin**: Upravlja modelima, članovima, API ključevima i naplatom
+- **Member**: Upravlja samo vlastitim API ključevima
 
-Admini mogu postaviti i mjesecne limite potrosnje po clanu da drze troskove pod kontrolom.
+Admini mogu postaviti i mjesečne limite potrošnje za svakog člana kako bi držali
+troškove pod kontrolom.
 
 ---
 
 ### Pristup modelima
 
-Admini mogu ukljuciti ili iskljuciti odredene modele za workspace. Zahtjevi prema iskljucenom modelu vracaju gresku.
+Admini mogu uključiti ili isključiti određene modele za workspace. Zahtjevi
+poslani prema isključenom modelu vratiće grešku.
 
-Ovo je korisno kada zelite zabraniti model koji prikuplja podatke.
+Ovo je korisno u slučajevima kada želite onemogućiti korištenje modela koji
+prikuplja podatke.
 
 ---
 
 ### Donesite vlastiti ključ
 
-Mozete koristiti vlastite OpenAI ili Anthropic API kljuceve i dalje koristiti ostale modele u Zen-u.
+Možete koristiti vlastite OpenAI ili Anthropic API ključeve dok i dalje imate
+pristup drugim modelima u Zen.
 
-Kada koristite vlastite kljuceve, tokene direktno naplacuje provajder, ne Zen.
+Kada koristite vlastite ključeve, tokene direktno naplaćuje provajder, a ne Zen.
 
-Na primjer, vasa organizacija mozda vec ima OpenAI ili Anthropic kljuc i zelite koristiti njega umjesto onog koji daje Zen.
+Na primjer, vaša organizacija možda već ima OpenAI ili Anthropic ključ i želite
+koristiti njega umjesto onog koji pruža Zen.
 
 ---
 
@@ -249,7 +281,7 @@ Na primjer, vasa organizacija mozda vec ima OpenAI ili Anthropic kljuc i zelite
 
 OpenCode Zen smo napravili da:
 
-1. **Benchmarkiramo** najbolje kombinacije model/provajder za coding agente.
-2. Omogucimo pristup opcijama **najviseg kvaliteta** bez degradacije performansi i preusmjeravanja na jeftinije provajdere.
-3. Prenesemo svaka **snizenja cijena** prodajom po trosku, tako da je jedini markup pokrice processing naknada.
-4. Obezbijedimo **bez lock-ina** tako da Zen mozete koristiti sa bilo kojim coding agentom, uz slobodu koristenja drugih provajdera u OpenCode.
+1. **Benchmarkiramo** najbolje modele/provajdere za coding agente.
+2. Imamo pristup opcijama **najvišeg kvaliteta** bez snižavanja performansi ili preusmjeravanja na jeftinije provajdere.
+3. Prenesemo sva **sniženja cijena** prodajom po stvarnom trošku; tako da je jedini markup pokrivanje naših processing naknada.
+4. Obezbijedimo **bez lock-ina** time što vam omogućavamo da ga koristite s bilo kojim drugim coding agentom. I da vam uvijek omogućimo da koristite bilo koji drugi provajder i u OpenCode.

packages/web/src/content/docs/da/zen.mdx

@@ -13,33 +13,33 @@ OpenCode Zen er en liste over testede og verificerede modeller leveret af OpenCo
 OpenCode Zen er i øjeblikket i beta.
 :::
 
-Zen fungerer som alle andre udbydere i OpenCode. Du logger på OpenCode Zen og får
-din API-nøgle. Den er **helt valgfri** og du behøver ikke bruge den for at bruge
+Zen fungerer som enhver anden udbyder i OpenCode. Du logger ind på OpenCode Zen og får
+din API-nøgle. Det er **helt valgfrit**, og du behøver ikke bruge det for at bruge
 OpenCode.
 
 ---
 
 ## Baggrund
 
-Der er et stort antal modeller derude, men kun få af dem
+Der findes et stort antal modeller, men kun få af dem
 fungerer godt som kodeagenter. Derudover er de fleste udbydere
 konfigureret meget forskelligt, så du får meget forskellig ydeevne og kvalitet.
 
 :::tip
-Vi testede en udvalgt gruppe modeller og udbydere, der fungerer godt med OpenCode.
+Vi testede en udvalgt gruppe modeller og udbydere, som fungerer godt med OpenCode.
 :::
 
 Så hvis du bruger en model gennem noget som OpenRouter, kan du aldrig være
-sikker på, om du får den bedste version af modellen, du ønsker.
+sikker på, om du får den bedste version af den model, du vil have.
 
-For at fikse dette gjorde vi et par ting:
+For at løse det gjorde vi et par ting:
 
 1. Vi testede en udvalgt gruppe modeller og talte med deres teams om, hvordan
-   man bedst kører dem.
-2. Vi samarbejdede derefter med nogle få udbydere for at sikre, at disse blev serveret
+   de bedst køres.
+2. Derefter arbejdede vi sammen med nogle få udbydere for at sikre, at de blev leveret
    korrekt.
-3. Til sidst benchmarkede vi kombinationen af model/udbyder og kom frem
-   til en liste, som vi har lyst til at anbefale.
+3. Til sidst benchmarkede vi kombinationen af model/udbyder og lavede
+   en liste, som vi trygt kan anbefale.
 
 OpenCode Zen er en AI gateway, der giver dig adgang til disse modeller.
 
@@ -47,14 +47,14 @@ OpenCode Zen er en AI gateway, der giver dig adgang til disse modeller.
 
 ## Sådan fungerer det
 
-OpenCode Zen fungerer som alle andre udbydere i OpenCode.
+OpenCode Zen fungerer som enhver anden udbyder i OpenCode.
 
-1. Du logger på **<a href={console}>OpenCode Zen</a>**, tilføjer faktureringsoplysninger
-   og kopierer API-nøglen.
-2. Du kører kommandoen `/connect` i TUI, vælger OpenCode Zen og indsætter API-nøglen.
+1. Du logger ind på **<a href={console}>OpenCode Zen</a>**, tilføjer dine faktureringsoplysninger
+   og kopierer din API-nøgle.
+2. Du kører kommandoen `/connect` i TUI, vælger OpenCode Zen og indsætter din API-nøgle.
 3. Kør `/models` i TUI for at se listen over modeller, vi anbefaler.
 
-Du opkræves per anmodning, og du kan tilføje kredit til din konto.
+Du bliver opkrævet pr. anmodning, og du kan tilføje kredit til din konto.
 
 ---
 
@@ -62,51 +62,52 @@ Du opkræves per anmodning, og du kan tilføje kredit til din konto.
 
 Du kan også få adgang til vores modeller gennem følgende API-endpoints.
 
-| Model               | Model ID           | Endpoint                                           | AI SDK Pakke                |
-| ------------------- | ------------------ | -------------------------------------------------- | --------------------------- |
-| GPT 5.4             | gpt-5.4            | `https://opencode.ai/zen/v1/responses`             | `@ai-sdk/openai`            |
-| GPT 5.3 Codex       | gpt-5.3-codex      | `https://opencode.ai/zen/v1/responses`             | `@ai-sdk/openai`            |
-| GPT 5.2             | gpt-5.2            | `https://opencode.ai/zen/v1/responses`             | `@ai-sdk/openai`            |
-| GPT 5.2 Codex       | gpt-5.2-codex      | `https://opencode.ai/zen/v1/responses`             | `@ai-sdk/openai`            |
-| GPT 5.1             | gpt-5.1            | `https://opencode.ai/zen/v1/responses`             | `@ai-sdk/openai`            |
-| GPT 5.1 Codex       | gpt-5.1-codex      | `https://opencode.ai/zen/v1/responses`             | `@ai-sdk/openai`            |
-| GPT 5.1 Codex Max   | gpt-5.1-codex-max  | `https://opencode.ai/zen/v1/responses`             | `@ai-sdk/openai`            |
-| GPT 5.1 Codex Mini  | gpt-5.1-codex-mini | `https://opencode.ai/zen/v1/responses`             | `@ai-sdk/openai`            |
-| GPT 5               | gpt-5              | `https://opencode.ai/zen/v1/responses`             | `@ai-sdk/openai`            |
-| GPT 5 Codex         | gpt-5-codex        | `https://opencode.ai/zen/v1/responses`             | `@ai-sdk/openai`            |
-| GPT 5 Nano          | gpt-5-nano         | `https://opencode.ai/zen/v1/responses`             | `@ai-sdk/openai`            |
-| Claude Opus 4.6     | claude-opus-4-6    | `https://opencode.ai/zen/v1/messages`              | `@ai-sdk/anthropic`         |
-| Claude Opus 4.5     | claude-opus-4-5    | `https://opencode.ai/zen/v1/messages`              | `@ai-sdk/anthropic`         |
-| Claude Opus 4.1     | claude-opus-4-1    | `https://opencode.ai/zen/v1/messages`              | `@ai-sdk/anthropic`         |
-| Claude Sonnet 4.6   | claude-sonnet-4-6  | `https://opencode.ai/zen/v1/messages`              | `@ai-sdk/anthropic`         |
-| Claude Sonnet 4.5   | claude-sonnet-4-5  | `https://opencode.ai/zen/v1/messages`              | `@ai-sdk/anthropic`         |
-| Claude Sonnet 4     | claude-sonnet-4    | `https://opencode.ai/zen/v1/messages`              | `@ai-sdk/anthropic`         |
-| Claude Haiku 4.5    | claude-haiku-4-5   | `https://opencode.ai/zen/v1/messages`              | `@ai-sdk/anthropic`         |
-| Claude Haiku 3.5    | claude-3-5-haiku   | `https://opencode.ai/zen/v1/messages`              | `@ai-sdk/anthropic`         |
-| Gemini 3.1 Pro      | gemini-3.1-pro     | `https://opencode.ai/zen/v1/models/gemini-3.1-pro` | `@ai-sdk/google`            |
-| Gemini 3 Pro        | gemini-3-pro       | `https://opencode.ai/zen/v1/models/gemini-3-pro`   | `@ai-sdk/google`            |
-| Gemini 3 Flash      | gemini-3-flash     | `https://opencode.ai/zen/v1/models/gemini-3-flash` | `@ai-sdk/google`            |
-| MiniMax M2.5        | minimax-m2.5       | `https://opencode.ai/zen/v1/chat/completions`      | `@ai-sdk/openai-compatible` |
-| MiniMax M2.5 Gratis | minimax-m2.5-free  | `https://opencode.ai/zen/v1/chat/completions`      | `@ai-sdk/openai-compatible` |
-| MiniMax M2.1        | minimax-m2.1       | `https://opencode.ai/zen/v1/chat/completions`      | `@ai-sdk/openai-compatible` |
-| GLM 5               | glm-5              | `https://opencode.ai/zen/v1/chat/completions`      | `@ai-sdk/openai-compatible` |
-| GLM 4.7             | glm-4.7            | `https://opencode.ai/zen/v1/chat/completions`      | `@ai-sdk/openai-compatible` |
-| GLM 4.6             | glm-4.6            | `https://opencode.ai/zen/v1/chat/completions`      | `@ai-sdk/openai-compatible` |
-| Kimi K2.5           | kimi-k2.5          | `https://opencode.ai/zen/v1/chat/completions`      | `@ai-sdk/openai-compatible` |
-| Kimi K2 Tenker      | kimi-k2-thinking   | `https://opencode.ai/zen/v1/chat/completions`      | `@ai-sdk/openai-compatible` |
-| Kimi K2             | kimi-k2            | `https://opencode.ai/zen/v1/chat/completions`      | `@ai-sdk/openai-compatible` |
-| Qwen3-koder 480B    | qwen3-coder        | `https://opencode.ai/zen/v1/chat/completions`      | `@ai-sdk/openai-compatible` |
-| Stor sylteagurk     | big-pickle         | `https://opencode.ai/zen/v1/chat/completions`      | `@ai-sdk/openai-compatible` |
+| Model                 | Model ID              | Endpoint                                           | AI SDK-pakke                |
+| --------------------- | --------------------- | -------------------------------------------------- | --------------------------- |
+| GPT 5.4               | gpt-5.4               | `https://opencode.ai/zen/v1/responses`             | `@ai-sdk/openai`            |
+| GPT 5.4 Pro           | gpt-5.4-pro           | `https://opencode.ai/zen/v1/responses`             | `@ai-sdk/openai`            |
+| GPT 5.4 Mini          | gpt-5.4-mini          | `https://opencode.ai/zen/v1/responses`             | `@ai-sdk/openai`            |
+| GPT 5.4 Nano          | gpt-5.4-nano          | `https://opencode.ai/zen/v1/responses`             | `@ai-sdk/openai`            |
+| GPT 5.3 Codex         | gpt-5.3-codex         | `https://opencode.ai/zen/v1/responses`             | `@ai-sdk/openai`            |
+| GPT 5.3 Codex Spark   | gpt-5.3-codex-spark   | `https://opencode.ai/zen/v1/responses`             | `@ai-sdk/openai`            |
+| GPT 5.2               | gpt-5.2               | `https://opencode.ai/zen/v1/responses`             | `@ai-sdk/openai`            |
+| GPT 5.2 Codex         | gpt-5.2-codex         | `https://opencode.ai/zen/v1/responses`             | `@ai-sdk/openai`            |
+| GPT 5.1               | gpt-5.1               | `https://opencode.ai/zen/v1/responses`             | `@ai-sdk/openai`            |
+| GPT 5.1 Codex         | gpt-5.1-codex         | `https://opencode.ai/zen/v1/responses`             | `@ai-sdk/openai`            |
+| GPT 5.1 Codex Max     | gpt-5.1-codex-max     | `https://opencode.ai/zen/v1/responses`             | `@ai-sdk/openai`            |
+| GPT 5.1 Codex Mini    | gpt-5.1-codex-mini    | `https://opencode.ai/zen/v1/responses`             | `@ai-sdk/openai`            |
+| GPT 5                 | gpt-5                 | `https://opencode.ai/zen/v1/responses`             | `@ai-sdk/openai`            |
+| GPT 5 Codex           | gpt-5-codex           | `https://opencode.ai/zen/v1/responses`             | `@ai-sdk/openai`            |
+| GPT 5 Nano            | gpt-5-nano            | `https://opencode.ai/zen/v1/responses`             | `@ai-sdk/openai`            |
+| Claude Opus 4.6       | claude-opus-4-6       | `https://opencode.ai/zen/v1/messages`              | `@ai-sdk/anthropic`         |
+| Claude Opus 4.5       | claude-opus-4-5       | `https://opencode.ai/zen/v1/messages`              | `@ai-sdk/anthropic`         |
+| Claude Opus 4.1       | claude-opus-4-1       | `https://opencode.ai/zen/v1/messages`              | `@ai-sdk/anthropic`         |
+| Claude Sonnet 4.6     | claude-sonnet-4-6     | `https://opencode.ai/zen/v1/messages`              | `@ai-sdk/anthropic`         |
+| Claude Sonnet 4.5     | claude-sonnet-4-5     | `https://opencode.ai/zen/v1/messages`              | `@ai-sdk/anthropic`         |
+| Claude Sonnet 4       | claude-sonnet-4       | `https://opencode.ai/zen/v1/messages`              | `@ai-sdk/anthropic`         |
+| Claude Haiku 4.5      | claude-haiku-4-5      | `https://opencode.ai/zen/v1/messages`              | `@ai-sdk/anthropic`         |
+| Claude Haiku 3.5      | claude-3-5-haiku      | `https://opencode.ai/zen/v1/messages`              | `@ai-sdk/anthropic`         |
+| Gemini 3.1 Pro        | gemini-3.1-pro        | `https://opencode.ai/zen/v1/models/gemini-3.1-pro` | `@ai-sdk/google`            |
+| Gemini 3 Flash        | gemini-3-flash        | `https://opencode.ai/zen/v1/models/gemini-3-flash` | `@ai-sdk/google`            |
+| MiniMax M2.5          | minimax-m2.5          | `https://opencode.ai/zen/v1/chat/completions`      | `@ai-sdk/openai-compatible` |
+| MiniMax M2.5 Free     | minimax-m2.5-free     | `https://opencode.ai/zen/v1/chat/completions`      | `@ai-sdk/openai-compatible` |
+| GLM 5                 | glm-5                 | `https://opencode.ai/zen/v1/chat/completions`      | `@ai-sdk/openai-compatible` |
+| Kimi K2.5             | kimi-k2.5             | `https://opencode.ai/zen/v1/chat/completions`      | `@ai-sdk/openai-compatible` |
+| Big Pickle            | big-pickle            | `https://opencode.ai/zen/v1/chat/completions`      | `@ai-sdk/openai-compatible` |
+| MiMo V2 Pro Free      | mimo-v2-pro-free      | `https://opencode.ai/zen/v1/chat/completions`      | `@ai-sdk/openai-compatible` |
+| MiMo V2 Omni Free     | mimo-v2-omni-free     | `https://opencode.ai/zen/v1/chat/completions`      | `@ai-sdk/openai-compatible` |
+| Qwen3.6 Plus Free     | qwen3.6-plus-free     | `https://opencode.ai/zen/v1/chat/completions`      | `@ai-sdk/openai-compatible` |
+| Nemotron 3 Super Free | nemotron-3-super-free | `https://opencode.ai/zen/v1/chat/completions`      | `@ai-sdk/openai-compatible` |
 
-[model-id](/docs/config/#models) i OpenCode-konfigurationen
-bruger formatet `opencode/<model-id>`. For eksempel, for GPT 5.2 Codex, ville du
-bruge `opencode/gpt-5.2-codex` i din konfiguration.
+[model id](/docs/config/#models) i din OpenCode-konfiguration
+bruger formatet `opencode/<model-id>`. For eksempel ville du for GPT 5.3 Codex
+bruge `opencode/gpt-5.3-codex` i din konfiguration.
 
 ---
 
 ### Modeller
 
-Du kan hente hele listen over tilgængelige modeller og deres metadata fra:
+Du kan hente den fulde liste over tilgængelige modeller og deres metadata fra:
 
 ```
 https://opencode.ai/zen/v1/models
@@ -116,81 +117,84 @@ https://opencode.ai/zen/v1/models
 
 ## Priser
 
-Vi støtter en pay-as-you-go-model. Nedenfor er priserne **per 1 million tokens**.
+Vi understøtter en pay-as-you-go-model. Nedenfor er priserne **pr. 1M tokens**.
 
-| Model                             | Input  | Output | Cached Læs | Cached Skriv |
-| --------------------------------- | ------ | ------ | ---------- | ------------ |
-| Stor sylteagurk                   | Gratis | Gratis | Gratis     | -            |
-| MiniMax M2.5 Gratis               | Gratis | Gratis | Gratis     | -            |
-| MiniMax M2.5                      | $0,30  | $1,20  | $0,06      | -            |
-| MiniMax M2.1                      | $0,30  | $1,20  | $0,10      | -            |
-| GLM 5                             | $1,00  | $3,20  | $0,20      | -            |
-| GLM 4.7                           | $0,60  | $2,20  | $0,10      | -            |
-| GLM 4.6                           | $0,60  | $2,20  | $0,10      | -            |
-| Kimi K2.5                         | $0,60  | $3,00  | $0,08      | -            |
-| Kimi K2 Tenker                    | $0,40  | $2,50  | -          | -            |
-| Kimi K2                           | $0,40  | $2,50  | -          | -            |
-| Qwen3-koder 480B                  | $0,45  | $1,50  | -          | -            |
-| Claude Opus 4.6 (≤ 200K tokens)   | $5,00  | $25,00 | $0,50      | $6,25        |
-| Claude Opus 4.6 (> 200K tokens)   | $10,00 | $37,50 | $1,00      | $12,50       |
-| Claude Opus 4.5                   | $5,00  | $25,00 | $0,50      | $6,25        |
-| Claude Opus 4.1                   | $15,00 | $75,00 | $1,50      | $18,75       |
-| Claude Sonnet 4.6 (≤ 200K tokens) | $3,00  | $15,00 | $0,30      | $3,75        |
-| Claude Sonnet 4.6 (> 200K tokens) | $6,00  | $22,50 | $0,60      | $7,50        |
-| Claude Sonnet 4.5 (≤ 200K tokens) | $3,00  | $15,00 | $0,30      | $3,75        |
-| Claude Sonnet 4.5 (> 200K tokens) | $6,00  | $22,50 | $0,60      | $7,50        |
-| Claude Sonnet 4 (≤ 200K tokens)   | $3,00  | $15,00 | $0,30      | $3,75        |
-| Claude Sonnet 4 (> 200K tokens)   | $6,00  | $22,50 | $0,60      | $7,50        |
-| Claude Haiku 4.5                  | $1,00  | $5,00  | $0,10      | $1,25        |
-| Claude Haiku 3.5                  | $0,80  | $4,00  | $0,08      | $1,00        |
-| Gemini 3.1 Pro (≤ 200K tokens)    | $2,00  | $12,00 | $0,20      | -            |
-| Gemini 3.1 Pro (> 200K tokens)    | $4,00  | $18,00 | $0,40      | -            |
-| Gemini 3 Pro (≤ 200K tokens)      | $2,00  | $12,00 | $0,20      | -            |
-| Gemini 3 Pro (> 200K tokens)      | $4,00  | $18,00 | $0,40      | -            |
-| Gemini 3 Flash                    | $0,50  | $3,00  | $0,05      | -            |
-| GPT 5.4                           | $2,50  | $15,00 | $0,25      | -            |
-| GPT 5.3 Codex                     | $1,75  | $14,00 | $0,175     | -            |
-| GPT 5.2                           | $1,75  | $14,00 | $0,175     | -            |
-| GPT 5.2 Codex                     | $1,75  | $14,00 | $0,175     | -            |
-| GPT 5.1                           | $1,07  | $8,50  | $0,107     | -            |
-| GPT 5.1 Codex                     | $1,07  | $8,50  | $0,107     | -            |
-| GPT 5.1 Codex Max                 | $1,25  | $10,00 | $0,125     | -            |
-| GPT 5.1 Codex Mini                | $0,25  | $2,00  | $0,025     | -            |
-| GPT 5                             | $1,07  | $8,50  | $0,107     | -            |
-| GPT 5 Codex                       | $1,07  | $8,50  | $0,107     | -            |
-| GPT 5 Nano                        | Gratis | Gratis | Gratis     | -            |
+| Model                             | Input  | Output  | Cached Read | Cached Write |
+| --------------------------------- | ------ | ------- | ----------- | ------------ |
+| Big Pickle                        | Free   | Free    | Free        | -            |
+| MiMo V2 Pro Free                  | Free   | Free    | Free        | -            |
+| MiMo V2 Omni Free                 | Free   | Free    | Free        | -            |
+| Qwen3.6 Plus Free                 | Free   | Free    | Free        | -            |
+| Nemotron 3 Super Free             | Free   | Free    | Free        | -            |
+| MiniMax M2.5 Free                 | Free   | Free    | Free        | -            |
+| MiniMax M2.5                      | $0.30  | $1.20   | $0.06       | $0.375       |
+| GLM 5                             | $1.00  | $3.20   | $0.20       | -            |
+| Kimi K2.5                         | $0.60  | $3.00   | $0.10       | -            |
+| Qwen3 Coder 480B                  | $0.45  | $1.50   | -           | -            |
+| Claude Opus 4.6                   | $5.00  | $25.00  | $0.50       | $6.25        |
+| Claude Opus 4.5                   | $5.00  | $25.00  | $0.50       | $6.25        |
+| Claude Opus 4.1                   | $15.00 | $75.00  | $1.50       | $18.75       |
+| Claude Sonnet 4.6                 | $3.00  | $15.00  | $0.30       | $3.75        |
+| Claude Sonnet 4.5 (≤ 200K tokens) | $3.00  | $15.00  | $0.30       | $3.75        |
+| Claude Sonnet 4.5 (> 200K tokens) | $6.00  | $22.50  | $0.60       | $7.50        |
+| Claude Sonnet 4 (≤ 200K tokens)   | $3.00  | $15.00  | $0.30       | $3.75        |
+| Claude Sonnet 4 (> 200K tokens)   | $6.00  | $22.50  | $0.60       | $7.50        |
+| Claude Haiku 4.5                  | $1.00  | $5.00   | $0.10       | $1.25        |
+| Claude Haiku 3.5                  | $0.80  | $4.00   | $0.08       | $1.00        |
+| Gemini 3.1 Pro (≤ 200K tokens)    | $2.00  | $12.00  | $0.20       | -            |
+| Gemini 3.1 Pro (> 200K tokens)    | $4.00  | $18.00  | $0.40       | -            |
+| Gemini 3 Flash                    | $0.50  | $3.00   | $0.05       | -            |
+| GPT 5.4                           | $2.50  | $15.00  | $0.25       | -            |
+| GPT 5.4 Pro                       | $30.00 | $180.00 | $30.00      | -            |
+| GPT 5.4 Mini                      | $0.75  | $4.50   | $0.075      | -            |
+| GPT 5.4 Nano                      | $0.20  | $1.25   | $0.02       | -            |
+| GPT 5.3 Codex Spark               | $1.75  | $14.00  | $0.175      | -            |
+| GPT 5.3 Codex                     | $1.75  | $14.00  | $0.175      | -            |
+| GPT 5.2                           | $1.75  | $14.00  | $0.175      | -            |
+| GPT 5.2 Codex                     | $1.75  | $14.00  | $0.175      | -            |
+| GPT 5.1                           | $1.07  | $8.50   | $0.107      | -            |
+| GPT 5.1 Codex                     | $1.07  | $8.50   | $0.107      | -            |
+| GPT 5.1 Codex Max                 | $1.25  | $10.00  | $0.125      | -            |
+| GPT 5.1 Codex Mini                | $0.25  | $2.00   | $0.025      | -            |
+| GPT 5                             | $1.07  | $8.50   | $0.107      | -            |
+| GPT 5 Codex                       | $1.07  | $8.50   | $0.107      | -            |
+| GPT 5 Nano                        | Free   | Free    | Free        | -            |
 
-Du bemærker måske _Claude Haiku 3.5_ i din brugshistorik. Dette er en [lavprismodel](/docs/config/#models), som bruges til at generere titlerne på dine sessioner.
+Du vil måske bemærke _Claude Haiku 3.5_ i din brugshistorik. Det er en [lavprismodel](/docs/config/#models), som bruges til at generere titlerne på dine sessioner.
 
 :::note
-Kreditkortgebyrer overføres til kostpris (4,4 % + $0,30 per transaktion); vi opkræver ikke noget udover det.
+Kreditkortgebyrer videregives til kostpris (4.4% + $0.30 pr. transaktion); vi opkræver ikke noget ud over det.
 :::
 
 De gratis modeller:
 
-- MiniMax M2.5 Gratis er tilgængelig på OpenCode i en begrænset periode. Teamet bruger denne tid til at samle feedback og forbedre modellen.
-- Stor sylteagurk er en stealth-model som er gratis på OpenCode i en begrænset periode. Teamet bruger denne tid til at samle feedback og forbedre modellen.
+- MiniMax M2.5 Free er tilgængelig på OpenCode i en begrænset periode. Teamet bruger denne tid til at indsamle feedback og forbedre modellen.
+- MiMo V2 Pro Free er tilgængelig på OpenCode i en begrænset periode. Teamet bruger denne tid til at indsamle feedback og forbedre modellen.
+- MiMo V2 Omni Free er tilgængelig på OpenCode i en begrænset periode. Teamet bruger denne tid til at indsamle feedback og forbedre modellen.
+- Qwen3.6 Plus Free er tilgængelig på OpenCode i en begrænset periode. Teamet bruger denne tid til at indsamle feedback og forbedre modellen.
+- Nemotron 3 Super Free er tilgængelig på OpenCode i en begrænset periode. Teamet bruger denne tid til at indsamle feedback og forbedre modellen.
+- Big Pickle er en stealth-model, som er gratis på OpenCode i en begrænset periode. Teamet bruger denne tid til at indsamle feedback og forbedre modellen.
 
-<a href={email}>Kontakt os</a> hvis du har spørgsmål.
+<a href={email}>Kontakt os</a>, hvis du har spørgsmål.
 
 ---
 
-### Automatisk opfyldning
+### Automatisk genopfyldning
 
-Hvis din saldo går under $5, vil Zen automatisk laste $20 ind på nytt.
+Hvis din saldo kommer under $5, genopfylder Zen automatisk med $20.
 
-Du kan ændre beløbet for automatisk påfyldning. Du kan også deaktivere automatisk genindlæsning helt.
+Du kan ændre beløbet for automatisk genopfyldning. Du kan også deaktivere automatisk genopfyldning helt.
 
 ---
 
 ### Månedlige grænser
 
-Du kan også angive en månedlig brugsgrænse for hele arbejdsområdet og for hvert
+Du kan også sætte en månedlig forbrugsgrænse for hele arbejdsområdet og for hvert
 medlem af dit team.
 
-Lad os for eksempel sige, at du sætter en månedlig brugsgrænse til $20, Zen vil ikke bruge
-mere end $20 på en måned. Men hvis du har automatisk genindlæsning aktiveret, kan Zen ende med
-at opkræve dig mere end $20, hvis din saldo går under $5.
+Lad os for eksempel sige, at du sætter en månedlig forbrugsgrænse til $20. Zen vil ikke bruge
+mere end $20 på en måned. Men hvis du har automatisk genopfyldning aktiveret, kan Zen ende med
+at opkræve dig mere end $20, hvis din saldo kommer under $5.
 
 ---
 
@@ -198,43 +202,48 @@ at opkræve dig mere end $20, hvis din saldo går under $5.
 
 | Model            | Udfasningsdato |
 | ---------------- | -------------- |
-| Qwen3-koder 480B | 6. feb. 2026   |
-| Kimi K2 Tenker   | 6. marts 2026  |
-| Kimi K2          | 6. marts 2026  |
-| MiniMax M2.1     | 15. marts 2026 |
-| GLM 4.7          | 15. marts 2026 |
-| GLM 4.6          | 15. marts 2026 |
+| MiniMax M2.1     | March 15, 2026 |
+| GLM 4.7          | March 15, 2026 |
+| GLM 4.6          | March 15, 2026 |
+| Gemini 3 Pro     | March 9, 2026  |
+| Kimi K2 Thinking | March 6, 2026  |
+| Kimi K2          | March 6, 2026  |
+| Qwen3 Coder 480B | Feb 6, 2026    |
 
 ---
 
 ## Privatliv
 
-Alle vores modeller er hostet i USA. Vores udbydere følger en nul-opbevaringspolitik og bruger ikke dine data til modeltræning, med følgende undtagelser:
+Alle vores modeller hostes i US. Vores udbydere følger en nul-opbevaringspolitik og bruger ikke dine data til modeltræning, med følgende undtagelser:
 
-- Stor sylteagurk: I løbet af gratisperioden kan indsamlede data bruges til at forbedre modellen.
-- MiniMax M2.5 Gratis: I løbet af gratisperioden kan indsamlede data bruges til at forbedre modellen.
-- OpenAI API'er: Anmodninger opbevares i 30 dage i overensstemmelse med [OpenAIs datapolitikker](https://platform.openai.com/docs/guides/your-data).
-- Anthropic API'er: Anmodninger opbevares i 30 dage i overensstemmelse med [Anthropics datapolitikker](https://docs.anthropic.com/en/docs/claude-code/data-usage).
+- Big Pickle: I den gratis periode kan indsamlede data blive brugt til at forbedre modellen.
+- MiniMax M2.5 Free: I den gratis periode kan indsamlede data blive brugt til at forbedre modellen.
+- MiMo V2 Pro Free: I den gratis periode kan indsamlede data blive brugt til at forbedre modellen.
+- MiMo V2 Omni Free: I den gratis periode kan indsamlede data blive brugt til at forbedre modellen.
+- Qwen3.6 Plus Free: I den gratis periode kan indsamlede data blive brugt til at forbedre modellen.
+- Nemotron 3 Super Free: I den gratis periode kan indsamlede data blive brugt til at forbedre modellen.
+- OpenAI APIs: Anmodninger opbevares i 30 dage i overensstemmelse med [OpenAI's Data Policies](https://platform.openai.com/docs/guides/your-data).
+- Anthropic APIs: Anmodninger opbevares i 30 dage i overensstemmelse med [Anthropic's Data Policies](https://docs.anthropic.com/en/docs/claude-code/data-usage).
 
 ---
 
 ## For teams
 
-Zen fungerer også fremragende for teams. Du kan invitere holdkammerater, tildele roller, kuratere
-de modeller dit hold bruger, og mere.
+Zen fungerer også rigtig godt for teams. Du kan invitere teammedlemmer, tildele roller, kuratere
+de modeller dit team bruger, og mere.
 
 :::note
-Arbejdsområder er i øjeblikket gratis for teams som en del af betaversionen.
+Arbejdsområder er i øjeblikket gratis for teams som en del af betaen.
 :::
 
-At administrere dit arbejdsområde er i øjeblikket gratis for teams som en del af betaversionen. Vi vil
-dele mere information om priserne snart.
+Det er i øjeblikket gratis for teams at administrere deres arbejdsområde som en del af betaen. Vi deler
+flere detaljer om priserne snart.
 
 ---
 
 ### Roller
 
-Du kan invitere holdkammerater til dit arbejdsområde og tildele roller:
+Du kan invitere teammedlemmer til dit arbejdsområde og tildele roller:
 
 - **Admin**: Administrer modeller, medlemmer, API-nøgler og fakturering
 - **Medlem**: Administrer kun sine egne API-nøgler
@@ -245,29 +254,29 @@ Administratorer kan også sætte månedlige forbrugsgrænser for hvert medlem fo
 
 ### Modeladgang
 
-Administratorer kan aktivere eller deaktivere specifikke modeller for arbejdsområdet. Anmodninger til en deaktiveret model vil returnere en fejl.
+Administratorer kan aktivere eller deaktivere bestemte modeller for arbejdsområdet. Anmodninger til en deaktiveret model returnerer en fejl.
 
-Dette er nyttigt for tilfælde, hvor du ønsker at deaktivere brugen af en model, som
-samler ind data.
+Dette er nyttigt i tilfælde, hvor du vil deaktivere brugen af en model, der
+indsamler data.
 
 ---
 
-### Medbring din egen nøgle
+### Brug din egen nøgle
 
 Du kan bruge dine egne OpenAI- eller Anthropic API-nøgler, mens du stadig har adgang til andre modeller i Zen.
 
-Når du bruger dine egne nøgler, faktureres tokens direkte af udbyderen, ikke af Zen.
+Når du bruger dine egne nøgler, bliver tokens faktureret direkte af udbyderen, ikke af Zen.
 
-For eksempel kan din organisation allerede have en nøgle for OpenAI eller Anthropic
-og du vil bruge det i stedet for det Zen giver.
+For eksempel kan din organisation allerede have en nøgle til OpenAI eller Anthropic,
+og du vil bruge den i stedet for den, som Zen stiller til rådighed.
 
 ---
 
 ## Mål
 
-Vi oprettede OpenCode Zen for at:
+Vi skabte OpenCode Zen for at:
 
-1. **Benchmarke** de bedste modeller/udbydere for kodningsagenter.
-2. Have adgang til mulighederne for **højeste kvalitet** og ikke nedgradere ydeevnen eller rute til billigere udbydere.
-3. Give videre eventuelle **prisfald** ved at sælge til kostpris; så den eneste mark-up er for at dække vores behandlingsgebyrer.
-4. Have **ingen fastlåsning** ved at lade dig bruge den med en hvilken som helst anden kodeagent. Og lade dig altid bruge en hvilken som helst anden udbyder med OpenCode også.
+1. **Benchmarke** de bedste modeller/udbydere til kodeagenter.
+2. Give adgang til muligheder af **højeste kvalitet** uden at nedgradere ydeevnen eller route til billigere udbydere.
+3. Videreføre eventuelle **prisfald** ved at sælge til kostpris; så den eneste markup er for at dække vores behandlingsgebyrer.
+4. Have **ingen lock-in** ved at lade dig bruge det med enhver anden kodeagent. Og altid lade dig bruge enhver anden udbyder med OpenCode også.

packages/web/src/content/docs/de/zen.mdx

@@ -7,98 +7,96 @@ import config from "../../../../config.mjs"
 export const console = config.console
 export const email = `mailto:${config.email}`
 
-OpenCode Zen ist eine vom OpenCode-Team getestete und verifizierte Modellliste.
+OpenCode Zen ist eine Liste von getesteten und verifizierten Modellen, die vom OpenCode-Team bereitgestellt wird.
 
 :::note
-OpenCode Zen befindet sich aktuell in der Beta.
+OpenCode Zen befindet sich derzeit in der Beta.
 :::
 
-Zen funktioniert wie jeder andere Provider in OpenCode.
-Du meldest dich bei OpenCode Zen an, holst dir deinen API-Key und nutzt ihn optional.
+Zen funktioniert wie jeder andere Provider in OpenCode. Du meldest dich bei OpenCode Zen an und erhältst deinen API-Key. Es ist **vollständig optional** und du musst es nicht verwenden, um OpenCode zu nutzen.
 
 ---
 
 ## Hintergrund
 
-Es gibt sehr viele Modelle, aber nur ein Teil davon eignet sich wirklich gut als Coding-Agent.
-Ausserdem konfigurieren Provider Modelle sehr unterschiedlich, was die Qualitaet stark beeinflusst.
+Es gibt eine große Anzahl an Modellen, aber nur wenige davon eignen sich gut als Coding-Agenten. Außerdem sind die meisten Provider sehr unterschiedlich konfiguriert, sodass du sehr unterschiedliche Leistung und Qualität erhältst.
 
 :::tip
-Wir haben eine Auswahl aus Modellen und Providern getestet, die gut mit OpenCode funktionieren.
+Wir haben eine ausgewählte Gruppe von Modellen und Providern getestet, die gut mit OpenCode funktionieren.
 :::
 
-Wenn du Modelle ueber Gateways wie OpenRouter nutzt, ist oft unklar, ob du die beste Ausfuehrung eines Modells bekommst.
+Wenn du also ein Modell über etwas wie OpenRouter nutzt, kannst du nie sicher sein, ob du die beste Version des gewünschten Modells bekommst.
 
-Um das zu verbessern, haben wir:
+Um das zu beheben, haben wir ein paar Dinge getan:
 
-1. Eine Auswahl an Modellen getestet und mit den Teams ueber optimale Laufzeit-Setups gesprochen
-2. Mit Providern zusammengearbeitet, damit diese Modelle korrekt ausgeliefert werden
-3. Modell/Provider-Kombinationen gebenchmarkt und eine empfehlenswerte Liste erstellt
+1. Wir haben eine ausgewählte Gruppe von Modellen getestet und mit ihren Teams darüber gesprochen, wie man sie am besten betreibt.
+2. Anschließend haben wir mit einigen Providern zusammengearbeitet, um sicherzustellen, dass diese korrekt ausgeliefert werden.
+3. Schließlich haben wir die Kombination aus Modell und Provider benchmarked und eine Liste erstellt, die wir guten Gewissens empfehlen können.
 
-OpenCode Zen ist ein AI-Gateway, das dir Zugriff auf genau diese Modelle gibt.
+OpenCode Zen ist ein AI-Gateway, das dir Zugriff auf diese Modelle gibt.
 
 ---
 
-## Funktionsweise
+## So funktioniert es
 
 OpenCode Zen funktioniert wie jeder andere Provider in OpenCode.
 
-1. Melde dich bei **<a href={console}>OpenCode Zen</a>** an, hinterlege Zahlungsdaten und kopiere deinen API-Key.
-2. Fuehre in der TUI `/connect` aus, waehle OpenCode Zen und fuege den API-Key ein.
-3. Starte `/models` in der TUI, um empfohlene Modelle zu sehen.
+1. Du meldest dich bei **<a href={console}>OpenCode Zen</a>** an, hinterlegst deine Zahlungsdaten und kopierst deinen API-Key.
+2. Du führst den Befehl `/connect` in der TUI aus, wählst OpenCode Zen aus und fügst deinen API-Key ein.
+3. Führe `/models` in der TUI aus, um die Liste der Modelle zu sehen, die wir empfehlen.
 
-Abgerechnet wird pro Anfrage, Guthaben kannst du jederzeit aufladen.
+Dir wird pro Anfrage berechnet, und du kannst deinem Konto Guthaben hinzufügen.
 
 ---
 
 ## Endpunkte
 
-Du kannst unsere Modelle auch ueber die folgenden API-Endpunkte aufrufen.
+Du kannst auch über die folgenden API-Endpunkte auf unsere Modelle zugreifen.
 
-| Model              | Model ID           | Endpoint                                           | AI SDK Package              |
-| ------------------ | ------------------ | -------------------------------------------------- | --------------------------- |
-| GPT 5.4            | gpt-5.4            | `https://opencode.ai/zen/v1/responses`             | `@ai-sdk/openai`            |
-| GPT 5.3 Codex      | gpt-5.3-codex      | `https://opencode.ai/zen/v1/responses`             | `@ai-sdk/openai`            |
-| GPT 5.2            | gpt-5.2            | `https://opencode.ai/zen/v1/responses`             | `@ai-sdk/openai`            |
-| GPT 5.2 Codex      | gpt-5.2-codex      | `https://opencode.ai/zen/v1/responses`             | `@ai-sdk/openai`            |
-| GPT 5.1            | gpt-5.1            | `https://opencode.ai/zen/v1/responses`             | `@ai-sdk/openai`            |
-| GPT 5.1 Codex      | gpt-5.1-codex      | `https://opencode.ai/zen/v1/responses`             | `@ai-sdk/openai`            |
-| GPT 5.1 Codex Max  | gpt-5.1-codex-max  | `https://opencode.ai/zen/v1/responses`             | `@ai-sdk/openai`            |
-| GPT 5.1 Codex Mini | gpt-5.1-codex-mini | `https://opencode.ai/zen/v1/responses`             | `@ai-sdk/openai`            |
-| GPT 5              | gpt-5              | `https://opencode.ai/zen/v1/responses`             | `@ai-sdk/openai`            |
-| GPT 5 Codex        | gpt-5-codex        | `https://opencode.ai/zen/v1/responses`             | `@ai-sdk/openai`            |
-| GPT 5 Nano         | gpt-5-nano         | `https://opencode.ai/zen/v1/responses`             | `@ai-sdk/openai`            |
-| Claude Opus 4.6    | claude-opus-4-6    | `https://opencode.ai/zen/v1/messages`              | `@ai-sdk/anthropic`         |
-| Claude Opus 4.5    | claude-opus-4-5    | `https://opencode.ai/zen/v1/messages`              | `@ai-sdk/anthropic`         |
-| Claude Opus 4.1    | claude-opus-4-1    | `https://opencode.ai/zen/v1/messages`              | `@ai-sdk/anthropic`         |
-| Claude Sonnet 4.6  | claude-sonnet-4-6  | `https://opencode.ai/zen/v1/messages`              | `@ai-sdk/anthropic`         |
-| Claude Sonnet 4.5  | claude-sonnet-4-5  | `https://opencode.ai/zen/v1/messages`              | `@ai-sdk/anthropic`         |
-| Claude Sonnet 4    | claude-sonnet-4    | `https://opencode.ai/zen/v1/messages`              | `@ai-sdk/anthropic`         |
-| Claude Haiku 4.5   | claude-haiku-4-5   | `https://opencode.ai/zen/v1/messages`              | `@ai-sdk/anthropic`         |
-| Claude Haiku 3.5   | claude-3-5-haiku   | `https://opencode.ai/zen/v1/messages`              | `@ai-sdk/anthropic`         |
-| Gemini 3.1 Pro     | gemini-3.1-pro     | `https://opencode.ai/zen/v1/models/gemini-3.1-pro` | `@ai-sdk/google`            |
-| Gemini 3 Pro       | gemini-3-pro       | `https://opencode.ai/zen/v1/models/gemini-3-pro`   | `@ai-sdk/google`            |
-| Gemini 3 Flash     | gemini-3-flash     | `https://opencode.ai/zen/v1/models/gemini-3-flash` | `@ai-sdk/google`            |
-| MiniMax M2.5       | minimax-m2.5       | `https://opencode.ai/zen/v1/chat/completions`      | `@ai-sdk/openai-compatible` |
-| MiniMax M2.5 Free  | minimax-m2.5-free  | `https://opencode.ai/zen/v1/chat/completions`      | `@ai-sdk/openai-compatible` |
-| MiniMax M2.1       | minimax-m2.1       | `https://opencode.ai/zen/v1/chat/completions`      | `@ai-sdk/openai-compatible` |
-| GLM 5              | glm-5              | `https://opencode.ai/zen/v1/chat/completions`      | `@ai-sdk/openai-compatible` |
-| GLM 4.7            | glm-4.7            | `https://opencode.ai/zen/v1/chat/completions`      | `@ai-sdk/openai-compatible` |
-| GLM 4.6            | glm-4.6            | `https://opencode.ai/zen/v1/chat/completions`      | `@ai-sdk/openai-compatible` |
-| Kimi K2.5          | kimi-k2.5          | `https://opencode.ai/zen/v1/chat/completions`      | `@ai-sdk/openai-compatible` |
-| Kimi K2 Thinking   | kimi-k2-thinking   | `https://opencode.ai/zen/v1/chat/completions`      | `@ai-sdk/openai-compatible` |
-| Kimi K2            | kimi-k2            | `https://opencode.ai/zen/v1/chat/completions`      | `@ai-sdk/openai-compatible` |
-| Qwen3 Coder 480B   | qwen3-coder        | `https://opencode.ai/zen/v1/chat/completions`      | `@ai-sdk/openai-compatible` |
-| Big Pickle         | big-pickle         | `https://opencode.ai/zen/v1/chat/completions`      | `@ai-sdk/openai-compatible` |
+| Model                 | Model ID              | Endpoint                                           | AI SDK Package              |
+| --------------------- | --------------------- | -------------------------------------------------- | --------------------------- |
+| GPT 5.4               | gpt-5.4               | `https://opencode.ai/zen/v1/responses`             | `@ai-sdk/openai`            |
+| GPT 5.4 Pro           | gpt-5.4-pro           | `https://opencode.ai/zen/v1/responses`             | `@ai-sdk/openai`            |
+| GPT 5.4 Mini          | gpt-5.4-mini          | `https://opencode.ai/zen/v1/responses`             | `@ai-sdk/openai`            |
+| GPT 5.4 Nano          | gpt-5.4-nano          | `https://opencode.ai/zen/v1/responses`             | `@ai-sdk/openai`            |
+| GPT 5.3 Codex         | gpt-5.3-codex         | `https://opencode.ai/zen/v1/responses`             | `@ai-sdk/openai`            |
+| GPT 5.3 Codex Spark   | gpt-5.3-codex-spark   | `https://opencode.ai/zen/v1/responses`             | `@ai-sdk/openai`            |
+| GPT 5.2               | gpt-5.2               | `https://opencode.ai/zen/v1/responses`             | `@ai-sdk/openai`            |
+| GPT 5.2 Codex         | gpt-5.2-codex         | `https://opencode.ai/zen/v1/responses`             | `@ai-sdk/openai`            |
+| GPT 5.1               | gpt-5.1               | `https://opencode.ai/zen/v1/responses`             | `@ai-sdk/openai`            |
+| GPT 5.1 Codex         | gpt-5.1-codex         | `https://opencode.ai/zen/v1/responses`             | `@ai-sdk/openai`            |
+| GPT 5.1 Codex Max     | gpt-5.1-codex-max     | `https://opencode.ai/zen/v1/responses`             | `@ai-sdk/openai`            |
+| GPT 5.1 Codex Mini    | gpt-5.1-codex-mini    | `https://opencode.ai/zen/v1/responses`             | `@ai-sdk/openai`            |
+| GPT 5                 | gpt-5                 | `https://opencode.ai/zen/v1/responses`             | `@ai-sdk/openai`            |
+| GPT 5 Codex           | gpt-5-codex           | `https://opencode.ai/zen/v1/responses`             | `@ai-sdk/openai`            |
+| GPT 5 Nano            | gpt-5-nano            | `https://opencode.ai/zen/v1/responses`             | `@ai-sdk/openai`            |
+| Claude Opus 4.6       | claude-opus-4-6       | `https://opencode.ai/zen/v1/messages`              | `@ai-sdk/anthropic`         |
+| Claude Opus 4.5       | claude-opus-4-5       | `https://opencode.ai/zen/v1/messages`              | `@ai-sdk/anthropic`         |
+| Claude Opus 4.1       | claude-opus-4-1       | `https://opencode.ai/zen/v1/messages`              | `@ai-sdk/anthropic`         |
+| Claude Sonnet 4.6     | claude-sonnet-4-6     | `https://opencode.ai/zen/v1/messages`              | `@ai-sdk/anthropic`         |
+| Claude Sonnet 4.5     | claude-sonnet-4-5     | `https://opencode.ai/zen/v1/messages`              | `@ai-sdk/anthropic`         |
+| Claude Sonnet 4       | claude-sonnet-4       | `https://opencode.ai/zen/v1/messages`              | `@ai-sdk/anthropic`         |
+| Claude Haiku 4.5      | claude-haiku-4-5      | `https://opencode.ai/zen/v1/messages`              | `@ai-sdk/anthropic`         |
+| Claude Haiku 3.5      | claude-3-5-haiku      | `https://opencode.ai/zen/v1/messages`              | `@ai-sdk/anthropic`         |
+| Gemini 3.1 Pro        | gemini-3.1-pro        | `https://opencode.ai/zen/v1/models/gemini-3.1-pro` | `@ai-sdk/google`            |
+| Gemini 3 Flash        | gemini-3-flash        | `https://opencode.ai/zen/v1/models/gemini-3-flash` | `@ai-sdk/google`            |
+| MiniMax M2.5          | minimax-m2.5          | `https://opencode.ai/zen/v1/chat/completions`      | `@ai-sdk/openai-compatible` |
+| MiniMax M2.5 Free     | minimax-m2.5-free     | `https://opencode.ai/zen/v1/chat/completions`      | `@ai-sdk/openai-compatible` |
+| GLM 5                 | glm-5                 | `https://opencode.ai/zen/v1/chat/completions`      | `@ai-sdk/openai-compatible` |
+| Kimi K2.5             | kimi-k2.5             | `https://opencode.ai/zen/v1/chat/completions`      | `@ai-sdk/openai-compatible` |
+| Big Pickle            | big-pickle            | `https://opencode.ai/zen/v1/chat/completions`      | `@ai-sdk/openai-compatible` |
+| MiMo V2 Pro Free      | mimo-v2-pro-free      | `https://opencode.ai/zen/v1/chat/completions`      | `@ai-sdk/openai-compatible` |
+| MiMo V2 Omni Free     | mimo-v2-omni-free     | `https://opencode.ai/zen/v1/chat/completions`      | `@ai-sdk/openai-compatible` |
+| Qwen3.6 Plus Free     | qwen3.6-plus-free     | `https://opencode.ai/zen/v1/chat/completions`      | `@ai-sdk/openai-compatible` |
+| Nemotron 3 Super Free | nemotron-3-super-free | `https://opencode.ai/zen/v1/chat/completions`      | `@ai-sdk/openai-compatible` |
 
-Die [Model-ID](/docs/config/#models) in deiner OpenCode-Konfiguration hat das Format `opencode/<model-id>`.
-Fuer GPT 5.2 Codex verwendest du zum Beispiel `opencode/gpt-5.2-codex`.
+Die [Model-ID](/docs/config/#models) in deiner OpenCode-Konfiguration verwendet das Format `opencode/<model-id>`. Für GPT 5.3 Codex würdest du zum Beispiel `opencode/gpt-5.3-codex` in deiner Konfiguration verwenden.
 
 ---
 
 ### Models
 
-Die komplette Liste verfuegbarer Modelle inklusive Metadaten findest du unter:
+Du kannst die vollständige Liste der verfügbaren Modelle und ihrer Metadaten hier abrufen:
 
 ```
 https://opencode.ai/zen/v1/models
@@ -106,160 +104,161 @@ https://opencode.ai/zen/v1/models
 
 ---
 
-## Preisgestaltung
+## Preise
 
-Wir nutzen ein Pay-as-you-go-Modell.
-Unten siehst du die Preise **pro 1 Mio. Tokens**.
+Wir unterstützen ein Pay-as-you-go-Modell. Unten findest du die Preise **pro 1M Tokens**.
 
-| Model                             | Input  | Output | Cached Read | Cached Write |
-| --------------------------------- | ------ | ------ | ----------- | ------------ |
-| Big Pickle                        | Free   | Free   | Free        | -            |
-| MiniMax M2.5 Free                 | Free   | Free   | Free        | -            |
-| MiniMax M2.5                      | $0.30  | $1.20  | $0.06       | $0.375       |
-| MiniMax M2.1                      | $0.30  | $1.20  | $0.10       | -            |
-| GLM 5                             | $1.00  | $3.20  | $0.20       | -            |
-| GLM 4.7                           | $0.60  | $2.20  | $0.10       | -            |
-| GLM 4.6                           | $0.60  | $2.20  | $0.10       | -            |
-| Kimi K2.5                         | $0.60  | $3.00  | $0.10       | -            |
-| Kimi K2 Thinking                  | $0.40  | $2.50  | -           | -            |
-| Kimi K2                           | $0.40  | $2.50  | -           | -            |
-| Qwen3 Coder 480B                  | $0.45  | $1.50  | -           | -            |
-| Claude Opus 4.6 (≤ 200K tokens)   | $5.00  | $25.00 | $0.50       | $6.25        |
-| Claude Opus 4.6 (> 200K tokens)   | $10.00 | $37.50 | $1.00       | $12.50       |
-| Claude Opus 4.5                   | $5.00  | $25.00 | $0.50       | $6.25        |
-| Claude Opus 4.1                   | $15.00 | $75.00 | $1.50       | $18.75       |
-| Claude Sonnet 4.6 (≤ 200K tokens) | $3.00  | $15.00 | $0.30       | $3.75        |
-| Claude Sonnet 4.6 (> 200K tokens) | $6.00  | $22.50 | $0.60       | $7.50        |
-| Claude Sonnet 4.5 (≤ 200K tokens) | $3.00  | $15.00 | $0.30       | $3.75        |
-| Claude Sonnet 4.5 (> 200K tokens) | $6.00  | $22.50 | $0.60       | $7.50        |
-| Claude Sonnet 4 (≤ 200K tokens)   | $3.00  | $15.00 | $0.30       | $3.75        |
-| Claude Sonnet 4 (> 200K tokens)   | $6.00  | $22.50 | $0.60       | $7.50        |
-| Claude Haiku 4.5                  | $1.00  | $5.00  | $0.10       | $1.25        |
-| Claude Haiku 3.5                  | $0.80  | $4.00  | $0.08       | $1.00        |
-| Gemini 3.1 Pro (≤ 200K tokens)    | $2.00  | $12.00 | $0.20       | -            |
-| Gemini 3.1 Pro (> 200K tokens)    | $4.00  | $18.00 | $0.40       | -            |
-| Gemini 3 Pro (≤ 200K tokens)      | $2.00  | $12.00 | $0.20       | -            |
-| Gemini 3 Pro (> 200K tokens)      | $4.00  | $18.00 | $0.40       | -            |
-| Gemini 3 Flash                    | $0.50  | $3.00  | $0.05       | -            |
-| GPT 5.4                           | $2.50  | $15.00 | $0.25       | -            |
-| GPT 5.3 Codex                     | $1.75  | $14.00 | $0.175      | -            |
-| GPT 5.2                           | $1.75  | $14.00 | $0.175      | -            |
-| GPT 5.2 Codex                     | $1.75  | $14.00 | $0.175      | -            |
-| GPT 5.1                           | $1.07  | $8.50  | $0.107      | -            |
-| GPT 5.1 Codex                     | $1.07  | $8.50  | $0.107      | -            |
-| GPT 5.1 Codex Max                 | $1.25  | $10.00 | $0.125      | -            |
-| GPT 5.1 Codex Mini                | $0.25  | $2.00  | $0.025      | -            |
-| GPT 5                             | $1.07  | $8.50  | $0.107      | -            |
-| GPT 5 Codex                       | $1.07  | $8.50  | $0.107      | -            |
-| GPT 5 Nano                        | Free   | Free   | Free        | -            |
+| Model                             | Input  | Output  | Cached Read | Cached Write |
+| --------------------------------- | ------ | ------- | ----------- | ------------ |
+| Big Pickle                        | Free   | Free    | Free        | -            |
+| MiMo V2 Pro Free                  | Free   | Free    | Free        | -            |
+| MiMo V2 Omni Free                 | Free   | Free    | Free        | -            |
+| Qwen3.6 Plus Free                 | Free   | Free    | Free        | -            |
+| Nemotron 3 Super Free             | Free   | Free    | Free        | -            |
+| MiniMax M2.5 Free                 | Free   | Free    | Free        | -            |
+| MiniMax M2.5                      | $0.30  | $1.20   | $0.06       | $0.375       |
+| GLM 5                             | $1.00  | $3.20   | $0.20       | -            |
+| Kimi K2.5                         | $0.60  | $3.00   | $0.10       | -            |
+| Qwen3 Coder 480B                  | $0.45  | $1.50   | -           | -            |
+| Claude Opus 4.6                   | $5.00  | $25.00  | $0.50       | $6.25        |
+| Claude Opus 4.5                   | $5.00  | $25.00  | $0.50       | $6.25        |
+| Claude Opus 4.1                   | $15.00 | $75.00  | $1.50       | $18.75       |
+| Claude Sonnet 4.6                 | $3.00  | $15.00  | $0.30       | $3.75        |
+| Claude Sonnet 4.5 (≤ 200K tokens) | $3.00  | $15.00  | $0.30       | $3.75        |
+| Claude Sonnet 4.5 (> 200K tokens) | $6.00  | $22.50  | $0.60       | $7.50        |
+| Claude Sonnet 4 (≤ 200K tokens)   | $3.00  | $15.00  | $0.30       | $3.75        |
+| Claude Sonnet 4 (> 200K tokens)   | $6.00  | $22.50  | $0.60       | $7.50        |
+| Claude Haiku 4.5                  | $1.00  | $5.00   | $0.10       | $1.25        |
+| Claude Haiku 3.5                  | $0.80  | $4.00   | $0.08       | $1.00        |
+| Gemini 3.1 Pro (≤ 200K tokens)    | $2.00  | $12.00  | $0.20       | -            |
+| Gemini 3.1 Pro (> 200K tokens)    | $4.00  | $18.00  | $0.40       | -            |
+| Gemini 3 Flash                    | $0.50  | $3.00   | $0.05       | -            |
+| GPT 5.4                           | $2.50  | $15.00  | $0.25       | -            |
+| GPT 5.4 Pro                       | $30.00 | $180.00 | $30.00      | -            |
+| GPT 5.4 Mini                      | $0.75  | $4.50   | $0.075      | -            |
+| GPT 5.4 Nano                      | $0.20  | $1.25   | $0.02       | -            |
+| GPT 5.3 Codex Spark               | $1.75  | $14.00  | $0.175      | -            |
+| GPT 5.3 Codex                     | $1.75  | $14.00  | $0.175      | -            |
+| GPT 5.2                           | $1.75  | $14.00  | $0.175      | -            |
+| GPT 5.2 Codex                     | $1.75  | $14.00  | $0.175      | -            |
+| GPT 5.1                           | $1.07  | $8.50   | $0.107      | -            |
+| GPT 5.1 Codex                     | $1.07  | $8.50   | $0.107      | -            |
+| GPT 5.1 Codex Max                 | $1.25  | $10.00  | $0.125      | -            |
+| GPT 5.1 Codex Mini                | $0.25  | $2.00   | $0.025      | -            |
+| GPT 5                             | $1.07  | $8.50   | $0.107      | -            |
+| GPT 5 Codex                       | $1.07  | $8.50   | $0.107      | -            |
+| GPT 5 Nano                        | Free   | Free    | Free        | -            |
 
-In deinem Verlauf siehst du eventuell _Claude Haiku 3.5_.
-Das ist ein [guenstiges Modell](/docs/config/#models), das fuer Session-Titel verwendet wird.
+Möglicherweise siehst du _Claude Haiku 3.5_ in deinem Nutzungsverlauf. Das ist ein [kostengünstiges Modell](/docs/config/#models), das verwendet wird, um die Titel deiner Sessions zu generieren.
 
 :::note
-Kreditkartengebuehren geben wir zum Selbstkostenpreis weiter (4,4 % + $0.30 pro Transaktion), ohne Aufschlag.
+Kreditkartengebühren werden zum Selbstkostenpreis weitergegeben (4.4% + $0.30 pro Transaktion); wir berechnen darüber hinaus nichts.
 :::
 
 Die kostenlosen Modelle:
 
-- MiniMax M2.5 Free ist fuer begrenzte Zeit auf OpenCode verfuegbar. Das Team nutzt diese Zeit, um Feedback zu sammeln und das Modell zu verbessern.
-- Big Pickle ist ein Stealth-Modell, das fuer begrenzte Zeit kostenlos auf OpenCode verfuegbar ist. Das Team nutzt diese Zeit, um Feedback zu sammeln und das Modell zu verbessern.
+- MiniMax M2.5 Free ist für begrenzte Zeit auf OpenCode verfügbar. Das Team nutzt diese Zeit, um Feedback zu sammeln und das Modell zu verbessern.
+- MiMo V2 Pro Free ist für begrenzte Zeit auf OpenCode verfügbar. Das Team nutzt diese Zeit, um Feedback zu sammeln und das Modell zu verbessern.
+- MiMo V2 Omni Free ist für begrenzte Zeit auf OpenCode verfügbar. Das Team nutzt diese Zeit, um Feedback zu sammeln und das Modell zu verbessern.
+- Qwen3.6 Plus Free ist für begrenzte Zeit auf OpenCode verfügbar. Das Team nutzt diese Zeit, um Feedback zu sammeln und das Modell zu verbessern.
+- Nemotron 3 Super Free ist für begrenzte Zeit auf OpenCode verfügbar. Das Team nutzt diese Zeit, um Feedback zu sammeln und das Modell zu verbessern.
+- Big Pickle ist ein Stealth-Modell, das für begrenzte Zeit kostenlos auf OpenCode verfügbar ist. Das Team nutzt diese Zeit, um Feedback zu sammeln und das Modell zu verbessern.
 
-Wenn du Fragen hast, <a href={email}>kontaktiere uns</a>.
+<a href={email}>Kontaktiere uns</a>, wenn du Fragen hast.
 
 ---
 
 ### Auto-reload
 
-Wenn dein Guthaben unter $5 faellt, laedt Zen automatisch $20 nach.
+Wenn dein Guthaben unter $5 fällt, lädt Zen automatisch $20 nach.
 
-Du kannst den Betrag anpassen oder Auto-Reload komplett deaktivieren.
+Du kannst den Auto-reload-Betrag ändern. Du kannst Auto-reload auch vollständig deaktivieren.
 
 ---
 
-### Monatslimits
+### Monatliche Limits
 
-Du kannst monatliche Limits fuer den gesamten Workspace und pro Teammitglied festlegen.
+Du kannst auch ein monatliches Nutzungslimit für den gesamten Workspace und für jedes Mitglied deines Teams festlegen.
 
-Wenn du z. B. ein Monatslimit von $20 setzt, verbraucht Zen nicht mehr als $20 pro Monat.
-Mit aktiviertem Auto-Reload kann die Abrechnung dennoch darueber liegen, falls das Guthaben unter $5 sinkt.
+Angenommen, du setzt ein monatliches Nutzungslimit von $20, dann wird Zen in einem Monat nicht mehr als $20 verwenden. Wenn du jedoch Auto-reload aktiviert hast, kann es sein, dass Zen dir mehr als $20 berechnet, wenn dein Guthaben unter $5 fällt.
 
 ---
 
 ### Veraltete Modelle
 
-| Model            | Datum der Abschaltung |
-| ---------------- | --------------------- |
-| Qwen3 Coder 480B | 6. Feb. 2026          |
-| Kimi K2 Thinking | 6. Maerz 2026         |
-| Kimi K2          | 6. Maerz 2026         |
-| MiniMax M2.1     | 15. Maerz 2026        |
-| GLM 4.7          | 15. Maerz 2026        |
-| GLM 4.6          | 15. Maerz 2026        |
+| Model            | Deprecation date |
+| ---------------- | ---------------- |
+| MiniMax M2.1     | March 15, 2026   |
+| GLM 4.7          | March 15, 2026   |
+| GLM 4.6          | March 15, 2026   |
+| Gemini 3 Pro     | March 9, 2026    |
+| Kimi K2 Thinking | March 6, 2026    |
+| Kimi K2          | March 6, 2026    |
+| Qwen3 Coder 480B | Feb 6, 2026      |
 
 ---
 
 ## Datenschutz
 
-Alle Modelle werden in den USA gehostet.
-Unsere Provider arbeiten grundsaetzlich mit Zero-Retention und nutzen deine Daten nicht zum Training, mit folgenden Ausnahmen:
+Alle unsere Modelle werden in den USA gehostet. Unsere Provider folgen einer Zero-Retention-Richtlinie und verwenden deine Daten nicht zum Trainieren von Modellen, mit den folgenden Ausnahmen:
 
-- Big Pickle: Waehrend der kostenlosen Phase koennen gesammelte Daten zur Verbesserung des Modells genutzt werden.
-- MiniMax M2.5 Free: Waehrend der kostenlosen Phase koennen gesammelte Daten zur Verbesserung des Modells genutzt werden.
-- OpenAI APIs: Anfragen werden fuer 30 Tage gemaess [OpenAI's Data Policies](https://platform.openai.com/docs/guides/your-data) aufbewahrt.
-- Anthropic APIs: Anfragen werden fuer 30 Tage gemaess [Anthropic's Data Policies](https://docs.anthropic.com/en/docs/claude-code/data-usage) aufbewahrt.
+- Big Pickle: Während des kostenlosen Zeitraums können gesammelte Daten zur Verbesserung des Modells verwendet werden.
+- MiniMax M2.5 Free: Während des kostenlosen Zeitraums können gesammelte Daten zur Verbesserung des Modells verwendet werden.
+- MiMo V2 Pro Free: Während des kostenlosen Zeitraums können gesammelte Daten zur Verbesserung des Modells verwendet werden.
+- MiMo V2 Omni Free: Während des kostenlosen Zeitraums können gesammelte Daten zur Verbesserung des Modells verwendet werden.
+- Qwen3.6 Plus Free: Während des kostenlosen Zeitraums können gesammelte Daten zur Verbesserung des Modells verwendet werden.
+- Nemotron 3 Super Free: Während des kostenlosen Zeitraums können gesammelte Daten zur Verbesserung des Modells verwendet werden.
+- OpenAI APIs: Anfragen werden in Übereinstimmung mit [OpenAI's Data Policies](https://platform.openai.com/docs/guides/your-data) 30 Tage lang gespeichert.
+- Anthropic APIs: Anfragen werden in Übereinstimmung mit [Anthropic's Data Policies](https://docs.anthropic.com/en/docs/claude-code/data-usage) 30 Tage lang gespeichert.
 
 ---
 
 ## Für Teams
 
-Zen eignet sich auch gut fuer Teams.
-Du kannst Mitglieder einladen, Rollen vergeben und den Modellzugriff fuer dein Team steuern.
+Zen funktioniert auch hervorragend für Teams. Du kannst Teammitglieder einladen, Rollen zuweisen, die Modelle kuratieren, die dein Team verwendet, und mehr.
 
 :::note
-Workspaces sind fuer Teams waehrend der Beta derzeit kostenlos.
+Workspaces sind für Teams derzeit als Teil der Beta kostenlos.
 :::
 
-Workspace-Verwaltung ist in der Beta kostenlos.
-Details zur spaeteren Preisgestaltung folgen.
+Die Verwaltung deines Workspace ist für Teams derzeit als Teil der Beta kostenlos. Wir werden bald mehr Details zur Preisgestaltung teilen.
 
 ---
 
 ### Rollen
 
-Du kannst Teammitglieder einladen und Rollen vergeben:
+Du kannst Teammitglieder in deinen Workspace einladen und Rollen zuweisen:
 
-- **Admin**: Verwalten Modelle, Mitglieder, API-Keys und Abrechnung
-- **Member**: Verwalten nur eigene API-Keys
+- **Admin**: Verwaltet Modelle, Mitglieder, API-Keys und Abrechnung
+- **Member**: Verwaltet nur die eigenen API-Keys
 
-Admins koennen zusaetzlich monatliche Ausgabenlimits pro Mitglied setzen.
+Admins können außerdem monatliche Ausgabenlimits für jedes Mitglied festlegen, um die Kosten unter Kontrolle zu halten.
 
 ---
 
 ### Modellzugriff
 
-Admins koennen einzelne Modelle fuer den Workspace aktivieren oder deaktivieren.
-Anfragen an deaktivierte Modelle liefern einen Fehler.
+Admins können bestimmte Modelle für den Workspace aktivieren oder deaktivieren. Anfragen an ein deaktiviertes Modell geben einen Fehler zurück.
 
-Das ist hilfreich, wenn bestimmte datenverarbeitende Modelle ausgeschlossen werden sollen.
+Das ist nützlich, wenn du die Verwendung eines Modells deaktivieren möchtest, das Daten sammelt.
 
 ---
 
 ### Eigenen Key mitbringen
 
-Du kannst eigene OpenAI- oder Anthropic-API-Keys verwenden und trotzdem andere Zen-Modelle nutzen.
+Du kannst deine eigenen OpenAI- oder Anthropic-API-Keys verwenden und trotzdem auf andere Modelle in Zen zugreifen.
 
-Bei eigenen Keys erfolgt die Token-Abrechnung direkt ueber den Provider, nicht ueber Zen.
+Wenn du deine eigenen Keys verwendest, werden Tokens direkt vom Provider abgerechnet, nicht von Zen.
 
-Zum Beispiel hat deine Organisation vielleicht bereits einen Key fuer OpenAI oder Anthropic und du moechtest diesen anstelle des von Zen bereitgestellten nutzen.
+Zum Beispiel hat deine Organisation möglicherweise bereits einen Key für OpenAI oder Anthropic und du möchtest diesen anstelle des von Zen bereitgestellten verwenden.
 
 ---
 
 ## Ziele
 
-Wir haben OpenCode Zen entwickelt, um:
+Wir haben OpenCode Zen geschaffen, um:
 
-1. Die besten Modell/Provider-Kombinationen fuer Coding-Agenten zu **benchmarken**
-2. Stets **hohe Qualitaet** ohne Downgrades oder versteckte Umleitungen auf billigere Provider zu liefern
-3. **Preissenkungen** zum Selbstkostenpreis weiterzugeben, mit Aufschlag nur fuer Zahlungsgebuehren
-4. **Kein Lock-in** zu erzwingen, damit du Zen mit anderen Coding-Agents und OpenCode weiter mit anderen Providern nutzen kannst
+1. die besten Modelle/Provider für Coding-Agenten zu **benchmarken**.
+2. Zugriff auf die **höchste Qualität** zu haben und die Leistung nicht zu verschlechtern oder an günstigere Provider weiterzuleiten.
+3. **Preissenkungen** weiterzugeben, indem wir zum Selbstkostenpreis verkaufen; der einzige Aufschlag dient also dazu, unsere Bearbeitungsgebühren abzudecken.
+4. **keinen Lock-in** zu haben, indem du es mit jedem anderen Coding-Agenten verwenden kannst. Und dir außerdem immer zu erlauben, mit OpenCode auch jeden anderen Provider zu verwenden.

packages/web/src/content/docs/es/zen.mdx

@@ -7,39 +7,41 @@ import config from "../../../../config.mjs"
 export const console = config.console
 export const email = `mailto:${config.email}`
 
-OpenCode Zen es una lista de modelos probados y verificados proporcionada por el equipo OpenCode.
+OpenCode Zen es una lista de modelos probados y verificados proporcionada por el equipo de OpenCode.
 
 :::note
-OpenCode Zen se encuentra actualmente en versión beta.
+OpenCode Zen está actualmente en beta.
 :::
 
-Zen funciona como cualquier otro proveedor en OpenCode. Inicias sesion en OpenCode Zen y obtienes
-tu API key. Es **completamente opcional** y no necesitas usarlo para usar OpenCode.
+Zen funciona como cualquier otro proveedor en OpenCode. Inicias sesión en OpenCode Zen y obtienes
+tu API key. Es **completamente opcional** y no necesitas usarlo para usar
+OpenCode.
 
 ---
 
 ## Contexto
 
-Hay una gran cantidad de modelos, pero solo unos pocos funcionan bien como agentes de coding.
-Ademas, la mayoria de proveedores se configura de forma diferente, asi que el rendimiento y la calidad varian mucho.
+Existe una gran cantidad de modelos, pero solo unos pocos de
+estos modelos funcionan bien como agentes de coding. Además, la mayoría de los proveedores están
+configurados de forma muy diferente; por eso obtienes un rendimiento y una calidad muy distintos.
 
 :::tip
 Probamos un grupo selecto de modelos y proveedores que funcionan bien con OpenCode.
 :::
 
-Si usas un modelo a traves de algo como OpenRouter, nunca puedes estar
-seguro de que recibes la mejor version del modelo que quieres.
+Así que, si estás usando un modelo a través de algo como OpenRouter, nunca puedes estar
+seguro de si estás obteniendo la mejor versión del modelo que quieres.
 
-Para solucionar este problema, hicimos un par de cosas:
+Para solucionar esto, hicimos un par de cosas:
 
 1. Probamos un grupo selecto de modelos y hablamos con sus equipos sobre cómo
-   mejor ejecutarlos.
-2. Luego trabajamos con algunos proveedores para asegurarnos de que estuvieran siendo atendidos.
+   ejecutarlos de la mejor manera.
+2. Luego trabajamos con algunos proveedores para asegurarnos de que se estuvieran sirviendo
    correctamente.
-3. Finalmente, comparamos la combinación modelo/proveedor y llegamos
-   con una lista que nos sentimos bien recomendando.
+3. Por último, evaluamos la combinación de modelo/proveedor y elaboramos
+   una lista que creemos que vale la pena recomendar.
 
-OpenCode Zen es una puerta de enlace de IA que le brinda acceso a estos modelos.
+OpenCode Zen es un gateway de AI que te da acceso a estos modelos.
 
 ---
 
@@ -47,64 +49,65 @@ OpenCode Zen es una puerta de enlace de IA que le brinda acceso a estos modelos.
 
 OpenCode Zen funciona como cualquier otro proveedor en OpenCode.
 
-1. Inicias sesión en **<a href={console}>OpenCode Zen</a>**, agregas tu facturación
-   detalles y copie su clave API.
-2. Ejecuta el comando `/connect` en TUI, selecciona OpenCode Zen y pega tu clave API.
-3. Ejecute `/models` en TUI para ver la lista de modelos que recomendamos.
+1. Inicias sesión en **<a href={console}>OpenCode Zen</a>**, agregas tus datos de facturación
+   y copias tu API key.
+2. Ejecutas el comando `/connect` en la TUI, seleccionas OpenCode Zen y pegas tu API key.
+3. Ejecuta `/models` en la TUI para ver la lista de modelos que recomendamos.
 
-Se le cobra por solicitud y puede agregar créditos a su cuenta.
+Se te cobra por solicitud y puedes agregar créditos a tu cuenta.
 
 ---
 
-## Puntos finales
+## Endpoints
 
-También puede acceder a nuestros modelos a través de los siguientes puntos finales API.
+También puedes acceder a nuestros modelos a través de los siguientes endpoints de API.
 
-| Modelo             | Model ID           | Endpoint                                           | AI SDK package              |
-| ------------------ | ------------------ | -------------------------------------------------- | --------------------------- |
-| GPT 5.4            | gpt-5.4            | `https://opencode.ai/zen/v1/responses`             | `@ai-sdk/openai`            |
-| GPT 5.3 Codex      | gpt-5.3-codex      | `https://opencode.ai/zen/v1/responses`             | `@ai-sdk/openai`            |
-| GPT 5.2            | gpt-5.2            | `https://opencode.ai/zen/v1/responses`             | `@ai-sdk/openai`            |
-| GPT 5.2 Codex      | gpt-5.2-codex      | `https://opencode.ai/zen/v1/responses`             | `@ai-sdk/openai`            |
-| GPT 5.1            | gpt-5.1            | `https://opencode.ai/zen/v1/responses`             | `@ai-sdk/openai`            |
-| GPT 5.1 Codex      | gpt-5.1-codex      | `https://opencode.ai/zen/v1/responses`             | `@ai-sdk/openai`            |
-| GPT 5.1 Codex Max  | gpt-5.1-codex-max  | `https://opencode.ai/zen/v1/responses`             | `@ai-sdk/openai`            |
-| GPT 5.1 Codex Mini | gpt-5.1-codex-mini | `https://opencode.ai/zen/v1/responses`             | `@ai-sdk/openai`            |
-| GPT 5              | gpt-5              | `https://opencode.ai/zen/v1/responses`             | `@ai-sdk/openai`            |
-| GPT 5 Codex        | gpt-5-codex        | `https://opencode.ai/zen/v1/responses`             | `@ai-sdk/openai`            |
-| GPT 5 Nano         | gpt-5-nano         | `https://opencode.ai/zen/v1/responses`             | `@ai-sdk/openai`            |
-| Claude Opus 4.6    | claude-opus-4-6    | `https://opencode.ai/zen/v1/messages`              | `@ai-sdk/anthropic`         |
-| Claude Opus 4.5    | claude-opus-4-5    | `https://opencode.ai/zen/v1/messages`              | `@ai-sdk/anthropic`         |
-| Claude Opus 4.1    | claude-opus-4-1    | `https://opencode.ai/zen/v1/messages`              | `@ai-sdk/anthropic`         |
-| Claude Sonnet 4.6  | claude-sonnet-4-6  | `https://opencode.ai/zen/v1/messages`              | `@ai-sdk/anthropic`         |
-| Claude Sonnet 4.5  | claude-sonnet-4-5  | `https://opencode.ai/zen/v1/messages`              | `@ai-sdk/anthropic`         |
-| Claude Sonnet 4    | claude-sonnet-4    | `https://opencode.ai/zen/v1/messages`              | `@ai-sdk/anthropic`         |
-| Claude Haiku 4.5   | claude-haiku-4-5   | `https://opencode.ai/zen/v1/messages`              | `@ai-sdk/anthropic`         |
-| Claude Haiku 3.5   | claude-3-5-haiku   | `https://opencode.ai/zen/v1/messages`              | `@ai-sdk/anthropic`         |
-| Gemini 3.1 Pro     | gemini-3.1-pro     | `https://opencode.ai/zen/v1/models/gemini-3.1-pro` | `@ai-sdk/google`            |
-| Gemini 3 Pro       | gemini-3-pro       | `https://opencode.ai/zen/v1/models/gemini-3-pro`   | `@ai-sdk/google`            |
-| Gemini 3 Flash     | gemini-3-flash     | `https://opencode.ai/zen/v1/models/gemini-3-flash` | `@ai-sdk/google`            |
-| MiniMax M2.5       | minimax-m2.5       | `https://opencode.ai/zen/v1/chat/completions`      | `@ai-sdk/openai-compatible` |
-| MiniMax M2.5 Free  | minimax-m2.5-free  | `https://opencode.ai/zen/v1/chat/completions`      | `@ai-sdk/openai-compatible` |
-| MiniMax M2.1       | minimax-m2.1       | `https://opencode.ai/zen/v1/chat/completions`      | `@ai-sdk/openai-compatible` |
-| GLM 5              | glm-5              | `https://opencode.ai/zen/v1/chat/completions`      | `@ai-sdk/openai-compatible` |
-| GLM 4.7            | glm-4.7            | `https://opencode.ai/zen/v1/chat/completions`      | `@ai-sdk/openai-compatible` |
-| GLM 4.6            | glm-4.6            | `https://opencode.ai/zen/v1/chat/completions`      | `@ai-sdk/openai-compatible` |
-| Kimi K2.5          | kimi-k2.5          | `https://opencode.ai/zen/v1/chat/completions`      | `@ai-sdk/openai-compatible` |
-| Kimi K2 Thinking   | kimi-k2-thinking   | `https://opencode.ai/zen/v1/chat/completions`      | `@ai-sdk/openai-compatible` |
-| Kimi K2            | kimi-k2            | `https://opencode.ai/zen/v1/chat/completions`      | `@ai-sdk/openai-compatible` |
-| Qwen3 Coder 480B   | qwen3-coder        | `https://opencode.ai/zen/v1/chat/completions`      | `@ai-sdk/openai-compatible` |
-| Big Pickle         | big-pickle         | `https://opencode.ai/zen/v1/chat/completions`      | `@ai-sdk/openai-compatible` |
+| Modelo                | Model ID              | Endpoint                                           | AI SDK Package              |
+| --------------------- | --------------------- | -------------------------------------------------- | --------------------------- |
+| GPT 5.4               | gpt-5.4               | `https://opencode.ai/zen/v1/responses`             | `@ai-sdk/openai`            |
+| GPT 5.4 Pro           | gpt-5.4-pro           | `https://opencode.ai/zen/v1/responses`             | `@ai-sdk/openai`            |
+| GPT 5.4 Mini          | gpt-5.4-mini          | `https://opencode.ai/zen/v1/responses`             | `@ai-sdk/openai`            |
+| GPT 5.4 Nano          | gpt-5.4-nano          | `https://opencode.ai/zen/v1/responses`             | `@ai-sdk/openai`            |
+| GPT 5.3 Codex         | gpt-5.3-codex         | `https://opencode.ai/zen/v1/responses`             | `@ai-sdk/openai`            |
+| GPT 5.3 Codex Spark   | gpt-5.3-codex-spark   | `https://opencode.ai/zen/v1/responses`             | `@ai-sdk/openai`            |
+| GPT 5.2               | gpt-5.2               | `https://opencode.ai/zen/v1/responses`             | `@ai-sdk/openai`            |
+| GPT 5.2 Codex         | gpt-5.2-codex         | `https://opencode.ai/zen/v1/responses`             | `@ai-sdk/openai`            |
+| GPT 5.1               | gpt-5.1               | `https://opencode.ai/zen/v1/responses`             | `@ai-sdk/openai`            |
+| GPT 5.1 Codex         | gpt-5.1-codex         | `https://opencode.ai/zen/v1/responses`             | `@ai-sdk/openai`            |
+| GPT 5.1 Codex Max     | gpt-5.1-codex-max     | `https://opencode.ai/zen/v1/responses`             | `@ai-sdk/openai`            |
+| GPT 5.1 Codex Mini    | gpt-5.1-codex-mini    | `https://opencode.ai/zen/v1/responses`             | `@ai-sdk/openai`            |
+| GPT 5                 | gpt-5                 | `https://opencode.ai/zen/v1/responses`             | `@ai-sdk/openai`            |
+| GPT 5 Codex           | gpt-5-codex           | `https://opencode.ai/zen/v1/responses`             | `@ai-sdk/openai`            |
+| GPT 5 Nano            | gpt-5-nano            | `https://opencode.ai/zen/v1/responses`             | `@ai-sdk/openai`            |
+| Claude Opus 4.6       | claude-opus-4-6       | `https://opencode.ai/zen/v1/messages`              | `@ai-sdk/anthropic`         |
+| Claude Opus 4.5       | claude-opus-4-5       | `https://opencode.ai/zen/v1/messages`              | `@ai-sdk/anthropic`         |
+| Claude Opus 4.1       | claude-opus-4-1       | `https://opencode.ai/zen/v1/messages`              | `@ai-sdk/anthropic`         |
+| Claude Sonnet 4.6     | claude-sonnet-4-6     | `https://opencode.ai/zen/v1/messages`              | `@ai-sdk/anthropic`         |
+| Claude Sonnet 4.5     | claude-sonnet-4-5     | `https://opencode.ai/zen/v1/messages`              | `@ai-sdk/anthropic`         |
+| Claude Sonnet 4       | claude-sonnet-4       | `https://opencode.ai/zen/v1/messages`              | `@ai-sdk/anthropic`         |
+| Claude Haiku 4.5      | claude-haiku-4-5      | `https://opencode.ai/zen/v1/messages`              | `@ai-sdk/anthropic`         |
+| Claude Haiku 3.5      | claude-3-5-haiku      | `https://opencode.ai/zen/v1/messages`              | `@ai-sdk/anthropic`         |
+| Gemini 3.1 Pro        | gemini-3.1-pro        | `https://opencode.ai/zen/v1/models/gemini-3.1-pro` | `@ai-sdk/google`            |
+| Gemini 3 Flash        | gemini-3-flash        | `https://opencode.ai/zen/v1/models/gemini-3-flash` | `@ai-sdk/google`            |
+| MiniMax M2.5          | minimax-m2.5          | `https://opencode.ai/zen/v1/chat/completions`      | `@ai-sdk/openai-compatible` |
+| MiniMax M2.5 Free     | minimax-m2.5-free     | `https://opencode.ai/zen/v1/chat/completions`      | `@ai-sdk/openai-compatible` |
+| GLM 5                 | glm-5                 | `https://opencode.ai/zen/v1/chat/completions`      | `@ai-sdk/openai-compatible` |
+| Kimi K2.5             | kimi-k2.5             | `https://opencode.ai/zen/v1/chat/completions`      | `@ai-sdk/openai-compatible` |
+| Big Pickle            | big-pickle            | `https://opencode.ai/zen/v1/chat/completions`      | `@ai-sdk/openai-compatible` |
+| MiMo V2 Pro Free      | mimo-v2-pro-free      | `https://opencode.ai/zen/v1/chat/completions`      | `@ai-sdk/openai-compatible` |
+| MiMo V2 Omni Free     | mimo-v2-omni-free     | `https://opencode.ai/zen/v1/chat/completions`      | `@ai-sdk/openai-compatible` |
+| Qwen3.6 Plus Free     | qwen3.6-plus-free     | `https://opencode.ai/zen/v1/chat/completions`      | `@ai-sdk/openai-compatible` |
+| Nemotron 3 Super Free | nemotron-3-super-free | `https://opencode.ai/zen/v1/chat/completions`      | `@ai-sdk/openai-compatible` |
 
-El [model ID](/docs/config/#models) en tu configuracion de OpenCode
-usa el formato `opencode/<model-id>`. Por ejemplo, para GPT 5.2 Codex, debes
-usar `opencode/gpt-5.2-codex` en tu configuracion.
+El [identificador del modelo](/docs/config/#models) en tu configuración de OpenCode
+usa el formato `opencode/<model-id>`. Por ejemplo, para GPT 5.3 Codex, usarías
+`opencode/gpt-5.3-codex` en tu configuración.
 
 ---
 
 ### Modelos
 
-Puede obtener la lista completa de modelos disponibles y sus metadatos en:
+Puedes obtener la lista completa de modelos disponibles y sus metadatos desde:
 
 ```
 https://opencode.ai/zen/v1/models
@@ -114,150 +117,158 @@ https://opencode.ai/zen/v1/models
 
 ## Precios
 
-Apoyamos un modelo de pago por uso. A continuación se muestran los precios **por 1 millón de tokens**.
+Admitimos un modelo de pago por uso. A continuación se muestran los precios **por 1M tokens**.
 
-| Modelo                               | Entrada | Salida | Lectura en caché | Escritura en caché |
-| ------------------------------------ | ------- | ------ | ---------------- | ------------------ |
-| Big Pickle                           | Gratis  | Gratis | Gratis           | -                  |
-| MiniMax M2.5 Free                    | Gratis  | Gratis | Gratis           | -                  |
-| MiniMax M2.5                         | $0,30   | $1,20  | $0,06            | -                  |
-| MiniMax M2.1                         | $0,30   | $1,20  | $0,10            | -                  |
-| GLM 5                                | $1,00   | $3,20  | $0,20            | -                  |
-| GLM 4.7                              | $0,60   | $2.20  | $0,10            | -                  |
-| GLM 4.6                              | $0,60   | $2.20  | $0,10            | -                  |
-| Kimi K2.5                            | $0,60   | $3.00  | $0,08            | -                  |
-| Kimi K2 Thinking                     | $0,40   | $2.50  | -                | -                  |
-| Kimi K2                              | $0,40   | $2.50  | -                | -                  |
-| Qwen3 Coder 480B                     | $0,45   | $1,50  | -                | -                  |
-| Claude Opus 4.6 (≤ 200.000 tokens)   | $5.00   | $25.00 | $0,50            | $6.25              |
-| Claude Opus 4.6 (> 200K tokens)      | $10.00  | $37,50 | $1.00            | $12,50             |
-| Claude Opus 4.5                      | $5.00   | $25.00 | $0,50            | $6.25              |
-| Claude Opus 4.1                      | $15.00  | $75.00 | $1,50            | $18,75             |
-| Claude Sonnet 4.6 (≤ 200.000 tokens) | $3.00   | $15.00 | $0,30            | $3,75              |
-| Claude Sonnet 4.6 (> 200.000 tokens) | $6.00   | $22,50 | $0,60            | $7.50              |
-| Claude Sonnet 4.5 (≤ 200.000 tokens) | $3.00   | $15.00 | $0,30            | $3,75              |
-| Claude Sonnet 4.5 (> 200.000 tokens) | $6.00   | $22,50 | $0,60            | $7.50              |
-| Claude Sonnet 4 (≤ 200K tokens)      | $3.00   | $15.00 | $0,30            | $3,75              |
-| Claude Sonnet 4 (> 200K tokens)      | $6.00   | $22,50 | $0,60            | $7.50              |
-| Claude Haiku 4.5                     | $1.00   | $5.00  | $0,10            | $1,25              |
-| Claude Haiku 3.5                     | $0,80   | $4.00  | $0,08            | $1.00              |
-| Gemini 3.1 Pro (≤ 200K tokens)       | $2.00   | $12.00 | $0,20            | -                  |
-| Gemini 3.1 Pro (> 200K tokens)       | $4.00   | $18.00 | $0,40            | -                  |
-| Gemini 3 Pro (≤ 200K tokens)         | $2.00   | $12.00 | $0,20            | -                  |
-| Gemini 3 Pro (> 200K tokens)         | $4.00   | $18.00 | $0,40            | -                  |
-| Gemini 3 Flash                       | $0,50   | $3.00  | $0,05            | -                  |
-| GPT 5.4                              | $2,50   | $15,00 | $0,25            | -                  |
-| GPT 5.3 Codex                        | $1,75   | $14.00 | $0,175           | -                  |
-| GPT 5.2                              | $1,75   | $14.00 | $0,175           | -                  |
-| GPT 5.2 Codex                        | $1,75   | $14.00 | $0,175           | -                  |
-| GPT 5.1                              | $1.07   | $8,50  | $0,107           | -                  |
-| GPT 5.1 Codex                        | $1.07   | $8,50  | $0,107           | -                  |
-| GPT 5.1 Codex Max                    | $1,25   | $10.00 | $0,125           | -                  |
-| GPT 5.1 Codex Mini                   | $0,25   | $2.00  | $0,025           | -                  |
-| GPT 5                                | $1.07   | $8,50  | $0,107           | -                  |
-| GPT 5 Codex                          | $1.07   | $8,50  | $0,107           | -                  |
-| GPT 5 Nano                           | Gratis  | Gratis | Gratis           | -                  |
+| Modelo                            | Entrada | Salida  | Lectura en caché | Escritura en caché |
+| --------------------------------- | ------- | ------- | ---------------- | ------------------ |
+| Big Pickle                        | Free    | Free    | Free             | -                  |
+| MiMo V2 Pro Free                  | Free    | Free    | Free             | -                  |
+| MiMo V2 Omni Free                 | Free    | Free    | Free             | -                  |
+| Qwen3.6 Plus Free                 | Free    | Free    | Free             | -                  |
+| Nemotron 3 Super Free             | Free    | Free    | Free             | -                  |
+| MiniMax M2.5 Free                 | Free    | Free    | Free             | -                  |
+| MiniMax M2.5                      | $0.30   | $1.20   | $0.06            | $0.375             |
+| GLM 5                             | $1.00   | $3.20   | $0.20            | -                  |
+| Kimi K2.5                         | $0.60   | $3.00   | $0.10            | -                  |
+| Qwen3 Coder 480B                  | $0.45   | $1.50   | -                | -                  |
+| Claude Opus 4.6                   | $5.00   | $25.00  | $0.50            | $6.25              |
+| Claude Opus 4.5                   | $5.00   | $25.00  | $0.50            | $6.25              |
+| Claude Opus 4.1                   | $15.00  | $75.00  | $1.50            | $18.75             |
+| Claude Sonnet 4.6                 | $3.00   | $15.00  | $0.30            | $3.75              |
+| Claude Sonnet 4.5 (≤ 200K tokens) | $3.00   | $15.00  | $0.30            | $3.75              |
+| Claude Sonnet 4.5 (> 200K tokens) | $6.00   | $22.50  | $0.60            | $7.50              |
+| Claude Sonnet 4 (≤ 200K tokens)   | $3.00   | $15.00  | $0.30            | $3.75              |
+| Claude Sonnet 4 (> 200K tokens)   | $6.00   | $22.50  | $0.60            | $7.50              |
+| Claude Haiku 4.5                  | $1.00   | $5.00   | $0.10            | $1.25              |
+| Claude Haiku 3.5                  | $0.80   | $4.00   | $0.08            | $1.00              |
+| Gemini 3.1 Pro (≤ 200K tokens)    | $2.00   | $12.00  | $0.20            | -                  |
+| Gemini 3.1 Pro (> 200K tokens)    | $4.00   | $18.00  | $0.40            | -                  |
+| Gemini 3 Flash                    | $0.50   | $3.00   | $0.05            | -                  |
+| GPT 5.4                           | $2.50   | $15.00  | $0.25            | -                  |
+| GPT 5.4 Pro                       | $30.00  | $180.00 | $30.00           | -                  |
+| GPT 5.4 Mini                      | $0.75   | $4.50   | $0.075           | -                  |
+| GPT 5.4 Nano                      | $0.20   | $1.25   | $0.02            | -                  |
+| GPT 5.3 Codex Spark               | $1.75   | $14.00  | $0.175           | -                  |
+| GPT 5.3 Codex                     | $1.75   | $14.00  | $0.175           | -                  |
+| GPT 5.2                           | $1.75   | $14.00  | $0.175           | -                  |
+| GPT 5.2 Codex                     | $1.75   | $14.00  | $0.175           | -                  |
+| GPT 5.1                           | $1.07   | $8.50   | $0.107           | -                  |
+| GPT 5.1 Codex                     | $1.07   | $8.50   | $0.107           | -                  |
+| GPT 5.1 Codex Max                 | $1.25   | $10.00  | $0.125           | -                  |
+| GPT 5.1 Codex Mini                | $0.25   | $2.00   | $0.025           | -                  |
+| GPT 5                             | $1.07   | $8.50   | $0.107           | -                  |
+| GPT 5 Codex                       | $1.07   | $8.50   | $0.107           | -                  |
+| GPT 5 Nano                        | Free    | Free    | Free             | -                  |
 
-Es posible que notes _Claude Haiku 3.5_ en tu historial de uso. Este es un [modelo de bajo costo](/docs/config/#models) que se utiliza para generar los títulos de tus sesiones.
+Puede que notes _Claude Haiku 3.5_ en tu historial de uso. Este es un [modelo de bajo costo](/docs/config/#models) que se usa para generar los títulos de tus sesiones.
 
 :::note
-Las tarifas de las tarjetas de crédito se trasladan al costo (4,4% + 0,30 dólares por transacción); No cobramos nada más allá de eso.
+Las comisiones de tarjeta de crédito se trasladan al costo (4.4% + $0.30 por transacción); no cobramos nada más allá de eso.
 :::
 
 Los modelos gratuitos:
 
-- MiniMax M2.5 Free está disponible en OpenCode por tiempo limitado. El equipo está aprovechando este tiempo para recopilar comentarios y mejorar el modelo.
-- Big Pickle es un modelo sigiloso gratuito en OpenCode por tiempo limitado. El equipo está aprovechando este tiempo para recopilar comentarios y mejorar el modelo.
+- MiniMax M2.5 Free está disponible en OpenCode por tiempo limitado. El equipo está usando este tiempo para recopilar comentarios y mejorar el modelo.
+- MiMo V2 Pro Free está disponible en OpenCode por tiempo limitado. El equipo está usando este tiempo para recopilar comentarios y mejorar el modelo.
+- MiMo V2 Omni Free está disponible en OpenCode por tiempo limitado. El equipo está usando este tiempo para recopilar comentarios y mejorar el modelo.
+- Qwen3.6 Plus Free está disponible en OpenCode por tiempo limitado. El equipo está usando este tiempo para recopilar comentarios y mejorar el modelo.
+- Nemotron 3 Super Free está disponible en OpenCode por tiempo limitado. El equipo está usando este tiempo para recopilar comentarios y mejorar el modelo.
+- Big Pickle es un modelo stealth que es gratuito en OpenCode por tiempo limitado. El equipo está usando este tiempo para recopilar comentarios y mejorar el modelo.
 
-<a href={email}>Contáctenos</a> si tiene alguna pregunta.
+<a href={email}>Contáctanos</a> si tienes alguna pregunta.
 
 ---
 
 ### Recarga automática
 
-Si su saldo es inferior a $5, Zen recargará automáticamente $20.
+Si tu saldo baja de $5, Zen recargará automáticamente $20.
 
-Puedes cambiar el monto de la recarga automática. También puedes desactivar la recarga automática por completo.
+Puedes cambiar el monto de la recarga automática. También puedes desactivar por completo la recarga automática.
 
 ---
 
 ### Límites mensuales
 
-También puede establecer un límite de uso mensual para todo el espacio de trabajo y para cada
+También puedes establecer un límite de uso mensual para todo el workspace y para cada
 miembro de tu equipo.
 
-Por ejemplo, digamos que establece un límite de uso mensual de $20, Zen no usará
+Por ejemplo, supongamos que estableces un límite de uso mensual de $20. Zen no usará
 más de $20 en un mes. Pero si tienes habilitada la recarga automática, Zen podría terminar
-cobrarle más de $20 si su saldo es inferior a $5.
+cobrándote más de $20 si tu saldo baja de $5.
 
 ---
 
 ### Modelos obsoletos
 
-| Modelo           | Fecha de retiro     |
-| ---------------- | ------------------- |
-| Qwen3 Coder 480B | 6 de feb. de 2026   |
-| Kimi K2 Thinking | 6 de marzo de 2026  |
-| Kimi K2          | 6 de marzo de 2026  |
-| MiniMax M2.1     | 15 de marzo de 2026 |
-| GLM 4.7          | 15 de marzo de 2026 |
-| GLM 4.6          | 15 de marzo de 2026 |
+| Modelo           | Fecha de retirada |
+| ---------------- | ----------------- |
+| MiniMax M2.1     | March 15, 2026    |
+| GLM 4.7          | March 15, 2026    |
+| GLM 4.6          | March 15, 2026    |
+| Gemini 3 Pro     | March 9, 2026     |
+| Kimi K2 Thinking | March 6, 2026     |
+| Kimi K2          | March 6, 2026     |
+| Qwen3 Coder 480B | Feb 6, 2026       |
 
 ---
 
 ## Privacidad
 
-Todos nuestros modelos están alojados en los EE. UU. Nuestros proveedores siguen una política de retención cero y no utilizan sus datos para la capacitación de modelos, con las siguientes excepciones:
+Todos nuestros modelos están alojados en US. Nuestros proveedores siguen una política de zero-retention y no usan tus datos para el entrenamiento de modelos, con las siguientes excepciones:
 
-- Big Pickle: Durante su periodo gratuito, los datos recopilados podrán utilizarse para mejorar el modelo.
-- MiniMax M2.5 Free: Durante su período gratuito, los datos recopilados podrán utilizarse para mejorar el modelo.
-- API de OpenAI: las solicitudes se conservan durante 30 días de acuerdo con las [Políticas de datos de OpenAI](https://platform.openai.com/docs/guides/your-data).
-- API de Anthropic: las solicitudes se conservan durante 30 días de acuerdo con las [Políticas de datos de Anthropic](https://docs.anthropic.com/en/docs/claude-code/data-usage).
+- Big Pickle: Durante su período gratuito, los datos recopilados pueden usarse para mejorar el modelo.
+- MiniMax M2.5 Free: Durante su período gratuito, los datos recopilados pueden usarse para mejorar el modelo.
+- MiMo V2 Pro Free: Durante su período gratuito, los datos recopilados pueden usarse para mejorar el modelo.
+- MiMo V2 Omni Free: Durante su período gratuito, los datos recopilados pueden usarse para mejorar el modelo.
+- Qwen3.6 Plus Free: Durante su período gratuito, los datos recopilados pueden usarse para mejorar el modelo.
+- Nemotron 3 Super Free: Durante su período gratuito, los datos recopilados pueden usarse para mejorar el modelo.
+- OpenAI APIs: Las solicitudes se conservan durante 30 días de acuerdo con [OpenAI's Data Policies](https://platform.openai.com/docs/guides/your-data).
+- Anthropic APIs: Las solicitudes se conservan durante 30 días de acuerdo con [Anthropic's Data Policies](https://docs.anthropic.com/en/docs/claude-code/data-usage).
 
 ---
 
 ## Para equipos
 
 Zen también funciona muy bien para equipos. Puedes invitar a compañeros de equipo, asignar roles, seleccionar
-los modelos que utiliza su equipo y más.
+los modelos que usa tu equipo y más.
 
 :::note
-Actualmente, los espacios de trabajo son gratuitos para los equipos como parte de la versión beta.
+Los workspaces son actualmente gratuitos para equipos como parte de la beta.
 :::
 
-Actualmente, administrar su espacio de trabajo es gratuito para equipos como parte de la versión beta. estaremos
-compartiremos más detalles sobre los precios pronto.
+Gestionar tu workspace es actualmente gratuito para equipos como parte de la beta. Compartiremos
+más detalles sobre los precios pronto.
 
 ---
 
 ### Roles
 
-Puedes invitar a compañeros de equipo a tu espacio de trabajo y asignar roles:
+Puedes invitar a compañeros de equipo a tu workspace y asignar roles:
 
-- **Administrador**: administra modelos, miembros, claves API y facturación
-- **Miembro**: administra solo sus propias claves API
+- **Admin**: Gestiona modelos, miembros, API keys y facturación
+- **Member**: Gestiona solo sus propias API keys
 
-Los administradores también pueden establecer límites de gastos mensuales para cada miembro para mantener los costos bajo control.
+Los admins también pueden establecer límites mensuales de gasto para cada miembro para mantener los costos bajo control.
 
 ---
 
-### Acceso al modelo
+### Acceso a modelos
 
-Los administradores pueden habilitar o deshabilitar modelos específicos para el espacio de trabajo. Las solicitudes realizadas a un modelo deshabilitado devolverán un error.
+Los admins pueden habilitar o deshabilitar modelos específicos para el workspace. Las solicitudes realizadas a un modelo deshabilitado devolverán un error.
 
-Esto es útil para los casos en los que desea desactivar el uso de un modelo que
+Esto resulta útil en los casos en los que quieres deshabilitar el uso de un modelo que
 recopila datos.
 
 ---
 
-### Trae tu propia clave API
+### Bring your own key
 
-Puedes usar tus propias API keys de OpenAI o Anthropic mientras accedes a otros modelos en Zen.
+Puedes usar tus propias API keys de OpenAI o Anthropic mientras sigues accediendo a otros modelos en Zen.
 
-Cuando utiliza sus propias claves, los tokens los factura directamente el proveedor, no Zen.
+Cuando usas tus propias keys, los tokens te los factura directamente el proveedor, no Zen.
 
-Por ejemplo, es posible que su organización ya tenga una clave para OpenAI o Anthropic.
-y quieres usar ese en lugar del que proporciona Zen.
+Por ejemplo, tu organización podría ya tener una key para OpenAI o Anthropic
+y quieres usarla en lugar de la que proporciona Zen.
 
 ---
 
@@ -265,7 +276,7 @@ y quieres usar ese en lugar del que proporciona Zen.
 
 Creamos OpenCode Zen para:
 
-1. **Evaluar** los mejores modelos/proveedores de agentes de codificación.
-2. Tener acceso a las opciones de **más alta calidad** y no degradar el rendimiento ni recurrir a proveedores más baratos.
-3. Transmitir cualquier **bajada de precio** vendiendo al costo; por lo que el único margen de beneficio es para cubrir nuestras tarifas de procesamiento.
-4. **No tener ningún bloqueo** al permitirle usarlo con cualquier otro agente de codificación. Y siempre le permitirá utilizar cualquier otro proveedor con OpenCode también.
+1. **Benchmark** de los mejores modelos/proveedores para agentes de coding.
+2. Tener acceso a las opciones de **más alta calidad** y no degradar el rendimiento ni enrutar a proveedores más baratos.
+3. Trasladar cualquier **bajada de precio** vendiendo al costo; así que el único margen es para cubrir nuestras comisiones de procesamiento.
+4. Que no haya **lock-in** al permitirte usarlo con cualquier otro agente de coding. Y permitirte siempre usar también cualquier otro proveedor con OpenCode.

packages/web/src/content/docs/fr/zen.mdx

@@ -13,27 +13,27 @@ OpenCode Zen est une liste de modèles testés et vérifiés fournie par l'équi
 OpenCode Zen est actuellement en version bêta.
 :::
 
-Zen fonctionne comme n'importe quel autre fournisseur dans OpenCode. Vous vous connectez à OpenCode Zen et obtenez votre clé API. C'est **complètement facultatif** et vous n'avez pas besoin de l'utiliser pour utiliser OpenCode.
+Zen fonctionne comme n'importe quel autre fournisseur dans OpenCode. Vous vous connectez à OpenCode Zen et obtenez votre clé API. C'est **entièrement facultatif** et vous n'avez pas besoin de l'utiliser pour utiliser OpenCode.
 
 ---
 
 ## Arrière-plan
 
-Il existe un grand nombre de modèles, mais seulement quelques-uns d'entre eux fonctionnent bien comme agents de codage. De plus, la plupart des fournisseurs sont configurés très différemment; vous obtenez donc des performances et une qualité très différentes.
+Il existe un grand nombre de modèles, mais seuls quelques-uns fonctionnent bien comme agents de codage. De plus, la plupart des fournisseurs sont configurés très différemment ; vous obtenez donc des performances et une qualité très variables.
 
 :::tip
 Nous avons testé un groupe sélectionné de modèles et de fournisseurs qui fonctionnent bien avec OpenCode.
 :::
 
-Donc, si vous utilisez un modèle via quelque chose comme OpenRouter, vous ne pourrez jamais être sûr que vous obtenez la meilleure version du modèle que vous souhaitez.
+Donc, si vous utilisez un modèle via quelque chose comme OpenRouter, vous ne pouvez jamais être sûr d'obtenir la meilleure version du modèle que vous voulez.
 
-Pour résoudre ce problème, nous avons effectué plusieurs opérations :
+Pour corriger cela, nous avons fait plusieurs choses :
 
-1. Nous avons testé un groupe sélectionné de modèles et discuté avec leurs équipes de la manière de mieux les exécuter.
-2. Nous avons ensuite travaillé avec quelques prestataires pour nous assurer qu'ils étaient servis correctement.
-3. Enfin, nous avons comparé la combinaison modèle/fournisseur et sommes arrivés avec une liste que nous nous ferons un plaisir de recommander.
+1. Nous avons testé un groupe sélectionné de modèles et discuté avec leurs équipes de la meilleure façon de les faire fonctionner.
+2. Nous avons ensuite travaillé avec quelques fournisseurs pour nous assurer qu'ils étaient correctement servis.
+3. Enfin, nous avons comparé la combinaison modèle/fournisseur et établi une liste que nous recommandons en toute confiance.
 
-OpenCode Zen est une passerelle IA qui vous donne accès à ces modèles.
+OpenCode Zen est une passerelle AI qui vous donne accès à ces modèles.
 
 ---
 
@@ -45,7 +45,7 @@ OpenCode Zen fonctionne comme n'importe quel autre fournisseur dans OpenCode.
 2. Vous exécutez la commande `/connect` dans le TUI, sélectionnez OpenCode Zen et collez votre clé API.
 3. Exécutez `/models` dans le TUI pour voir la liste des modèles que nous recommandons.
 
-Vous êtes facturé par demande et vous pouvez ajouter des crédits à votre compte.
+La facturation se fait à la requête et vous pouvez ajouter des crédits à votre compte.
 
 ---
 
@@ -53,43 +53,44 @@ Vous êtes facturé par demande et vous pouvez ajouter des crédits à votre com
 
 Vous pouvez également accéder à nos modèles via les points de terminaison API suivants.
 
-| Modèle             | ID du modèle       | Point de terminaison                               | Package SDK IA              |
-| ------------------ | ------------------ | -------------------------------------------------- | --------------------------- |
-| GPT 5.4            | gpt-5.4            | `https://opencode.ai/zen/v1/responses`             | `@ai-sdk/openai`            |
-| GPT 5.3 Codex      | gpt-5.3-codex      | `https://opencode.ai/zen/v1/responses`             | `@ai-sdk/openai`            |
-| GPT 5.2            | gpt-5.2            | `https://opencode.ai/zen/v1/responses`             | `@ai-sdk/openai`            |
-| GPT 5.2 Codex      | gpt-5.2-codex      | `https://opencode.ai/zen/v1/responses`             | `@ai-sdk/openai`            |
-| GPT 5.1            | gpt-5.1            | `https://opencode.ai/zen/v1/responses`             | `@ai-sdk/openai`            |
-| GPT 5.1 Codex      | gpt-5.1-codex      | `https://opencode.ai/zen/v1/responses`             | `@ai-sdk/openai`            |
-| GPT 5.1 Codex Max  | gpt-5.1-codex-max  | `https://opencode.ai/zen/v1/responses`             | `@ai-sdk/openai`            |
-| GPT 5.1 Codex Mini | gpt-5.1-codex-mini | `https://opencode.ai/zen/v1/responses`             | `@ai-sdk/openai`            |
-| GPT 5              | gpt-5              | `https://opencode.ai/zen/v1/responses`             | `@ai-sdk/openai`            |
-| GPT 5 Codex        | gpt-5-codex        | `https://opencode.ai/zen/v1/responses`             | `@ai-sdk/openai`            |
-| GPT 5 Nano         | gpt-5-nano         | `https://opencode.ai/zen/v1/responses`             | `@ai-sdk/openai`            |
-| Claude Opus 4.6    | claude-opus-4-6    | `https://opencode.ai/zen/v1/messages`              | `@ai-sdk/anthropic`         |
-| Claude Opus 4.5    | claude-opus-4-5    | `https://opencode.ai/zen/v1/messages`              | `@ai-sdk/anthropic`         |
-| Claude Opus 4.1    | claude-opus-4-1    | `https://opencode.ai/zen/v1/messages`              | `@ai-sdk/anthropic`         |
-| Claude Sonnet 4.6  | claude-sonnet-4-6  | `https://opencode.ai/zen/v1/messages`              | `@ai-sdk/anthropic`         |
-| Claude Sonnet 4.5  | claude-sonnet-4-5  | `https://opencode.ai/zen/v1/messages`              | `@ai-sdk/anthropic`         |
-| Claude Sonnet 4    | claude-sonnet-4    | `https://opencode.ai/zen/v1/messages`              | `@ai-sdk/anthropic`         |
-| Claude Haiku 4.5   | claude-haiku-4-5   | `https://opencode.ai/zen/v1/messages`              | `@ai-sdk/anthropic`         |
-| Claude Haiku 3.5   | claude-3-5-haiku   | `https://opencode.ai/zen/v1/messages`              | `@ai-sdk/anthropic`         |
-| Gemini 3.1 Pro     | gemini-3.1-pro     | `https://opencode.ai/zen/v1/models/gemini-3.1-pro` | `@ai-sdk/google`            |
-| Gemini 3 Pro       | gemini-3-pro       | `https://opencode.ai/zen/v1/models/gemini-3-pro`   | `@ai-sdk/google`            |
-| Gemini 3 Flash     | gemini-3-flash     | `https://opencode.ai/zen/v1/models/gemini-3-flash` | `@ai-sdk/google`            |
-| MiniMax M2.5       | minimax-m2.5       | `https://opencode.ai/zen/v1/chat/completions`      | `@ai-sdk/openai-compatible` |
-| MiniMax M2.5 Free  | minimax-m2.5-free  | `https://opencode.ai/zen/v1/chat/completions`      | `@ai-sdk/openai-compatible` |
-| MiniMax M2.1       | minimax-m2.1       | `https://opencode.ai/zen/v1/chat/completions`      | `@ai-sdk/openai-compatible` |
-| GLM 5              | glm-5              | `https://opencode.ai/zen/v1/chat/completions`      | `@ai-sdk/openai-compatible` |
-| GLM 4.7            | glm-4.7            | `https://opencode.ai/zen/v1/chat/completions`      | `@ai-sdk/openai-compatible` |
-| GLM 4.6            | glm-4.6            | `https://opencode.ai/zen/v1/chat/completions`      | `@ai-sdk/openai-compatible` |
-| Kimi K2.5          | kimi-k2.5          | `https://opencode.ai/zen/v1/chat/completions`      | `@ai-sdk/openai-compatible` |
-| Kimi K2 Thinking   | kimi-k2-thinking   | `https://opencode.ai/zen/v1/chat/completions`      | `@ai-sdk/openai-compatible` |
-| Kimi K2            | kimi-k2            | `https://opencode.ai/zen/v1/chat/completions`      | `@ai-sdk/openai-compatible` |
-| Qwen3 Coder 480B   | qwen3-coder        | `https://opencode.ai/zen/v1/chat/completions`      | `@ai-sdk/openai-compatible` |
-| Big Pickle         | big-pickle         | `https://opencode.ai/zen/v1/chat/completions`      | `@ai-sdk/openai-compatible` |
+| Modèle                | ID du modèle          | Point de terminaison                               | Package AI SDK              |
+| --------------------- | --------------------- | -------------------------------------------------- | --------------------------- |
+| GPT 5.4               | gpt-5.4               | `https://opencode.ai/zen/v1/responses`             | `@ai-sdk/openai`            |
+| GPT 5.4 Pro           | gpt-5.4-pro           | `https://opencode.ai/zen/v1/responses`             | `@ai-sdk/openai`            |
+| GPT 5.4 Mini          | gpt-5.4-mini          | `https://opencode.ai/zen/v1/responses`             | `@ai-sdk/openai`            |
+| GPT 5.4 Nano          | gpt-5.4-nano          | `https://opencode.ai/zen/v1/responses`             | `@ai-sdk/openai`            |
+| GPT 5.3 Codex         | gpt-5.3-codex         | `https://opencode.ai/zen/v1/responses`             | `@ai-sdk/openai`            |
+| GPT 5.3 Codex Spark   | gpt-5.3-codex-spark   | `https://opencode.ai/zen/v1/responses`             | `@ai-sdk/openai`            |
+| GPT 5.2               | gpt-5.2               | `https://opencode.ai/zen/v1/responses`             | `@ai-sdk/openai`            |
+| GPT 5.2 Codex         | gpt-5.2-codex         | `https://opencode.ai/zen/v1/responses`             | `@ai-sdk/openai`            |
+| GPT 5.1               | gpt-5.1               | `https://opencode.ai/zen/v1/responses`             | `@ai-sdk/openai`            |
+| GPT 5.1 Codex         | gpt-5.1-codex         | `https://opencode.ai/zen/v1/responses`             | `@ai-sdk/openai`            |
+| GPT 5.1 Codex Max     | gpt-5.1-codex-max     | `https://opencode.ai/zen/v1/responses`             | `@ai-sdk/openai`            |
+| GPT 5.1 Codex Mini    | gpt-5.1-codex-mini    | `https://opencode.ai/zen/v1/responses`             | `@ai-sdk/openai`            |
+| GPT 5                 | gpt-5                 | `https://opencode.ai/zen/v1/responses`             | `@ai-sdk/openai`            |
+| GPT 5 Codex           | gpt-5-codex           | `https://opencode.ai/zen/v1/responses`             | `@ai-sdk/openai`            |
+| GPT 5 Nano            | gpt-5-nano            | `https://opencode.ai/zen/v1/responses`             | `@ai-sdk/openai`            |
+| Claude Opus 4.6       | claude-opus-4-6       | `https://opencode.ai/zen/v1/messages`              | `@ai-sdk/anthropic`         |
+| Claude Opus 4.5       | claude-opus-4-5       | `https://opencode.ai/zen/v1/messages`              | `@ai-sdk/anthropic`         |
+| Claude Opus 4.1       | claude-opus-4-1       | `https://opencode.ai/zen/v1/messages`              | `@ai-sdk/anthropic`         |
+| Claude Sonnet 4.6     | claude-sonnet-4-6     | `https://opencode.ai/zen/v1/messages`              | `@ai-sdk/anthropic`         |
+| Claude Sonnet 4.5     | claude-sonnet-4-5     | `https://opencode.ai/zen/v1/messages`              | `@ai-sdk/anthropic`         |
+| Claude Sonnet 4       | claude-sonnet-4       | `https://opencode.ai/zen/v1/messages`              | `@ai-sdk/anthropic`         |
+| Claude Haiku 4.5      | claude-haiku-4-5      | `https://opencode.ai/zen/v1/messages`              | `@ai-sdk/anthropic`         |
+| Claude Haiku 3.5      | claude-3-5-haiku      | `https://opencode.ai/zen/v1/messages`              | `@ai-sdk/anthropic`         |
+| Gemini 3.1 Pro        | gemini-3.1-pro        | `https://opencode.ai/zen/v1/models/gemini-3.1-pro` | `@ai-sdk/google`            |
+| Gemini 3 Flash        | gemini-3-flash        | `https://opencode.ai/zen/v1/models/gemini-3-flash` | `@ai-sdk/google`            |
+| MiniMax M2.5          | minimax-m2.5          | `https://opencode.ai/zen/v1/chat/completions`      | `@ai-sdk/openai-compatible` |
+| MiniMax M2.5 Free     | minimax-m2.5-free     | `https://opencode.ai/zen/v1/chat/completions`      | `@ai-sdk/openai-compatible` |
+| GLM 5                 | glm-5                 | `https://opencode.ai/zen/v1/chat/completions`      | `@ai-sdk/openai-compatible` |
+| Kimi K2.5             | kimi-k2.5             | `https://opencode.ai/zen/v1/chat/completions`      | `@ai-sdk/openai-compatible` |
+| Big Pickle            | big-pickle            | `https://opencode.ai/zen/v1/chat/completions`      | `@ai-sdk/openai-compatible` |
+| MiMo V2 Pro Free      | mimo-v2-pro-free      | `https://opencode.ai/zen/v1/chat/completions`      | `@ai-sdk/openai-compatible` |
+| MiMo V2 Omni Free     | mimo-v2-omni-free     | `https://opencode.ai/zen/v1/chat/completions`      | `@ai-sdk/openai-compatible` |
+| Qwen3.6 Plus Free     | qwen3.6-plus-free     | `https://opencode.ai/zen/v1/chat/completions`      | `@ai-sdk/openai-compatible` |
+| Nemotron 3 Super Free | nemotron-3-super-free | `https://opencode.ai/zen/v1/chat/completions`      | `@ai-sdk/openai-compatible` |
 
-Le [id du modèle](/docs/config/#models) dans votre configuration OpenCode utilise le format `opencode/<model-id>`. Par exemple, pour GPT 5.2 Codex, vous devez utilisez `opencode/gpt-5.2-codex` dans votre configuration.
+Le [model id](/docs/config/#models) dans votre configuration OpenCode utilise le format `opencode/<model-id>`. Par exemple, pour GPT 5.3 Codex, vous utiliseriez `opencode/gpt-5.3-codex` dans votre configuration.
 
 ---
 
@@ -103,62 +104,65 @@ https://opencode.ai/zen/v1/models
 
 ---
 
-## Tarifs
+## Tarification
 
-Nous soutenons un modèle de paiement à l'utilisation. Vous trouverez ci-dessous les prix **par 1 million de jetons**.
+Nous prenons en charge un modèle de paiement à l'utilisation. Vous trouverez ci-dessous les prix **par 1M tokens**.
 
-| Modèle                            | Entrée  | Sortie  | Lecture en cache | Écriture en cache |
-| --------------------------------- | ------- | ------- | ---------------- | ----------------- |
-| Big Pickle                        | Gratuit | Gratuit | Gratuit          | -                 |
-| MiniMax M2.5 Free                 | Gratuit | Gratuit | Gratuit          | -                 |
-| MiniMax M2.5                      | 0,30 $  | 1,20 $  | 0,06 $           | -                 |
-| MiniMax M2.1                      | 0,30 $  | 1,20 $  | 0,10 $           | -                 |
-| GLM 5                             | 1,00 $  | 3,20 $  | 0,20 $           | -                 |
-| GLM 4.7                           | 0,60 $  | 2,20 $  | 0,10 $           | -                 |
-| GLM 4.6                           | 0,60 $  | 2,20 $  | 0,10 $           | -                 |
-| Kimi K2.5                         | 0,60 $  | 3,00 $  | 0,08 $           | -                 |
-| Kimi K2 Thinking                  | 0,40 $  | 2,50 $  | -                | -                 |
-| Kimi K2                           | 0,40 $  | 2,50 $  | -                | -                 |
-| Qwen3 Coder 480B                  | 0,45 $  | 1,50 $  | -                | -                 |
-| Claude Opus 4.6 (≤ 200K jetons)   | 5,00 $  | 25,00 $ | 0,50 $           | 6,25 $            |
-| Claude Opus 4.6 (> 200K jetons)   | 10,00 $ | 37,50 $ | 1,00 $           | 12,50 $           |
-| Claude Opus 4.5                   | 5,00 $  | 25,00 $ | 0,50 $           | 6,25 $            |
-| Claude Opus 4.1                   | 15,00 $ | 75,00 $ | 1,50 $           | 18,75 $           |
-| Claude Sonnet 4.6 (≤ 200K jetons) | 3,00 $  | 15,00 $ | 0,30 $           | 3,75 $            |
-| Claude Sonnet 4.6 (> 200K jetons) | 6,00 $  | 22,50 $ | 0,60 $           | 7,50 $            |
-| Claude Sonnet 4.5 (≤ 200K jetons) | 3,00 $  | 15,00 $ | 0,30 $           | 3,75 $            |
-| Claude Sonnet 4.5 (> 200K jetons) | 6,00 $  | 22,50 $ | 0,60 $           | 7,50 $            |
-| Claude Sonnet 4 (≤ 200K jetons)   | 3,00 $  | 15,00 $ | 0,30 $           | 3,75 $            |
-| Claude Sonnet 4 (> 200K jetons)   | 6,00 $  | 22,50 $ | 0,60 $           | 7,50 $            |
-| Claude Haiku 4.5                  | 1,00 $  | 5,00 $  | 0,10 $           | 1,25 $            |
-| Claude Haiku 3.5                  | 0,80 $  | 4,00 $  | 0,08 $           | 1,00 $            |
-| Gemini 3.1 Pro (≤ 200K jetons)    | 2,00 $  | 12,00 $ | 0,20 $           | -                 |
-| Gemini 3.1 Pro (> 200K jetons)    | 4,00 $  | 18,00 $ | 0,40 $           | -                 |
-| Gemini 3 Pro (≤ 200K jetons)      | 2,00 $  | 12,00 $ | 0,20 $           | -                 |
-| Gemini 3 Pro (> 200K jetons)      | 4,00 $  | 18,00 $ | 0,40 $           | -                 |
-| Gemini 3 Flash                    | 0,50 $  | 3,00 $  | 0,05 $           | -                 |
-| GPT 5.4                           | 2,50 $  | 15,00 $ | 0,25 $           | -                 |
-| GPT 5.3 Codex                     | 1,75 $  | 14,00 $ | 0,175 $          | -                 |
-| GPT 5.2                           | 1,75 $  | 14,00 $ | 0,175 $          | -                 |
-| GPT 5.2 Codex                     | 1,75 $  | 14,00 $ | 0,175 $          | -                 |
-| GPT 5.1                           | 1,07 $  | 8,50 $  | 0,107 $          | -                 |
-| GPT 5.1 Codex                     | 1,07 $  | 8,50 $  | 0,107 $          | -                 |
-| GPT 5.1 Codex Max                 | 1,25 $  | 10,00 $ | 0,125 $          | -                 |
-| GPT 5.1 Codex Mini                | 0,25 $  | 2,00 $  | 0,025 $          | -                 |
-| GPT 5                             | 1,07 $  | 8,50 $  | 0,107 $          | -                 |
-| GPT 5 Codex                       | 1,07 $  | 8,50 $  | 0,107 $          | -                 |
-| GPT 5 Nano                        | Gratuit | Gratuit | Gratuit          | -                 |
+| Modèle                            | Input  | Output  | Cached Read | Cached Write |
+| --------------------------------- | ------ | ------- | ----------- | ------------ |
+| Big Pickle                        | Free   | Free    | Free        | -            |
+| MiMo V2 Pro Free                  | Free   | Free    | Free        | -            |
+| MiMo V2 Omni Free                 | Free   | Free    | Free        | -            |
+| Qwen3.6 Plus Free                 | Free   | Free    | Free        | -            |
+| Nemotron 3 Super Free             | Free   | Free    | Free        | -            |
+| MiniMax M2.5 Free                 | Free   | Free    | Free        | -            |
+| MiniMax M2.5                      | $0.30  | $1.20   | $0.06       | $0.375       |
+| GLM 5                             | $1.00  | $3.20   | $0.20       | -            |
+| Kimi K2.5                         | $0.60  | $3.00   | $0.10       | -            |
+| Qwen3 Coder 480B                  | $0.45  | $1.50   | -           | -            |
+| Claude Opus 4.6                   | $5.00  | $25.00  | $0.50       | $6.25        |
+| Claude Opus 4.5                   | $5.00  | $25.00  | $0.50       | $6.25        |
+| Claude Opus 4.1                   | $15.00 | $75.00  | $1.50       | $18.75       |
+| Claude Sonnet 4.6                 | $3.00  | $15.00  | $0.30       | $3.75        |
+| Claude Sonnet 4.5 (≤ 200K tokens) | $3.00  | $15.00  | $0.30       | $3.75        |
+| Claude Sonnet 4.5 (> 200K tokens) | $6.00  | $22.50  | $0.60       | $7.50        |
+| Claude Sonnet 4 (≤ 200K tokens)   | $3.00  | $15.00  | $0.30       | $3.75        |
+| Claude Sonnet 4 (> 200K tokens)   | $6.00  | $22.50  | $0.60       | $7.50        |
+| Claude Haiku 4.5                  | $1.00  | $5.00   | $0.10       | $1.25        |
+| Claude Haiku 3.5                  | $0.80  | $4.00   | $0.08       | $1.00        |
+| Gemini 3.1 Pro (≤ 200K tokens)    | $2.00  | $12.00  | $0.20       | -            |
+| Gemini 3.1 Pro (> 200K tokens)    | $4.00  | $18.00  | $0.40       | -            |
+| Gemini 3 Flash                    | $0.50  | $3.00   | $0.05       | -            |
+| GPT 5.4                           | $2.50  | $15.00  | $0.25       | -            |
+| GPT 5.4 Pro                       | $30.00 | $180.00 | $30.00      | -            |
+| GPT 5.4 Mini                      | $0.75  | $4.50   | $0.075      | -            |
+| GPT 5.4 Nano                      | $0.20  | $1.25   | $0.02       | -            |
+| GPT 5.3 Codex Spark               | $1.75  | $14.00  | $0.175      | -            |
+| GPT 5.3 Codex                     | $1.75  | $14.00  | $0.175      | -            |
+| GPT 5.2                           | $1.75  | $14.00  | $0.175      | -            |
+| GPT 5.2 Codex                     | $1.75  | $14.00  | $0.175      | -            |
+| GPT 5.1                           | $1.07  | $8.50   | $0.107      | -            |
+| GPT 5.1 Codex                     | $1.07  | $8.50   | $0.107      | -            |
+| GPT 5.1 Codex Max                 | $1.25  | $10.00  | $0.125      | -            |
+| GPT 5.1 Codex Mini                | $0.25  | $2.00   | $0.025      | -            |
+| GPT 5                             | $1.07  | $8.50   | $0.107      | -            |
+| GPT 5 Codex                       | $1.07  | $8.50   | $0.107      | -            |
+| GPT 5 Nano                        | Free   | Free    | Free        | -            |
 
-Vous remarquerez peut-être _Claude Haiku 3.5_ dans votre historique d'utilisation. Il s'agit d'un [modèle low cost](/docs/config/#models) qui est utilisé pour générer les titres de vos sessions.
+Vous remarquerez peut-être _Claude Haiku 3.5_ dans votre historique d'utilisation. Il s'agit d'un [modèle à faible coût](/docs/config/#models) utilisé pour générer les titres de vos sessions.
 
 :::note
-Les frais de carte de crédit sont répercutés au prix coûtant (4,4 % + 0,30 $ par transaction) ; nous ne facturons rien au-delà.
+Les frais de carte de crédit sont répercutés au prix coûtant (4.4% + $0.30 par transaction) ; nous ne facturons rien au-delà.
 :::
 
 Les modèles gratuits :
 
-- MiniMax M2.5 Free est disponible sur OpenCode pour une durée limitée. L’équipe profite de ce temps pour recueillir des commentaires et améliorer le modèle.
-- Big Pickle est un modèle furtif gratuit sur OpenCode pour une durée limitée. L’équipe profite de ce temps pour recueillir des commentaires et améliorer le modèle.
+- MiniMax M2.5 Free est disponible sur OpenCode pour une durée limitée. L'équipe utilise cette période pour recueillir des retours et améliorer le modèle.
+- MiMo V2 Pro Free est disponible sur OpenCode pour une durée limitée. L'équipe utilise cette période pour recueillir des retours et améliorer le modèle.
+- MiMo V2 Omni Free est disponible sur OpenCode pour une durée limitée. L'équipe utilise cette période pour recueillir des retours et améliorer le modèle.
+- Qwen3.6 Plus Free est disponible sur OpenCode pour une durée limitée. L'équipe utilise cette période pour recueillir des retours et améliorer le modèle.
+- Nemotron 3 Super Free est disponible sur OpenCode pour une durée limitée. L'équipe utilise cette période pour recueillir des retours et améliorer le modèle.
+- Big Pickle est un modèle stealth gratuit sur OpenCode pour une durée limitée. L'équipe utilise cette période pour recueillir des retours et améliorer le modèle.
 
 <a href={email}>Contactez-nous</a> si vous avez des questions.
 
@@ -166,7 +170,7 @@ Les modèles gratuits :
 
 ### Rechargement automatique
 
-Si votre solde descend en dessous de 5 $, Zen rechargera automatiquement 20 $.
+Si votre solde passe sous $5, Zen rechargera automatiquement $20.
 
 Vous pouvez modifier le montant du rechargement automatique. Vous pouvez également désactiver complètement le rechargement automatique.
 
@@ -176,7 +180,7 @@ Vous pouvez modifier le montant du rechargement automatique. Vous pouvez égalem
 
 Vous pouvez également définir une limite d'utilisation mensuelle pour l'ensemble de l'espace de travail et pour chaque membre de votre équipe.
 
-Par exemple, disons que vous définissez une limite d'utilisation mensuelle à 20 $, Zen n'utilisera pas plus de 20 $ par mois. Mais si le rechargement automatique est activé, Zen pourrait finir par vous facturant plus de 20 $ si votre solde descend en dessous de 5 $.
+Par exemple, si vous définissez une limite d'utilisation mensuelle à $20, Zen n'utilisera pas plus de $20 sur un mois. Mais si le rechargement automatique est activé, Zen peut finir par vous facturer plus de $20 si votre solde passe sous $5.
 
 ---
 
@@ -184,35 +188,40 @@ Par exemple, disons que vous définissez une limite d'utilisation mensuelle à 2
 
 | Modèle           | Date de dépréciation |
 | ---------------- | -------------------- |
-| Qwen3 Coder 480B | 6 février 2026       |
-| Kimi K2 Thinking | 6 mars 2026          |
-| Kimi K2          | 6 mars 2026          |
-| MiniMax M2.1     | 15 mars 2026         |
-| GLM 4.7          | 15 mars 2026         |
-| GLM 4.6          | 15 mars 2026         |
+| MiniMax M2.1     | March 15, 2026       |
+| GLM 4.7          | March 15, 2026       |
+| GLM 4.6          | March 15, 2026       |
+| Gemini 3 Pro     | March 9, 2026        |
+| Kimi K2 Thinking | March 6, 2026        |
+| Kimi K2          | March 6, 2026        |
+| Qwen3 Coder 480B | Feb 6, 2026          |
 
 ---
 
 ## Confidentialité
 
-Tous nos modèles sont hébergés aux États-Unis. Nos fournisseurs suivent une politique de rétention zéro et n'utilisent pas vos données pour la formation de modèles, avec les exceptions suivantes :
+Tous nos modèles sont hébergés aux US. Nos fournisseurs suivent une politique de rétention zéro et n'utilisent pas vos données pour l'entraînement des modèles, avec les exceptions suivantes :
 
 - Big Pickle : Pendant sa période gratuite, les données collectées peuvent être utilisées pour améliorer le modèle.
 - MiniMax M2.5 Free : Pendant sa période gratuite, les données collectées peuvent être utilisées pour améliorer le modèle.
-- API OpenAI : Les demandes sont conservées pendant 30 jours conformément aux politiques de données de [OpenAI](https://platform.openai.com/docs/guides/your-data).
-- API Anthropic : Les demandes sont conservées pendant 30 jours conformément aux [Politiques de données d'Anthropic](https://docs.anthropic.com/en/docs/claude-code/data-usage).
+- MiMo V2 Pro Free : Pendant sa période gratuite, les données collectées peuvent être utilisées pour améliorer le modèle.
+- MiMo V2 Omni Free : Pendant sa période gratuite, les données collectées peuvent être utilisées pour améliorer le modèle.
+- Qwen3.6 Plus Free : Pendant sa période gratuite, les données collectées peuvent être utilisées pour améliorer le modèle.
+- Nemotron 3 Super Free : Pendant sa période gratuite, les données collectées peuvent être utilisées pour améliorer le modèle.
+- OpenAI APIs : Les requêtes sont conservées pendant 30 jours conformément à [OpenAI's Data Policies](https://platform.openai.com/docs/guides/your-data).
+- Anthropic APIs : Les requêtes sont conservées pendant 30 jours conformément à [Anthropic's Data Policies](https://docs.anthropic.com/en/docs/claude-code/data-usage).
 
 ---
 
 ## Pour les équipes
 
-Zen fonctionne également très bien pour les équipes. Vous pouvez inviter des coéquipiers, attribuer des rôles, organiser les modèles utilisés par votre équipe, et bien plus encore.
+Zen fonctionne aussi très bien pour les équipes. Vous pouvez inviter des coéquipiers, attribuer des rôles, sélectionner les modèles que votre équipe utilise, et plus encore.
 
 :::note
 Les espaces de travail sont actuellement gratuits pour les équipes dans le cadre de la version bêta.
 :::
 
-La gestion de votre espace de travail est actuellement gratuite pour les équipes dans le cadre de la version bêta. Nous partagerons bientôt plus de détails sur les prix.
+La gestion de votre espace de travail est actuellement gratuite pour les équipes dans le cadre de la version bêta. Nous partagerons bientôt plus de détails sur la tarification.
 
 ---
 
@@ -220,28 +229,28 @@ La gestion de votre espace de travail est actuellement gratuite pour les équipe
 
 Vous pouvez inviter des coéquipiers dans votre espace de travail et attribuer des rôles :
 
-- **Administrateur** : gérez les modèles, les membres, les clés API et la facturation.
-- **Membre** : gérer uniquement ses propres clés API
+- **Admin** : Gérer les modèles, les membres, les clés API et la facturation
+- **Member** : Gérer uniquement ses propres clés API
 
 Les administrateurs peuvent également définir des limites de dépenses mensuelles pour chaque membre afin de garder les coûts sous contrôle.
 
 ---
 
-### Accès au modèle
+### Accès aux modèles
 
-Les administrateurs peuvent activer ou désactiver des modèles spécifiques pour l'espace de travail. Les requêtes adressées à un modèle désactivé renverront une erreur.
+Les administrateurs peuvent activer ou désactiver des modèles spécifiques pour l'espace de travail. Les requêtes effectuées vers un modèle désactivé renverront une erreur.
 
-Ceci est utile dans les cas où vous souhaitez désactiver l'utilisation d'un modèle qui collecte des données.
+Cela est utile si vous souhaitez désactiver l'utilisation d'un modèle qui collecte des données.
 
 ---
 
 ### Apportez votre propre clé
 
-Vous pouvez utiliser vos propres clés OpenAI ou Anthropic API tout en accédant à d'autres modèles dans Zen.
+Vous pouvez utiliser vos propres clés API OpenAI ou Anthropic tout en accédant à d'autres modèles dans Zen.
 
-Lorsque vous utilisez vos propres clés, les tokens sont facturés directement par le fournisseur et non par Zen.
+Lorsque vous utilisez vos propres clés, les tokens sont facturés directement par le fournisseur, pas par Zen.
 
-Par exemple, votre organisation dispose peut-être déjà d'une clé pour OpenAI ou Anthropic et vous souhaitez l'utiliser à la place de celui fourni par Zen.
+Par exemple, votre organisation a peut-être déjà une clé pour OpenAI ou Anthropic et vous souhaitez l'utiliser à la place de celle fournie par Zen.
 
 ---
 
@@ -249,7 +258,7 @@ Par exemple, votre organisation dispose peut-être déjà d'une clé pour OpenAI
 
 Nous avons créé OpenCode Zen pour :
 
-1. **Benchmark** les meilleurs modèles/fournisseurs d'agents de codage.
-2. Ayez accès aux options de **la plus haute qualité** et ne dégradez pas les performances ni ne vous dirigez vers des fournisseurs moins chers.
-3. Répercutez toute **baisse de prix** en vendant au prix coûtant ; la seule majoration est donc pour couvrir nos frais de traitement.
-4. N'ayez **aucun verrouillage** en vous permettant de l'utiliser avec n'importe quel autre agent de codage. Et laissez-vous toujours utiliser n'importe quel autre fournisseur avec OpenCode également.
+1. **Benchmark** les meilleurs modèles/fournisseurs pour les agents de codage.
+2. Avoir accès aux options de **la plus haute qualité** sans dégrader les performances ni basculer vers des fournisseurs moins chers.
+3. Répercuter toute **baisse de prix** en vendant au prix coûtant ; la seule marge sert à couvrir nos frais de traitement.
+4. N'avoir **aucun lock-in** en vous permettant de l'utiliser avec n'importe quel autre agent de codage. Et toujours vous permettre d'utiliser n'importe quel autre fournisseur avec OpenCode également.

packages/web/src/content/docs/it/zen.mdx

@@ -7,31 +7,39 @@ import config from "../../../../config.mjs"
 export const console = config.console
 export const email = `mailto:${config.email}`
 
-OpenCode Zen è una lista di modelli testati e verificati dal team di OpenCode.
+OpenCode Zen è un elenco di modelli testati e verificati forniti dal team di OpenCode.
 
 :::note
 OpenCode Zen è attualmente in beta.
 :::
 
-Zen funziona come qualunque altro provider in OpenCode. Accedi a OpenCode Zen e ottieni la tua chiave API. È **completamente opzionale** e non devi usarlo per usare OpenCode.
+Zen funziona come qualsiasi altro provider in OpenCode. Accedi a OpenCode Zen e ottieni
+la tua chiave API. È **completamente opzionale** e non ti serve usarlo per usare
+OpenCode.
 
 ---
 
 ## Contesto
 
-Ci sono moltissimi modelli, ma solo pochi funzionano bene come agenti di coding. Inoltre, la maggior parte dei provider è configurata in modo molto diverso, quindi prestazioni e qualità possono variare parecchio.
+Esiste un gran numero di modelli, ma solo pochi di
+questi modelli funzionano bene come agenti di coding. Inoltre, la maggior parte dei provider è
+configurata in modo molto diverso; quindi ottieni prestazioni e qualità molto diverse.
 
 :::tip
 Abbiamo testato un gruppo selezionato di modelli e provider che funzionano bene con OpenCode.
 :::
 
-Quindi, se usi un modello tramite qualcosa come OpenRouter, non puoi mai essere sicuro di ottenere la migliore versione del modello che vuoi.
+Quindi, se usi un modello tramite qualcosa come OpenRouter, non puoi mai essere
+sicuro di ottenere la migliore versione del modello che vuoi.
 
-Per risolvere, abbiamo fatto alcune cose:
+Per risolvere questo problema, abbiamo fatto un paio di cose:
 
-1. Abbiamo testato un gruppo selezionato di modelli e parlato coi loro team su come eseguirli al meglio.
-2. Poi abbiamo lavorato con alcuni provider per assicurarci che venissero serviti correttamente.
-3. Infine, abbiamo fatto benchmark delle combinazioni modello/provider e creato una lista che ci sentiamo di raccomandare.
+1. Abbiamo testato un gruppo selezionato di modelli e parlato con i loro team di come
+   farli funzionare al meglio.
+2. Poi abbiamo collaborato con alcuni provider per assicurarci che fossero serviti
+   correttamente.
+3. Infine, abbiamo fatto benchmark della combinazione modello/provider e creato
+   un elenco che ci sentiamo di raccomandare.
 
 OpenCode Zen è un gateway AI che ti dà accesso a questi modelli.
 
@@ -39,13 +47,14 @@ OpenCode Zen è un gateway AI che ti dà accesso a questi modelli.
 
 ## Come funziona
 
-OpenCode Zen funziona come qualunque altro provider in OpenCode.
+OpenCode Zen funziona come qualsiasi altro provider in OpenCode.
 
-1. Accedi a **<a href={console}>OpenCode Zen</a>**, aggiungi i dettagli di fatturazione e copia la chiave API.
-2. Esegui il comando `/connect` nella TUI, seleziona OpenCode Zen e incolla la chiave API.
-3. Esegui `/models` nella TUI per vedere la lista dei modelli che consigliamo.
+1. Accedi a **<a href={console}>OpenCode Zen</a>**, aggiungi i dettagli di fatturazione,
+   e copia la tua chiave API.
+2. Esegui il comando `/connect` nella TUI, seleziona OpenCode Zen, e incolla la tua chiave API.
+3. Esegui `/models` nella TUI per vedere l'elenco dei modelli che raccomandiamo.
 
-Paghi per richiesta e puoi aggiungere credito al tuo account.
+Ti viene addebitato ogni richiesta e puoi aggiungere credito al tuo account.
 
 ---
 
@@ -53,7 +62,7 @@ Paghi per richiesta e puoi aggiungere credito al tuo account.
 
 Puoi anche accedere ai nostri modelli tramite i seguenti endpoint API.
 
-| Modello               | ID modello            | Endpoint                                           | Pacchetto AI SDK            |
+| Modello               | Model ID              | Endpoint                                           | Pacchetto AI SDK            |
 | --------------------- | --------------------- | -------------------------------------------------- | --------------------------- |
 | GPT 5.4               | gpt-5.4               | `https://opencode.ai/zen/v1/responses`             | `@ai-sdk/openai`            |
 | GPT 5.4 Pro           | gpt-5.4-pro           | `https://opencode.ai/zen/v1/responses`             | `@ai-sdk/openai`            |
@@ -85,16 +94,20 @@ Puoi anche accedere ai nostri modelli tramite i seguenti endpoint API.
 | GLM 5                 | glm-5                 | `https://opencode.ai/zen/v1/chat/completions`      | `@ai-sdk/openai-compatible` |
 | Kimi K2.5             | kimi-k2.5             | `https://opencode.ai/zen/v1/chat/completions`      | `@ai-sdk/openai-compatible` |
 | Big Pickle            | big-pickle            | `https://opencode.ai/zen/v1/chat/completions`      | `@ai-sdk/openai-compatible` |
-| MiMo V2 Flash Free    | mimo-v2-flash-free    | `https://opencode.ai/zen/v1/chat/completions`      | `@ai-sdk/openai-compatible` |
+| MiMo V2 Pro Free      | mimo-v2-pro-free      | `https://opencode.ai/zen/v1/chat/completions`      | `@ai-sdk/openai-compatible` |
+| MiMo V2 Omni Free     | mimo-v2-omni-free     | `https://opencode.ai/zen/v1/chat/completions`      | `@ai-sdk/openai-compatible` |
+| Qwen3.6 Plus Free     | qwen3.6-plus-free     | `https://opencode.ai/zen/v1/chat/completions`      | `@ai-sdk/openai-compatible` |
 | Nemotron 3 Super Free | nemotron-3-super-free | `https://opencode.ai/zen/v1/chat/completions`      | `@ai-sdk/openai-compatible` |
 
-Il [model id](/docs/config/#models) nella config di OpenCode usa il formato `opencode/<model-id>`. Per esempio, per GPT 5.3 Codex, useresti `opencode/gpt-5.3-codex` nella config.
+Il [model id](/docs/config/#models) nella config di OpenCode
+usa il formato `opencode/<model-id>`. Per esempio, per GPT 5.3 Codex, useresti
+`opencode/gpt-5.3-codex` nella tua config.
 
 ---
 
 ### Modelli
 
-Puoi recuperare la lista completa dei modelli disponibili e i relativi metadati da:
+Puoi recuperare l'elenco completo dei modelli disponibili e i relativi metadati da:
 
 ```
 https://opencode.ai/zen/v1/models
@@ -106,57 +119,61 @@ https://opencode.ai/zen/v1/models
 
 Supportiamo un modello pay-as-you-go. Qui sotto trovi i prezzi **per 1M token**.
 
-| Modello                           | Input  | Output  | Lettura in cache | Scrittura in cache |
-| --------------------------------- | ------ | ------- | ---------------- | ------------------ |
-| Big Pickle                        | Gratis | Gratis  | Gratis           | -                  |
-| MiMo V2 Flash Free                | Gratis | Gratis  | Gratis           | -                  |
-| Nemotron 3 Super Free             | Gratis | Gratis  | Gratis           | -                  |
-| MiniMax M2.5 Free                 | Gratis | Gratis  | Gratis           | -                  |
-| MiniMax M2.5                      | $0.30  | $1.20   | $0.06            | $0.375             |
-| GLM 5                             | $1.00  | $3.20   | $0.20            | -                  |
-| Kimi K2.5                         | $0.60  | $3.00   | $0.10            | -                  |
-| Qwen3 Coder 480B                  | $0.45  | $1.50   | -                | -                  |
-| Claude Opus 4.6                   | $5.00  | $25.00  | $0.50            | $6.25              |
-| Claude Opus 4.5                   | $5.00  | $25.00  | $0.50            | $6.25              |
-| Claude Opus 4.1                   | $15.00 | $75.00  | $1.50            | $18.75             |
-| Claude Sonnet 4.6                 | $3.00  | $15.00  | $0.30            | $3.75              |
-| Claude Sonnet 4.5 (≤ 200K tokens) | $3.00  | $15.00  | $0.30            | $3.75              |
-| Claude Sonnet 4.5 (> 200K tokens) | $6.00  | $22.50  | $0.60            | $7.50              |
-| Claude Sonnet 4 (≤ 200K tokens)   | $3.00  | $15.00  | $0.30            | $3.75              |
-| Claude Sonnet 4 (> 200K tokens)   | $6.00  | $22.50  | $0.60            | $7.50              |
-| Claude Haiku 4.5                  | $1.00  | $5.00   | $0.10            | $1.25              |
-| Claude Haiku 3.5                  | $0.80  | $4.00   | $0.08            | $1.00              |
-| Gemini 3.1 Pro (≤ 200K tokens)    | $2.00  | $12.00  | $0.20            | -                  |
-| Gemini 3.1 Pro (> 200K tokens)    | $4.00  | $18.00  | $0.40            | -                  |
-| Gemini 3 Flash                    | $0.50  | $3.00   | $0.05            | -                  |
-| GPT 5.4                           | $2.50  | $15.00  | $0.25            | -                  |
-| GPT 5.4 Pro                       | $30.00 | $180.00 | $30.00           | -                  |
-| GPT 5.4 Mini                      | $0.75  | $4.50   | $0.075           | -                  |
-| GPT 5.4 Nano                      | $0.20  | $1.25   | $0.02            | -                  |
-| GPT 5.3 Codex Spark               | $1.75  | $14.00  | $0.175           | -                  |
-| GPT 5.3 Codex                     | $1.75  | $14.00  | $0.175           | -                  |
-| GPT 5.2                           | $1.75  | $14.00  | $0.175           | -                  |
-| GPT 5.2 Codex                     | $1.75  | $14.00  | $0.175           | -                  |
-| GPT 5.1                           | $1.07  | $8.50   | $0.107           | -                  |
-| GPT 5.1 Codex                     | $1.07  | $8.50   | $0.107           | -                  |
-| GPT 5.1 Codex Max                 | $1.25  | $10.00  | $0.125           | -                  |
-| GPT 5.1 Codex Mini                | $0.25  | $2.00   | $0.025           | -                  |
-| GPT 5                             | $1.07  | $8.50   | $0.107           | -                  |
-| GPT 5 Codex                       | $1.07  | $8.50   | $0.107           | -                  |
-| GPT 5 Nano                        | Gratis | Gratis  | Gratis           | -                  |
+| Modello                           | Input  | Output  | Cached Read | Cached Write |
+| --------------------------------- | ------ | ------- | ----------- | ------------ |
+| Big Pickle                        | Free   | Free    | Free        | -            |
+| MiMo V2 Pro Free                  | Free   | Free    | Free        | -            |
+| MiMo V2 Omni Free                 | Free   | Free    | Free        | -            |
+| Qwen3.6 Plus Free                 | Free   | Free    | Free        | -            |
+| Nemotron 3 Super Free             | Free   | Free    | Free        | -            |
+| MiniMax M2.5 Free                 | Free   | Free    | Free        | -            |
+| MiniMax M2.5                      | $0.30  | $1.20   | $0.06       | $0.375       |
+| GLM 5                             | $1.00  | $3.20   | $0.20       | -            |
+| Kimi K2.5                         | $0.60  | $3.00   | $0.10       | -            |
+| Qwen3 Coder 480B                  | $0.45  | $1.50   | -           | -            |
+| Claude Opus 4.6                   | $5.00  | $25.00  | $0.50       | $6.25        |
+| Claude Opus 4.5                   | $5.00  | $25.00  | $0.50       | $6.25        |
+| Claude Opus 4.1                   | $15.00 | $75.00  | $1.50       | $18.75       |
+| Claude Sonnet 4.6                 | $3.00  | $15.00  | $0.30       | $3.75        |
+| Claude Sonnet 4.5 (≤ 200K tokens) | $3.00  | $15.00  | $0.30       | $3.75        |
+| Claude Sonnet 4.5 (> 200K tokens) | $6.00  | $22.50  | $0.60       | $7.50        |
+| Claude Sonnet 4 (≤ 200K tokens)   | $3.00  | $15.00  | $0.30       | $3.75        |
+| Claude Sonnet 4 (> 200K tokens)   | $6.00  | $22.50  | $0.60       | $7.50        |
+| Claude Haiku 4.5                  | $1.00  | $5.00   | $0.10       | $1.25        |
+| Claude Haiku 3.5                  | $0.80  | $4.00   | $0.08       | $1.00        |
+| Gemini 3.1 Pro (≤ 200K tokens)    | $2.00  | $12.00  | $0.20       | -            |
+| Gemini 3.1 Pro (> 200K tokens)    | $4.00  | $18.00  | $0.40       | -            |
+| Gemini 3 Flash                    | $0.50  | $3.00   | $0.05       | -            |
+| GPT 5.4                           | $2.50  | $15.00  | $0.25       | -            |
+| GPT 5.4 Pro                       | $30.00 | $180.00 | $30.00      | -            |
+| GPT 5.4 Mini                      | $0.75  | $4.50   | $0.075      | -            |
+| GPT 5.4 Nano                      | $0.20  | $1.25   | $0.02       | -            |
+| GPT 5.3 Codex Spark               | $1.75  | $14.00  | $0.175      | -            |
+| GPT 5.3 Codex                     | $1.75  | $14.00  | $0.175      | -            |
+| GPT 5.2                           | $1.75  | $14.00  | $0.175      | -            |
+| GPT 5.2 Codex                     | $1.75  | $14.00  | $0.175      | -            |
+| GPT 5.1                           | $1.07  | $8.50   | $0.107      | -            |
+| GPT 5.1 Codex                     | $1.07  | $8.50   | $0.107      | -            |
+| GPT 5.1 Codex Max                 | $1.25  | $10.00  | $0.125      | -            |
+| GPT 5.1 Codex Mini                | $0.25  | $2.00   | $0.025      | -            |
+| GPT 5                             | $1.07  | $8.50   | $0.107      | -            |
+| GPT 5 Codex                       | $1.07  | $8.50   | $0.107      | -            |
+| GPT 5 Nano                        | Free   | Free    | Free        | -            |
 
-Potresti notare _Claude Haiku 3.5_ nella cronologia d'uso. È un [modello a basso costo](/docs/config/#models) usato per generare i titoli delle sessioni.
+Potresti notare _Claude Haiku 3.5_ nella cronologia di utilizzo. È un [modello a basso costo](/docs/config/#models) usato per generare i titoli delle tue sessioni.
 
 :::note
-Le commissioni della carta di credito vengono ribaltate al costo (4.4% + $0.30 per transazione); non addebitiamo nulla oltre a questo.
+Le commissioni della carta di credito vengono trasferite al costo (4.4% + $0.30 per transazione); non addebitiamo nulla oltre a questo.
 :::
 
 I modelli gratuiti:
 
-- MiniMax M2.5 Free è disponibile su OpenCode per un periodo limitato. Il team usa questo tempo per raccogliere feedback e migliorare il modello.
-- MiMo V2 Flash Free è disponibile su OpenCode per un periodo limitato. Il team usa questo tempo per raccogliere feedback e migliorare il modello.
-- Nemotron 3 Super Free è disponibile su OpenCode per un periodo limitato. Il team usa questo tempo per raccogliere feedback e migliorare il modello.
-- Big Pickle è un modello stealth gratuito su OpenCode per un periodo limitato. Il team usa questo tempo per raccogliere feedback e migliorare il modello.
+- MiniMax M2.5 Free è disponibile su OpenCode per un periodo limitato. Il team usa questo periodo per raccogliere feedback e migliorare il modello.
+- MiMo V2 Pro Free è disponibile su OpenCode per un periodo limitato. Il team usa questo periodo per raccogliere feedback e migliorare il modello.
+- MiMo V2 Omni Free è disponibile su OpenCode per un periodo limitato. Il team usa questo periodo per raccogliere feedback e migliorare il modello.
+- Qwen3.6 Plus Free è disponibile su OpenCode per un periodo limitato. Il team usa questo periodo per raccogliere feedback e migliorare il modello.
+- Nemotron 3 Super Free è disponibile su OpenCode per un periodo limitato. Il team usa questo periodo per raccogliere feedback e migliorare il modello.
+- Big Pickle è un modello stealth che è gratuito su OpenCode per un periodo limitato. Il team usa questo periodo per raccogliere feedback e migliorare il modello.
 
 <a href={email}>Contattaci</a> se hai domande.
 
@@ -164,17 +181,20 @@ I modelli gratuiti:
 
 ### Ricarica automatica
 
-Se il saldo scende sotto $5, Zen ricarica automaticamente $20.
+Se il tuo saldo scende sotto $5, Zen ricarica automaticamente $20.
 
-Puoi cambiare l'importo della ricarica automatica. Puoi anche disabilitare completamente l'auto-reload.
+Puoi modificare l'importo della ricarica automatica. Puoi anche disabilitare completamente la ricarica automatica.
 
 ---
 
 ### Limiti mensili
 
-Puoi anche impostare un limite mensile di utilizzo per l'intero workspace e per ogni membro del team.
+Puoi anche impostare un limite di utilizzo mensile per l'intero workspace e per ogni
+membro del tuo team.
 
-Per esempio, se imposti un limite mensile a $20, Zen non userà più di $20 in un mese. Ma se hai l'auto-reload attivo, Zen potrebbe finire per addebitarti più di $20 se il saldo scende sotto $5.
+Per esempio, supponiamo che tu imposti un limite di utilizzo mensile di $20, Zen non userà
+più di $20 in un mese. Ma se hai la ricarica automatica abilitata, Zen potrebbe finire
+per addebitarti più di $20 se il tuo saldo scende sotto $5.
 
 ---
 
@@ -182,67 +202,73 @@ Per esempio, se imposti un limite mensile a $20, Zen non userà più di $20 in u
 
 | Modello          | Data di deprecazione |
 | ---------------- | -------------------- |
-| MiniMax M2.1     | 15 mar 2026          |
-| GLM 4.7          | 15 mar 2026          |
-| GLM 4.6          | 15 mar 2026          |
-| Gemini 3 Pro     | 9 mar 2026           |
-| Kimi K2 Thinking | 6 mar 2026           |
-| Kimi K2          | 6 mar 2026           |
-| Qwen3 Coder 480B | 6 feb 2026           |
+| MiniMax M2.1     | March 15, 2026       |
+| GLM 4.7          | March 15, 2026       |
+| GLM 4.6          | March 15, 2026       |
+| Gemini 3 Pro     | March 9, 2026        |
+| Kimi K2 Thinking | March 6, 2026        |
+| Kimi K2          | March 6, 2026        |
+| Qwen3 Coder 480B | Feb 6, 2026          |
 
 ---
 
 ## Privacy
 
-Tutti i nostri modelli sono ospitati negli US. I nostri provider seguono una policy di zero-retention e non usano i tuoi dati per training dei modelli, con le seguenti eccezioni:
+Tutti i nostri modelli sono ospitati negli US. I nostri provider seguono una policy di zero-retention e non usano i tuoi dati per l'addestramento dei modelli, con le seguenti eccezioni:
 
-- Big Pickle: durante il periodo gratuito, i dati raccolti potrebbero essere usati per migliorare il modello.
-- MiniMax M2.5 Free: durante il periodo gratuito, i dati raccolti potrebbero essere usati per migliorare il modello.
-- MiniMax M2.5 Free: durante il periodo gratuito, i dati raccolti potrebbero essere usati per migliorare il modello.
-- Nemotron 3 Super Free: durante il periodo gratuito, i dati raccolti potrebbero essere usati per migliorare il modello.
-- OpenAI APIs: le richieste vengono conservate per 30 giorni in conformità alle [OpenAI's Data Policies](https://platform.openai.com/docs/guides/your-data).
-- Anthropic APIs: le richieste vengono conservate per 30 giorni in conformità alle [Anthropic's Data Policies](https://docs.anthropic.com/en/docs/claude-code/data-usage).
+- Big Pickle: durante il periodo gratuito, i dati raccolti possono essere usati per migliorare il modello.
+- MiniMax M2.5 Free: durante il periodo gratuito, i dati raccolti possono essere usati per migliorare il modello.
+- MiMo V2 Pro Free: durante il periodo gratuito, i dati raccolti possono essere usati per migliorare il modello.
+- MiMo V2 Omni Free: durante il periodo gratuito, i dati raccolti possono essere usati per migliorare il modello.
+- Qwen3.6 Plus Free: durante il periodo gratuito, i dati raccolti possono essere usati per migliorare il modello.
+- Nemotron 3 Super Free: durante il periodo gratuito, i dati raccolti possono essere usati per migliorare il modello.
+- OpenAI APIs: le richieste vengono conservate per 30 giorni in conformità con [OpenAI's Data Policies](https://platform.openai.com/docs/guides/your-data).
+- Anthropic APIs: le richieste vengono conservate per 30 giorni in conformità con [Anthropic's Data Policies](https://docs.anthropic.com/en/docs/claude-code/data-usage).
 
 ---
 
 ## Per i team
 
-Zen funziona benissimo anche per i team. Puoi invitare colleghi, assegnare ruoli, curare i modelli usati dal team e altro.
+Zen funziona benissimo anche per i team. Puoi invitare compagni di squadra, assegnare ruoli, curare
+i modelli che il tuo team usa, e altro ancora.
 
 :::note
 I workspace sono attualmente gratuiti per i team come parte della beta.
 :::
 
-Gestire il workspace è attualmente gratuito per i team come parte della beta. Condivideremo presto più dettagli sul pricing.
+La gestione del tuo workspace è attualmente gratuita per i team come parte della beta. Condivideremo presto
+più dettagli sui prezzi.
 
 ---
 
 ### Ruoli
 
-Puoi invitare colleghi nel workspace e assegnare ruoli:
+Puoi invitare compagni di squadra nel tuo workspace e assegnare ruoli:
 
-- **Admin**: gestisce modelli, membri, chiavi API e fatturazione
-- **Member**: gestisce solo le proprie chiavi API
+- **Admin**: Gestisce modelli, membri, chiavi API, e fatturazione
+- **Member**: Gestisce solo le proprie chiavi API
 
-Gli admin possono anche impostare limiti mensili di spesa per ogni membro per tenere i costi sotto controllo.
+Gli admin possono anche impostare limiti di spesa mensili per ogni membro per tenere i costi sotto controllo.
 
 ---
 
 ### Accesso ai modelli
 
-Gli admin possono abilitare o disabilitare modelli specifici per il workspace. Le richieste verso un modello disabilitato restituiscono un errore.
+Gli admin possono abilitare o disabilitare modelli specifici per il workspace. Le richieste fatte a un modello disabilitato restituiranno un errore.
 
-Questo è utile quando vuoi disabilitare l'uso di un modello che raccoglie dati.
+Questo è utile nei casi in cui vuoi disabilitare l'uso di un modello che
+raccoglie dati.
 
 ---
 
-### Usa le tue chiavi
+### Porta la tua chiave
 
 Puoi usare le tue chiavi API OpenAI o Anthropic continuando ad accedere agli altri modelli in Zen.
 
 Quando usi le tue chiavi, i token vengono fatturati direttamente dal provider, non da Zen.
 
-Per esempio, la tua organizzazione potrebbe avere già una chiave per OpenAI o Anthropic e vuoi usare quella invece di quella fornita da Zen.
+Per esempio, la tua organizzazione potrebbe avere già una chiave per OpenAI o Anthropic
+e vuoi usare quella invece di quella fornita da Zen.
 
 ---
 
@@ -251,6 +277,6 @@ Per esempio, la tua organizzazione potrebbe avere già una chiave per OpenAI o A
 Abbiamo creato OpenCode Zen per:
 
 1. Fare **benchmark** dei migliori modelli/provider per agenti di coding.
-2. Dare accesso alle opzioni di **massima qualità** senza ridurre le prestazioni o instradare verso provider più economici.
-3. Trasferire eventuali **riduzioni di prezzo** vendendo al costo; l'unico markup copre le commissioni di elaborazione.
-4. Evitare **lock-in** permettendoti di usarlo con qualunque altro agente di coding e lasciandoti sempre usare anche altri provider con OpenCode.
+2. Avere accesso alle opzioni di **massima qualità** e non degradare le prestazioni né instradare verso provider più economici.
+3. Trasferire eventuali **riduzioni di prezzo** vendendo al costo; quindi l'unico markup serve a coprire le nostre commissioni di elaborazione.
+4. Non avere **lock-in** permettendoti di usarlo con qualsiasi altro agente di coding. E permettendoti sempre di usare anche qualsiasi altro provider con OpenCode.

packages/web/src/content/docs/ja/zen.mdx

@@ -7,96 +7,96 @@ import config from "../../../../config.mjs"
 export const console = config.console
 export const email = `mailto:${config.email}`
 
-OpenCode Zen は、OpenCode チームによって提供される、テストおよび検証されたモデルのリストです。
+OpenCode Zen は、OpenCode チームが提供する、テスト済みかつ検証済みのモデルの一覧です。
 
 :::note
 OpenCode Zen は現在ベータ版です。
 :::
-Zen は OpenCode の他のプロバイダーと同様に機能します。 OpenCode Zen にログインすると、API キーを取得できます。これは **完全にオプション** であり、OpenCode を使用するために使用する必要はありません。
+
+Zen は OpenCode のほかのプロバイダーと同じように動作します。OpenCode Zen にログインして API キーを取得できます。これは **完全にオプション** であり、OpenCode を使うために必須ではありません。
 
 ---
 
 ## 背景
 
-世の中には多数のモデルがありますが、コーディングエージェントとしてうまく機能するのはごく一部です。さらに、ほとんどのプロバイダーは構成が大きく異なるため、パフォーマンスと品質も大きく異なります。
+世の中には非常に多くのモデルがありますが、コーディングエージェントとして十分に機能するものはごく一部です。さらに、ほとんどのプロバイダーは設定が大きく異なるため、得られるパフォーマンスや品質にも大きな差が出ます。
 
 :::tip
-OpenCode で適切に動作する、厳選されたモデルとプロバイダーのグループをテストしました。
+OpenCode でうまく動作する、厳選したモデルとプロバイダーの組み合わせをテストしました。
 :::
 
-OpenRouter などを通じてモデルを使用している場合、必要なモデルの最高のバージョンを取得できているか確信が持てません。
+そのため、OpenRouter のようなものを通じてモデルを使っている場合、欲しいモデルの最適なバージョンを本当に使えているかを判断するのは困難です。
 
-これを修正するために、いくつかのことを行いました。
+これを解決するために、私たちはいくつかのことを行いました。
 
-1. 選抜したモデルグループをテストし、それらを最適に実行する方法についてチームと話し合いました。
-2. その後、いくつかのプロバイダーと協力して、これらが確実に提供されるようにしました。
-3. 最後に、モデルとプロバイダーの組み合わせをベンチマークし、自信を持ってお勧めできるリストを作成しました。
+1. 厳選したモデル群をテストし、それぞれのチームと、最適な動かし方について話し合いました。
+2. そのうえで、いくつかのプロバイダーと連携し、それらが正しく提供されるようにしました。
+3. 最後に、モデルとプロバイダーの組み合わせをベンチマークし、自信を持っておすすめできる一覧を作成しました。
 
-OpenCode Zen は、これらのモデルへのアクセスを可能にする AI ゲートウェイです。
+OpenCode Zen は、これらのモデルにアクセスできる AI ゲートウェイです。
 
 ---
 
 ## 仕組み
 
-OpenCode Zen は、OpenCode の他のプロバイダーと同様に機能します。
+OpenCode Zen は、OpenCode のほかのプロバイダーと同じように動作します。
 
 1. **<a href={console}>OpenCode Zen</a>** にサインインし、請求情報を追加して、API キーをコピーします。
-2. TUI で `/connect` コマンドを実行し、OpenCode Zen を選択して API キーを貼り付けます。
-3. TUI で `/models` を実行すると、推奨されるモデルのリストが表示されます。
+2. TUI で `/connect` コマンドを実行し、OpenCode Zen を選んで API キーを貼り付けます。
+3. TUI で `/models` を実行すると、私たちがおすすめするモデルの一覧を確認できます。
 
-リクエストごとに料金が請求され、アカウントにクレジットを追加できます。
+料金はリクエストごとに発生し、アカウントにクレジットを追加できます。
 
 ---
 
 ## エンドポイント
 
-次の API エンドポイントを通じてモデルにアクセスすることもできます。
+以下の API エンドポイントを通じて、私たちのモデルにアクセスすることもできます。
 
-| Model              | Model ID           | Endpoint                                           | AI SDK Package              |
-| ------------------ | ------------------ | -------------------------------------------------- | --------------------------- |
-| GPT 5.4            | gpt-5.4            | `https://opencode.ai/zen/v1/responses`             | `@ai-sdk/openai`            |
-| GPT 5.3 Codex      | gpt-5.3-codex      | `https://opencode.ai/zen/v1/responses`             | `@ai-sdk/openai`            |
-| GPT 5.2            | gpt-5.2            | `https://opencode.ai/zen/v1/responses`             | `@ai-sdk/openai`            |
-| GPT 5.2 Codex      | gpt-5.2-codex      | `https://opencode.ai/zen/v1/responses`             | `@ai-sdk/openai`            |
-| GPT 5.1            | gpt-5.1            | `https://opencode.ai/zen/v1/responses`             | `@ai-sdk/openai`            |
-| GPT 5.1 Codex      | gpt-5.1-codex      | `https://opencode.ai/zen/v1/responses`             | `@ai-sdk/openai`            |
-| GPT 5.1 Codex Max  | gpt-5.1-codex-max  | `https://opencode.ai/zen/v1/responses`             | `@ai-sdk/openai`            |
-| GPT 5.1 Codex Mini | gpt-5.1-codex-mini | `https://opencode.ai/zen/v1/responses`             | `@ai-sdk/openai`            |
-| GPT 5              | gpt-5              | `https://opencode.ai/zen/v1/responses`             | `@ai-sdk/openai`            |
-| GPT 5 Codex        | gpt-5-codex        | `https://opencode.ai/zen/v1/responses`             | `@ai-sdk/openai`            |
-| GPT 5 Nano         | gpt-5-nano         | `https://opencode.ai/zen/v1/responses`             | `@ai-sdk/openai`            |
-| Claude Opus 4.6    | claude-opus-4-6    | `https://opencode.ai/zen/v1/messages`              | `@ai-sdk/anthropic`         |
-| Claude Opus 4.5    | claude-opus-4-5    | `https://opencode.ai/zen/v1/messages`              | `@ai-sdk/anthropic`         |
-| Claude Opus 4.1    | claude-opus-4-1    | `https://opencode.ai/zen/v1/messages`              | `@ai-sdk/anthropic`         |
-| Claude Sonnet 4.6  | claude-sonnet-4-6  | `https://opencode.ai/zen/v1/messages`              | `@ai-sdk/anthropic`         |
-| Claude Sonnet 4.5  | claude-sonnet-4-5  | `https://opencode.ai/zen/v1/messages`              | `@ai-sdk/anthropic`         |
-| Claude Sonnet 4    | claude-sonnet-4    | `https://opencode.ai/zen/v1/messages`              | `@ai-sdk/anthropic`         |
-| Claude Haiku 4.5   | claude-haiku-4-5   | `https://opencode.ai/zen/v1/messages`              | `@ai-sdk/anthropic`         |
-| Claude Haiku 3.5   | claude-3-5-haiku   | `https://opencode.ai/zen/v1/messages`              | `@ai-sdk/anthropic`         |
-| Gemini 3.1 Pro     | gemini-3.1-pro     | `https://opencode.ai/zen/v1/models/gemini-3.1-pro` | `@ai-sdk/google`            |
-| Gemini 3 Pro       | gemini-3-pro       | `https://opencode.ai/zen/v1/models/gemini-3-pro`   | `@ai-sdk/google`            |
-| Gemini 3 Flash     | gemini-3-flash     | `https://opencode.ai/zen/v1/models/gemini-3-flash` | `@ai-sdk/google`            |
-| MiniMax M2.5       | minimax-m2.5       | `https://opencode.ai/zen/v1/chat/completions`      | `@ai-sdk/openai-compatible` |
-| MiniMax M2.5 Free  | minimax-m2.5-free  | `https://opencode.ai/zen/v1/chat/completions`      | `@ai-sdk/openai-compatible` |
-| MiniMax M2.1       | minimax-m2.1       | `https://opencode.ai/zen/v1/chat/completions`      | `@ai-sdk/openai-compatible` |
-| GLM 5              | glm-5              | `https://opencode.ai/zen/v1/chat/completions`      | `@ai-sdk/openai-compatible` |
-| GLM 4.7            | glm-4.7            | `https://opencode.ai/zen/v1/chat/completions`      | `@ai-sdk/openai-compatible` |
-| GLM 4.6            | glm-4.6            | `https://opencode.ai/zen/v1/chat/completions`      | `@ai-sdk/openai-compatible` |
-| Kimi K2.5          | kimi-k2.5          | `https://opencode.ai/zen/v1/chat/completions`      | `@ai-sdk/openai-compatible` |
-| Kimi K2 Thinking   | kimi-k2-thinking   | `https://opencode.ai/zen/v1/chat/completions`      | `@ai-sdk/openai-compatible` |
-| Kimi K2            | kimi-k2            | `https://opencode.ai/zen/v1/chat/completions`      | `@ai-sdk/openai-compatible` |
-| Qwen3 Coder 480B   | qwen3-coder        | `https://opencode.ai/zen/v1/chat/completions`      | `@ai-sdk/openai-compatible` |
-| Big Pickle         | big-pickle         | `https://opencode.ai/zen/v1/chat/completions`      | `@ai-sdk/openai-compatible` |
+| Model                 | Model ID              | Endpoint                                           | AI SDK Package              |
+| --------------------- | --------------------- | -------------------------------------------------- | --------------------------- |
+| GPT 5.4               | gpt-5.4               | `https://opencode.ai/zen/v1/responses`             | `@ai-sdk/openai`            |
+| GPT 5.4 Pro           | gpt-5.4-pro           | `https://opencode.ai/zen/v1/responses`             | `@ai-sdk/openai`            |
+| GPT 5.4 Mini          | gpt-5.4-mini          | `https://opencode.ai/zen/v1/responses`             | `@ai-sdk/openai`            |
+| GPT 5.4 Nano          | gpt-5.4-nano          | `https://opencode.ai/zen/v1/responses`             | `@ai-sdk/openai`            |
+| GPT 5.3 Codex         | gpt-5.3-codex         | `https://opencode.ai/zen/v1/responses`             | `@ai-sdk/openai`            |
+| GPT 5.3 Codex Spark   | gpt-5.3-codex-spark   | `https://opencode.ai/zen/v1/responses`             | `@ai-sdk/openai`            |
+| GPT 5.2               | gpt-5.2               | `https://opencode.ai/zen/v1/responses`             | `@ai-sdk/openai`            |
+| GPT 5.2 Codex         | gpt-5.2-codex         | `https://opencode.ai/zen/v1/responses`             | `@ai-sdk/openai`            |
+| GPT 5.1               | gpt-5.1               | `https://opencode.ai/zen/v1/responses`             | `@ai-sdk/openai`            |
+| GPT 5.1 Codex         | gpt-5.1-codex         | `https://opencode.ai/zen/v1/responses`             | `@ai-sdk/openai`            |
+| GPT 5.1 Codex Max     | gpt-5.1-codex-max     | `https://opencode.ai/zen/v1/responses`             | `@ai-sdk/openai`            |
+| GPT 5.1 Codex Mini    | gpt-5.1-codex-mini    | `https://opencode.ai/zen/v1/responses`             | `@ai-sdk/openai`            |
+| GPT 5                 | gpt-5                 | `https://opencode.ai/zen/v1/responses`             | `@ai-sdk/openai`            |
+| GPT 5 Codex           | gpt-5-codex           | `https://opencode.ai/zen/v1/responses`             | `@ai-sdk/openai`            |
+| GPT 5 Nano            | gpt-5-nano            | `https://opencode.ai/zen/v1/responses`             | `@ai-sdk/openai`            |
+| Claude Opus 4.6       | claude-opus-4-6       | `https://opencode.ai/zen/v1/messages`              | `@ai-sdk/anthropic`         |
+| Claude Opus 4.5       | claude-opus-4-5       | `https://opencode.ai/zen/v1/messages`              | `@ai-sdk/anthropic`         |
+| Claude Opus 4.1       | claude-opus-4-1       | `https://opencode.ai/zen/v1/messages`              | `@ai-sdk/anthropic`         |
+| Claude Sonnet 4.6     | claude-sonnet-4-6     | `https://opencode.ai/zen/v1/messages`              | `@ai-sdk/anthropic`         |
+| Claude Sonnet 4.5     | claude-sonnet-4-5     | `https://opencode.ai/zen/v1/messages`              | `@ai-sdk/anthropic`         |
+| Claude Sonnet 4       | claude-sonnet-4       | `https://opencode.ai/zen/v1/messages`              | `@ai-sdk/anthropic`         |
+| Claude Haiku 4.5      | claude-haiku-4-5      | `https://opencode.ai/zen/v1/messages`              | `@ai-sdk/anthropic`         |
+| Claude Haiku 3.5      | claude-3-5-haiku      | `https://opencode.ai/zen/v1/messages`              | `@ai-sdk/anthropic`         |
+| Gemini 3.1 Pro        | gemini-3.1-pro        | `https://opencode.ai/zen/v1/models/gemini-3.1-pro` | `@ai-sdk/google`            |
+| Gemini 3 Flash        | gemini-3-flash        | `https://opencode.ai/zen/v1/models/gemini-3-flash` | `@ai-sdk/google`            |
+| MiniMax M2.5          | minimax-m2.5          | `https://opencode.ai/zen/v1/chat/completions`      | `@ai-sdk/openai-compatible` |
+| MiniMax M2.5 Free     | minimax-m2.5-free     | `https://opencode.ai/zen/v1/chat/completions`      | `@ai-sdk/openai-compatible` |
+| GLM 5                 | glm-5                 | `https://opencode.ai/zen/v1/chat/completions`      | `@ai-sdk/openai-compatible` |
+| Kimi K2.5             | kimi-k2.5             | `https://opencode.ai/zen/v1/chat/completions`      | `@ai-sdk/openai-compatible` |
+| Big Pickle            | big-pickle            | `https://opencode.ai/zen/v1/chat/completions`      | `@ai-sdk/openai-compatible` |
+| MiMo V2 Pro Free      | mimo-v2-pro-free      | `https://opencode.ai/zen/v1/chat/completions`      | `@ai-sdk/openai-compatible` |
+| MiMo V2 Omni Free     | mimo-v2-omni-free     | `https://opencode.ai/zen/v1/chat/completions`      | `@ai-sdk/openai-compatible` |
+| Qwen3.6 Plus Free     | qwen3.6-plus-free     | `https://opencode.ai/zen/v1/chat/completions`      | `@ai-sdk/openai-compatible` |
+| Nemotron 3 Super Free | nemotron-3-super-free | `https://opencode.ai/zen/v1/chat/completions`      | `@ai-sdk/openai-compatible` |
 
-OpenCode 設定の [model id](/docs/config/#models)
-`opencode/<model-id>` 形式を使用します。たとえば、GPT 5.2 Codex の場合は、
-設定で `opencode/gpt-5.2-codex` を使用してください。
+OpenCode 設定で使う [model id](/docs/config/#models) は `opencode/<model-id>` 形式です。たとえば、GPT 5.3 Codex では設定に `opencode/gpt-5.3-codex` を使用します。
 
 ---
 
 ### モデル
 
-利用可能なモデルとそのメタデータの完全なリストは、次から取得できます。
+利用可能なモデルとそのメタデータの完全な一覧は、次から取得できます。
 
 ```
 https://opencode.ai/zen/v1/models
@@ -106,78 +106,81 @@ https://opencode.ai/zen/v1/models
 
 ## 価格
 
-当社は従量課金制モデルをサポートしています。以下は **100 万トークンあたりの価格**です。
+従量課金制をサポートしています。以下は **100万トークンあたり** の価格です。
 
-| Model                             | Input  | Output | Cached Read | Cached Write |
-| --------------------------------- | ------ | ------ | ----------- | ------------ |
-| Big Pickle                        | Free   | Free   | Free        | -            |
-| MiniMax M2.5 Free                 | Free   | Free   | Free        | -            |
-| MiniMax M2.5                      | $0.30  | $1.20  | $0.06       | -            |
-| MiniMax M2.1                      | $0.30  | $1.20  | $0.10       | -            |
-| GLM 5                             | $1.00  | $3.20  | $0.20       | -            |
-| GLM 4.7                           | $0.60  | $2.20  | $0.10       | -            |
-| GLM 4.6                           | $0.60  | $2.20  | $0.10       | -            |
-| Kimi K2.5                         | $0.60  | $3.00  | $0.08       | -            |
-| Kimi K2 Thinking                  | $0.40  | $2.50  | -           | -            |
-| Kimi K2                           | $0.40  | $2.50  | -           | -            |
-| Qwen3 Coder 480B                  | $0.45  | $1.50  | -           | -            |
-| Claude Opus 4.6 (≤ 200K tokens)   | $5.00  | $25.00 | $0.50       | $6.25        |
-| Claude Opus 4.6 (> 200K tokens)   | $10.00 | $37.50 | $1.00       | $12.50       |
-| Claude Opus 4.5                   | $5.00  | $25.00 | $0.50       | $6.25        |
-| Claude Opus 4.1                   | $15.00 | $75.00 | $1.50       | $18.75       |
-| Claude Sonnet 4.6 (≤ 200K tokens) | $3.00  | $15.00 | $0.30       | $3.75        |
-| Claude Sonnet 4.6 (> 200K tokens) | $6.00  | $22.50 | $0.60       | $7.50        |
-| Claude Sonnet 4.5 (≤ 200K tokens) | $3.00  | $15.00 | $0.30       | $3.75        |
-| Claude Sonnet 4.5 (> 200K tokens) | $6.00  | $22.50 | $0.60       | $7.50        |
-| Claude Sonnet 4 (≤ 200K tokens)   | $3.00  | $15.00 | $0.30       | $3.75        |
-| Claude Sonnet 4 (> 200K tokens)   | $6.00  | $22.50 | $0.60       | $7.50        |
-| Claude Haiku 4.5                  | $1.00  | $5.00  | $0.10       | $1.25        |
-| Claude Haiku 3.5                  | $0.80  | $4.00  | $0.08       | $1.00        |
-| Gemini 3.1 Pro (≤ 200K tokens)    | $2.00  | $12.00 | $0.20       | -            |
-| Gemini 3.1 Pro (> 200K tokens)    | $4.00  | $18.00 | $0.40       | -            |
-| Gemini 3 Pro (≤ 200K tokens)      | $2.00  | $12.00 | $0.20       | -            |
-| Gemini 3 Pro (> 200K tokens)      | $4.00  | $18.00 | $0.40       | -            |
-| Gemini 3 Flash                    | $0.50  | $3.00  | $0.05       | -            |
-| GPT 5.4                           | $2.50  | $15.00 | $0.25       | -            |
-| GPT 5.3 Codex                     | $1.75  | $14.00 | $0.175      | -            |
-| GPT 5.2                           | $1.75  | $14.00 | $0.175      | -            |
-| GPT 5.2 Codex                     | $1.75  | $14.00 | $0.175      | -            |
-| GPT 5.1                           | $1.07  | $8.50  | $0.107      | -            |
-| GPT 5.1 Codex                     | $1.07  | $8.50  | $0.107      | -            |
-| GPT 5.1 Codex Max                 | $1.25  | $10.00 | $0.125      | -            |
-| GPT 5.1 Codex Mini                | $0.25  | $2.00  | $0.025      | -            |
-| GPT 5                             | $1.07  | $8.50  | $0.107      | -            |
-| GPT 5 Codex                       | $1.07  | $8.50  | $0.107      | -            |
-| GPT 5 Nano                        | Free   | Free   | Free        | -            |
+| Model                             | Input  | Output  | Cached Read | Cached Write |
+| --------------------------------- | ------ | ------- | ----------- | ------------ |
+| Big Pickle                        | Free   | Free    | Free        | -            |
+| MiMo V2 Pro Free                  | Free   | Free    | Free        | -            |
+| MiMo V2 Omni Free                 | Free   | Free    | Free        | -            |
+| Qwen3.6 Plus Free                 | Free   | Free    | Free        | -            |
+| Nemotron 3 Super Free             | Free   | Free    | Free        | -            |
+| MiniMax M2.5 Free                 | Free   | Free    | Free        | -            |
+| MiniMax M2.5                      | $0.30  | $1.20   | $0.06       | $0.375       |
+| GLM 5                             | $1.00  | $3.20   | $0.20       | -            |
+| Kimi K2.5                         | $0.60  | $3.00   | $0.10       | -            |
+| Qwen3 Coder 480B                  | $0.45  | $1.50   | -           | -            |
+| Claude Opus 4.6                   | $5.00  | $25.00  | $0.50       | $6.25        |
+| Claude Opus 4.5                   | $5.00  | $25.00  | $0.50       | $6.25        |
+| Claude Opus 4.1                   | $15.00 | $75.00  | $1.50       | $18.75       |
+| Claude Sonnet 4.6                 | $3.00  | $15.00  | $0.30       | $3.75        |
+| Claude Sonnet 4.5 (≤ 200K tokens) | $3.00  | $15.00  | $0.30       | $3.75        |
+| Claude Sonnet 4.5 (> 200K tokens) | $6.00  | $22.50  | $0.60       | $7.50        |
+| Claude Sonnet 4 (≤ 200K tokens)   | $3.00  | $15.00  | $0.30       | $3.75        |
+| Claude Sonnet 4 (> 200K tokens)   | $6.00  | $22.50  | $0.60       | $7.50        |
+| Claude Haiku 4.5                  | $1.00  | $5.00   | $0.10       | $1.25        |
+| Claude Haiku 3.5                  | $0.80  | $4.00   | $0.08       | $1.00        |
+| Gemini 3.1 Pro (≤ 200K tokens)    | $2.00  | $12.00  | $0.20       | -            |
+| Gemini 3.1 Pro (> 200K tokens)    | $4.00  | $18.00  | $0.40       | -            |
+| Gemini 3 Flash                    | $0.50  | $3.00   | $0.05       | -            |
+| GPT 5.4                           | $2.50  | $15.00  | $0.25       | -            |
+| GPT 5.4 Pro                       | $30.00 | $180.00 | $30.00      | -            |
+| GPT 5.4 Mini                      | $0.75  | $4.50   | $0.075      | -            |
+| GPT 5.4 Nano                      | $0.20  | $1.25   | $0.02       | -            |
+| GPT 5.3 Codex Spark               | $1.75  | $14.00  | $0.175      | -            |
+| GPT 5.3 Codex                     | $1.75  | $14.00  | $0.175      | -            |
+| GPT 5.2                           | $1.75  | $14.00  | $0.175      | -            |
+| GPT 5.2 Codex                     | $1.75  | $14.00  | $0.175      | -            |
+| GPT 5.1                           | $1.07  | $8.50   | $0.107      | -            |
+| GPT 5.1 Codex                     | $1.07  | $8.50   | $0.107      | -            |
+| GPT 5.1 Codex Max                 | $1.25  | $10.00  | $0.125      | -            |
+| GPT 5.1 Codex Mini                | $0.25  | $2.00   | $0.025      | -            |
+| GPT 5                             | $1.07  | $8.50   | $0.107      | -            |
+| GPT 5 Codex                       | $1.07  | $8.50   | $0.107      | -            |
+| GPT 5 Nano                        | Free   | Free    | Free        | -            |
 
-使用履歴に _Claude Haiku 3.5_ が表示されるかもしれません。これは [セッションのタイトルを生成するために使用される低コストモデル](/docs/config/#models) です。
+使用履歴に _Claude Haiku 3.5_ が表示されることがあります。これはセッションのタイトル生成に使われる [low cost model](/docs/config/#models) です。
 
 :::note
-クレジットカード手数料は実費で引き継がれます (4.4% + 取引ごとに 0.30 ドル)。それ以上の料金はかかりません。
+クレジットカード手数料は実費で転嫁されます（各取引につき 4.4% + $0.30）。それ以外の上乗せはありません。
 :::
 
-無料のモデル:
+無料モデル:
 
-- MiniMax M2.5 Free は期間限定で OpenCode で利用可能です。チームはこの期間を利用してフィードバックを収集し、モデルを改善します。
-- Big Pickle は、期間限定で OpenCode で無料で利用できるステルスモデルです。チームはこの期間を利用してフィードバックを収集し、モデルを改善します。
+- MiniMax M2.5 Free は期間限定で OpenCode で利用できます。チームはこの期間中にフィードバックを集め、モデルを改善しています。
+- MiMo V2 Pro Free は期間限定で OpenCode で利用できます。チームはこの期間中にフィードバックを集め、モデルを改善しています。
+- MiMo V2 Omni Free は期間限定で OpenCode で利用できます。チームはこの期間中にフィードバックを集め、モデルを改善しています。
+- Qwen3.6 Plus Free は期間限定で OpenCode で利用できます。チームはこの期間中にフィードバックを集め、モデルを改善しています。
+- Nemotron 3 Super Free は期間限定で OpenCode で利用できます。チームはこの期間中にフィードバックを集め、モデルを改善しています。
+- Big Pickle はステルスモデルで、期間限定で OpenCode で無料提供されています。チームはこの期間中にフィードバックを集め、モデルを改善しています。
 
-ご質問がございましたら、<a href={email}>お問い合わせ</a>ください。
+ご不明な点があれば、<a href={email}>お問い合わせください</a>。
 
 ---
 
 ### 自動リロード
 
-残高が 5 ドルを下回ると、Zen は自動的に 20 ドルをリロードします。
+残高が $5 を下回ると、Zen は自動的に $20 をリロードします。
 
-自動リロード量を変更できます。自動リロードを完全に無効にすることもできます。
+自動リロード額は変更できます。また、自動リロード自体を完全に無効化することもできます。
 
 ---
 
-### 月ごとの制限
+### 月次上限
 
-ワークスペース全体およびチームの各メンバーの月ごとの使用制限を設定することもできます。
+ワークスペース全体とチームの各メンバーに対して、月ごとの利用上限を設定することもできます。
 
-たとえば、毎月の使用制限を 20 ドルに設定したとします。Zen は月に 20 ドル以上を使用しません。ただし、自動リロードを有効にしている場合、残高が 5 ドルを下回ると、Zen が 20 ドル以上の請求を行う可能性があります。
+たとえば、月次利用上限を $20 に設定した場合、Zen は 1 か月で $20 を超えて使われることはありません。ただし、自動リロードが有効になっていると、残高が $5 を下回った時点で Zen が $20 を自動チャージするため、結果として $20 を超えて請求される可能性があります。
 
 ---
 
@@ -185,74 +188,77 @@ https://opencode.ai/zen/v1/models
 
 | Model            | Deprecation date |
 | ---------------- | ---------------- |
-| Qwen3 Coder 480B | 2026年2月6日     |
-| Kimi K2 Thinking | 2026年3月6日     |
-| Kimi K2          | 2026年3月6日     |
-| MiniMax M2.1     | 2026年3月15日    |
-| GLM 4.7          | 2026年3月15日    |
-| GLM 4.6          | 2026年3月15日    |
+| MiniMax M2.1     | March 15, 2026   |
+| GLM 4.7          | March 15, 2026   |
+| GLM 4.6          | March 15, 2026   |
+| Gemini 3 Pro     | March 9, 2026    |
+| Kimi K2 Thinking | March 6, 2026    |
+| Kimi K2          | March 6, 2026    |
+| Qwen3 Coder 480B | Feb 6, 2026      |
 
 ---
 
 ## プライバシー
 
-すべてのモデルは米国でホストされています。当社のプロバイダーはゼロ保持ポリシーに従い、次の例外を除いて、モデルのトレーニングにデータを使用しません。
+すべてのモデルは米国でホストされています。私たちのプロバイダーはゼロ保持ポリシーに従っており、以下の例外を除き、モデル学習にあなたのデータを使用しません。
 
-- Big Pickle: 無料期間中、収集されたデータはモデルの改善に使用される場合があります。
-- MiniMax M2.5 Free: 無料期間中、収集されたデータはモデルの改善に使用される場合があります。
-- OpenAI API: リクエストは [OpenAI のデータポリシー](https://platform.openai.com/docs/guides/your-data) に従い、30 日間保持されます。
-- Anthropic API: リクエストは [Anthropic のデータポリシー](https://docs.anthropic.com/en/docs/claude-code/data-usage) に従い、30 日間保持されます。
+- Big Pickle: 無料提供期間中、収集されたデータがモデル改善に使われる場合があります。
+- MiniMax M2.5 Free: 無料提供期間中、収集されたデータがモデル改善に使われる場合があります。
+- MiMo V2 Pro Free: 無料提供期間中、収集されたデータがモデル改善に使われる場合があります。
+- MiMo V2 Omni Free: 無料提供期間中、収集されたデータがモデル改善に使われる場合があります。
+- Qwen3.6 Plus Free: 無料提供期間中、収集されたデータがモデル改善に使われる場合があります。
+- Nemotron 3 Super Free: 無料提供期間中、収集されたデータがモデル改善に使われる場合があります。
+- OpenAI APIs: リクエストは [OpenAI's Data Policies](https://platform.openai.com/docs/guides/your-data) に従って 30 日間保持されます。
+- Anthropic APIs: リクエストは [Anthropic's Data Policies](https://docs.anthropic.com/en/docs/claude-code/data-usage) に従って 30 日間保持されます。
 
 ---
 
 ## チーム向け
 
-Zen はチームにも効果的です。チームメイトを招待し、役割を割り当て、チームが使用するモデルをキュレートすることなどができます。
+Zen はチームでも非常に便利です。チームメイトを招待し、ロールを割り当て、チームで使うモデルを厳選するなど、さまざまなことができます。
 
 :::note
-ワークスペースは現在、ベータ版の一部としてチームに無料で提供されています。
+ワークスペースは現在、ベータの一環としてチーム向けに無料です。
 :::
 
-ワークスペースの管理は現在、ベータ版の一部としてチームに無料で提供されています。価格の詳細については近日中にお知らせします。
+ワークスペース管理は現在、ベータの一環としてチーム向けに無料です。価格についての詳細は近日中に共有します。
 
 ---
 
 ### ロール
 
-チームメイトをワークスペースに招待し、役割を割り当てることができます。
+チームメイトをワークスペースに招待し、ロールを割り当てられます。
 
-- **管理者**: モデル、メンバー、API キー、請求を管理します。
-- **メンバー**: 自分の API キーのみを管理します
+- **Admin**: モデル、メンバー、API キー、請求を管理できます
+- **Member**: 自分自身の API キーのみ管理できます
 
-管理者は、コストを管理するために各メンバーの毎月の支出制限を設定することもできます。
+Admin は、コストを抑えるために各メンバーの月次支出上限を設定することもできます。
 
 ---
 
 ### モデルアクセス
 
-管理者は、ワークスペースの特定のモデルを有効または無効にすることができます。無効なモデルに対してリクエストを行うと、エラーが返されます。
+Admin は、ワークスペースで特定のモデルを有効化または無効化できます。無効化されたモデルへのリクエストはエラーになります。
 
-これは、モデルの使用を無効にしたい場合に便利です。
-データを収集します。
+これは、データを収集するモデルの利用を無効にしたい場合に便利です。
 
 ---
 
-### 自分のキーを持ち込む
+### Bring your own key
 
-Zen の他のモデルにアクセスしながら、独自の OpenAI または Anthropic API キーを使用できます。
+Zen のほかのモデルにアクセスしながら、自分の OpenAI または Anthropic API キーを使うこともできます。
 
-独自のキーを使用する場合、トークンは Zen ではなくプロバイダーによって直接請求されます。
+自分のキーを使う場合、トークン料金は Zen ではなくプロバイダーから直接請求されます。
 
-たとえば、組織はすでに OpenAI または Anthropic のキーを持っている可能性があります。
-Zen が提供するものの代わりにそれを使用したいとします。
+たとえば、あなたの組織がすでに OpenAI や Anthropic のキーを持っていて、Zen が提供するものではなくそちらを使いたい場合があります。
 
 ---
 
 ## 目標
 
-私たちは次の目的で OpenCode Zen を作成しました。
+私たちが OpenCode Zen を作った目的は次のとおりです。
 
-1. **ベンチマーク** コーディングエージェントに最適なモデル/プロバイダー。
-2. **最高品質**のオプションにアクセスでき、パフォーマンスをダウングレードしたり、より安価なプロバイダーにルートしたりする必要はありません。
-3. 原価で販売することで**価格下落**を転嫁します。したがって、唯一のマークアップは処理手数料をカバーすることです。
-4. 他のコーディングエージェントとの併用を許可することで、**ロックイン**がなくなります。また、常に OpenCode で他のプロバイダーも使用できるようにします。
+1. コーディングエージェントに最適なモデル/プロバイダーを **Benchmark** すること。
+2. **highest quality** の選択肢にアクセスできるようにし、性能を落としたり、より安価なプロバイダーへルーティングされたりしないこと。
+3. 原価で提供することで **price drops** をそのまま反映し、上乗せは処理手数料のカバー分だけにすること。
+4. ほかのコーディングエージェントでも使えるようにして **no lock-in** を実現し、さらに OpenCode では常にほかのプロバイダーも使えるようにすること。