feat(session-ui): render streaming markdown with Solid

Co-authored-by: Aiden Cline <63023139+rekram1-node@users.noreply.github.com>
Co-authored-by: opencode-agent[bot] <219766164+opencode-agent[bot]@users.noreply.github.com>

.changeset/device-authorization-url.md

@@ -0,0 +1,5 @@
+---
+"@opencode-ai/core": patch
+---
+
+Fix OpenCode Console device authorization URLs when the server returns an origin-rooted verification path.

.changeset/idempotent-prompt-reuse.md

@@ -0,0 +1,5 @@
+---
+"@opencode-ai/core": patch
+---
+
+Prompt and synthetic inbox ID reuse is now idempotent: reusing an ID within the same Session succeeds and returns the first admission, ignoring the retried payload, metadata, and delivery mode. Previously reuse with a differing payload failed with a conflict. Cross-Session and cross-type reuse still fail, and control items keep their operation-specific conflict behavior.

.changeset/nested-instruction-reinjection.md

@@ -0,0 +1,5 @@
+---
+"@opencode-ai/core": patch
+---
+
+Nested AGENTS.md instructions are re-injected after compaction. Previously the in-memory dedup claim outlived the synthetic message that compaction dropped from model-visible history, so nested instructions were silently lost for the rest of the process lifetime. The claim now only guards in-flight loads; the synthetic message metadata in durable history is the sole lasting ledger, so any history truncation (compaction, revert) self-heals on the next read in that subtree.

.changeset/share-session-model-request.md

@@ -0,0 +1,5 @@
+---
+"@opencode-ai/core": patch
+---
+
+Apply shared Session model-request preparation to transient generation.

.changeset/steer-resume-simplified.md

@@ -0,0 +1,5 @@
+---
+"@opencode-ai/core": patch
+---
+
+Simplify interrupt continuation: the steer-scoped resume decision now lives in SessionExecution as a post-cleanup inbox check, and the run coordinator drops its continuation state machine. Wakes arriving during cancellation cleanup now restart a normal full drain, and interrupting an idle session with continue now resumes pending steering input. Recovery-applied moves now end with the same full wake as inbox-admitted moves, retrying any stranded inbox work at the new location. Interrupting with continue now also resumes a next-in-line control item: between-turn manual compaction and moves run under any drain scope, while queued prompts remain parked.

.changeset/unify-request-envelope.md

@@ -0,0 +1,5 @@
+---
+"@opencode-ai/core": patch
+---
+
+Title generation and compaction summaries now build their model requests through the shared session request boundary, gaining unsupported-media filtering and image bounds while explicitly opting out of session context hooks: plugins that shape the agent conversation do not observe title or compaction requests. Title requests gain the fork-aware session prompt cache key, and compaction summaries in forked sessions reuse the fork root's prompt cache key instead of the fork's own.

.github/workflows/deploy-posts.yml

@@ -0,0 +1,37 @@
+name: deploy-posts
+
+on:
+  push:
+    branches:
+      - v2
+    paths:
+      - packages/posts/**
+      - bun.lock
+      - .github/workflows/deploy-posts.yml
+  workflow_dispatch:
+
+concurrency:
+  group: deploy-posts-${{ github.ref_name }}
+  cancel-in-progress: false
+
+permissions:
+  contents: read
+
+jobs:
+  deploy:
+    if: github.repository == 'anomalyco/opencode' && github.ref_name == 'v2'
+    runs-on: ubuntu-latest
+    steps:
+      - uses: actions/checkout@f43a0e5ff2bd294095638e18286ca9a3d1956744 # v3.6.0
+
+      - uses: ./.github/actions/setup-bun
+
+      - name: Build
+        working-directory: packages/posts
+        run: bun run build
+
+      - name: Deploy
+        working-directory: packages/posts
+        run: bun run deploy
+        env:
+          CLOUDFLARE_API_TOKEN: ${{ secrets.CLOUDFLARE_API_TOKEN }}

.github/workflows/deploy-www.yml

@@ -27,7 +27,6 @@ jobs:
         working-directory: packages/www
         run: bun run build
         env:
-          BLUME_ENV: ${{ github.ref_name == 'v2' && 'production' || 'dev' }}
           CLOUDFLARE_ENV: ${{ github.ref_name == 'v2' && 'production' || 'dev' }}
 
       - name: Deploy

.github/workflows/publish.yml

@@ -91,7 +91,7 @@ jobs:
 
       - uses: ./.github/actions/setup-bun
         with:
-          bun-version: canary # Bun 1.4 until its stable release is published
+          bun-version: 1.4.0
 
       - name: Setup git committer
         id: committer
@@ -113,7 +113,7 @@ jobs:
         id: build
         run: ./packages/cli/script/build.ts ${{ (github.ref_name == 'beta' && '--sourcemaps') || '' }}
         env:
-          BUN_COMPILE_RELEASE: canary
+          BUN_COMPILE_RELEASE: bun-v1.4.0
           OPENCODE_VERSION: ${{ needs.version.outputs.version }}
           OPENCODE_RELEASE: ${{ needs.version.outputs.release }}
           GH_REPO: ${{ needs.version.outputs.repo }}
@@ -195,9 +195,33 @@ jobs:
           path: packages/cli/dist/cli-*
           if-no-files-found: error
 
-  build-node-cli:
+  build-node-app-archive:
     needs: version
-    if: github.repository == 'anomalyco/opencode' && false # Temporarily disabled
+    runs-on: blacksmith-4vcpu-ubuntu-2404
+    timeout-minutes: 30
+    if: github.repository == 'anomalyco/opencode'
+    steps:
+      - uses: actions/checkout@f43a0e5ff2bd294095638e18286ca9a3d1956744 # v3.6.0
+
+      - uses: ./.github/actions/setup-bun
+
+      - name: Build app archive
+        run: bun packages/cli/script/build-node.ts --app-archive-only --app-archive=.cache/app-archive.bin --skip-install
+        env:
+          OPENCODE_VERSION: ${{ needs.version.outputs.version }}
+          OPENCODE_RELEASE: ${{ needs.version.outputs.release }}
+
+      - uses: actions/upload-artifact@ea165f8d65b6e75b540449e92b4886f43607fa02 # v4.6.2
+        with:
+          name: opencode-node-app-archive
+          path: packages/cli/.cache/app-archive.bin
+          if-no-files-found: error
+
+  build-node-cli:
+    needs:
+      - version
+      - build-node-app-archive
+    if: github.repository == 'anomalyco/opencode'
     strategy:
       fail-fast: false
       matrix:
@@ -227,8 +251,13 @@ jobs:
         with:
           node-version: "26.4.0"
 
+      - uses: actions/download-artifact@d3f86a106a0bac45b974a628896c90dbdf5c8093 # v4.3.0
+        with:
+          name: opencode-node-app-archive
+          path: packages/cli/.cache
+
       - name: Build
-        run: bun packages/cli/script/build-node.ts --target=${{ matrix.settings.target }} --skip-install --outdir=dist/node
+        run: bun packages/cli/script/build-node.ts --target=${{ matrix.settings.target }} --skip-install --outdir=dist/node --app-archive=.cache/app-archive.bin
         env:
           OPENCODE_VERSION: ${{ needs.version.outputs.version }}
           OPENCODE_RELEASE: ${{ needs.version.outputs.release }}
@@ -543,6 +572,7 @@ jobs:
       - version
       - build-cli
       - sign-cli-macos
+      - build-node-app-archive
       - build-node-cli
       - sign-cli-windows
       - build-electron

.github/workflows/test.yml

@@ -50,6 +50,14 @@ jobs:
       - name: Setup Bun
         uses: ./.github/actions/setup-bun
 
+      - name: Test Effect simplification rules
+        if: runner.os == 'Linux'
+        run: bun run test:effect-simplification-rules
+
+      - name: Check Effect simplifications
+        if: runner.os == 'Linux'
+        run: bun run lint:effect-simplifications
+
       - name: Configure git identity
         run: |
           git config --global user.email "bot@opencode.ai"

.opencode/tui.json

@@ -1,19 +1,3 @@
 {
-  "$schema": "https://opencode.ai/tui.json",
-  "plugin": [
-    [
-      "./plugins/tui-smoke.tsx",
-      {
-        "enabled": false,
-        "label": "workspace",
-        "keybinds": {
-          "smoke_modal": "ctrl+alt+m",
-          "smoke_screen": "ctrl+alt+o",
-          "smoke_screen_home": "escape,ctrl+shift+h",
-          "smoke_screen_modal": "ctrl+alt+m",
-          "smoke_dialog_close": "escape,q"
-        }
-      }
-    ]
-  ]
+  "$schema": "https://opencode.ai/tui.json"
 }

AGENTS.md

@@ -1,5 +1,5 @@
 - After changing the public Protocol or Server `HttpApi`, run `bun run generate` from `packages/client`. Do not edit generated client files directly.
-- Keep runtime dependencies directed from Schema to Core and Protocol, then from Core and Protocol to Server. Client runtime code may depend on Schema and Protocol but never Core or Server; `sdk-next` composes Client, Core, and Server.
+- Keep runtime dependencies directed from Schema to Core and Protocol, then from Core and Protocol to Server. Client runtime code may depend on Schema and Protocol but never Core or Server; `sdk` composes Client, Core, and Server.
 - Current implementation changes belong in `packages/core`, `packages/cli`, `packages/server`, `packages/protocol`, `packages/schema`, and related generated client surfaces when required.
 - The default branch in this repo is `v2`.
 - Base all new branches and worktrees on `v2`, or `origin/v2` when the local `v2` ref is unavailable. Do not base them on `dev`.
@@ -176,7 +176,7 @@ const table = sqliteTable("session", {
 
 - Keep durable events minimal: record irreducible new facts and do not repeat state derivable by folding the ordered aggregate history. Enrich projections and read models with previous or derived state when consumers need self-contained views.
 - Keep durable prompt admission separate from model execution. `Session.prompt(...)` publishes `session.inbox.enqueued`, whose projection inserts one durable `session_inbox` row, before scheduling advisory `SessionExecution.wake(sessionID)` unless `resume: false` requests admit-only behavior. Delivery publishes `session.inbox.delivered`; its projection consumes the inbox row and inserts the visible message in the same transaction. `session_inbox` stores only unconsumed work.
-- Reusing a Session ID adopts the existing Session. While a user or synthetic inbox item is pending, reusing its ID reconciles only when Session, type, complete payload, metadata, and delivery match; conflicting reuse fails. Once delivered, retry reconciliation for those message-producing items uses the projected message and does not require retained enqueue history or the original delivery mode. Control items keep their operation-specific conflict behavior.
+- Reusing a Session ID adopts the existing Session. Reusing a user or synthetic inbox item ID is idempotent when Session and type match: the first admission wins and the retried payload, metadata, and delivery mode are ignored, whether the item is still pending or already delivered (reconciled from the projected message without retained enqueue history). Cross-Session or cross-type reuse fails. Control items keep their operation-specific conflict behavior.
 - Keep `SessionExecution` process-global and Session-ID based. Its local implementation owns the process-local Session coordinator and discovers placement through `SessionStore` plus `LocationServiceMap.get(session.location)` only when a drain starts; no layer should take a Session ID. V2 interruption targets the active process-local ownership chain for that Session; interruption of a known but idle or locally unowned Session is a no-op, while the public API rejects an unknown Session.
 - Keep `SessionRunner`, model resolution, tool registry, permissions, and filesystem Location-scoped. Omitted `Location.workspaceID` means implicit-local placement; explicit workspace identity remains reserved for future placement semantics.
 - Preserve one explicit `llm.stream(request)` call per Physical Attempt and reload projected history before durable continuation. A logical Step may use generic pre-output retries, one full-context retry after continuation rejection, incomplete-stream continuation, or one overflow-compaction rebuild. Generic retries retain the logical step number and do not consume another agent-step allowance. Do not delegate orchestration to an in-memory tool loop.

nix/hashes.json

@@ -1,8 +1,8 @@
 {
   "nodeModules": {
-    "x86_64-linux": "sha256-IxkSw0gK/qkMHZGVHqjwgM9BKhzbQX6hyF9SWUNtpzg=",
-    "aarch64-linux": "sha256-YVjpbil0QswVwi6NtVYFq3xCqpsfveG1chlNVCVI0MU=",
-    "aarch64-darwin": "sha256-CdL2mI84pawH2H5i9qu8A6IWbkmKOYHlJS+DI/Mafdw=",
-    "x86_64-darwin": "sha256-NtswwfU5WYv99bEmI4XeLwjhBGcS9ZMYLRo4MQRNtLo="
+    "x86_64-linux": "sha256-tdNjQEuCMADQJpeAK0BajPyQkt8X2pS9foyQYxj+vgg=",
+    "aarch64-linux": "sha256-RXvYaZXpgTyZzgPu74fat1bbT9zJUf+S0kpm9rXtr6k=",
+    "aarch64-darwin": "sha256-MDinXCNrEn5pH75nmK/dsl78hcS4gu913Sf2fxj40r0=",
+    "x86_64-darwin": "sha256-pn9t+1sNDjQh25nfPozD11aIJJsHgxhU9uB4tR0UnJk="
   }
 }

package.json

@@ -15,42 +15,44 @@
     "dev:stats": "bun sst shell --stage=production -- bun run --cwd packages/stats/app dev",
     "dev:www": "bun run --cwd packages/www dev",
     "dev:storybook": "bun --cwd packages/storybook storybook",
+    "bench:devex": "bun run --cwd packages/app test:bench:devex",
     "lint": "oxlint",
     "lint:effect-patterns": "ast-grep scan -c script/ast-grep/sgconfig.yml packages/util/src packages/core/src packages/server/src packages/protocol/src packages/cli/src",
+    "lint:effect-simplifications": "ast-grep scan -c script/ast-grep/effect-simplifications/sgconfig.yml --off=unused-suppression packages",
     "test:lint-rules": "ast-grep test -c script/ast-grep/sgconfig.yml",
+    "test:effect-simplification-rules": "ast-grep test -c script/ast-grep/effect-simplifications/sgconfig.yml",
     "typecheck": "bun turbo typecheck --concurrency=3",
     "typecheck:profile": "bun script/profile-typecheck.ts",
     "typecheck:profile:packages": "bun script/profile-typecheck-packages.ts",
     "upgrade-opentui": "bun run script/upgrade-opentui.ts",
     "postinstall": "bun run --cwd packages/core fix-node-pty",
     "prepare": "husky",
+    "reserve-packages": "bun script/reserve-package-names.ts",
     "random": "echo 'Random script'",
     "sso": "aws sso login --sso-session=opencode --no-browser",
-    "translate:app": "bun run script/translate-app.ts",
     "test": "echo 'do not run tests from root' && exit 1"
   },
   "workspaces": {
     "packages": [
       "packages/*",
       "packages/console/*",
-      "packages/stats/*",
-      "packages/slack"
+      "packages/stats/*"
     ],
     "catalog": {
-      "@effect/opentelemetry": "4.0.0-beta.107",
-      "@effect/platform-node": "4.0.0-beta.107",
-      "@effect/platform-node-shared": "4.0.0-beta.107",
-      "@effect/sql-sqlite-bun": "4.0.0-beta.107",
+      "@effect/opentelemetry": "4.0.0-rc.110",
+      "@effect/platform-node": "4.0.0-rc.110",
+      "@effect/platform-node-shared": "4.0.0-rc.110",
+      "@effect/sql-sqlite-bun": "4.0.0-rc.110",
       "@npmcli/arborist": "9.4.0",
       "@types/bun": "1.3.13",
       "@types/cross-spawn": "6.0.6",
       "@octokit/rest": "22.0.0",
       "@hono/standard-validator": "0.2.0",
       "@hono/zod-validator": "0.4.2",
-      "@opentui/core": "0.5.4",
-      "@opentui/keymap": "0.5.4",
-      "@opentui/solid": "0.5.4",
-      "@tanstack/solid-virtual": "3.13.32",
+      "@opentui/core": "0.5.7",
+      "@opentui/keymap": "0.5.7",
+      "@opentui/solid": "0.5.7",
+      "@tanstack/solid-virtual": "3.13.37",
       "@shikijs/stream": "4.2.0",
       "@standard-schema/spec": "1.1.0",
       "ulid": "3.0.1",
@@ -71,7 +73,7 @@
       "dompurify": "3.3.1",
       "drizzle-kit": "1.0.0-rc.2",
       "drizzle-orm": "1.0.0-rc.2",
-      "effect": "4.0.0-beta.107",
+      "effect": "4.0.0-rc.110",
       "ai": "6.0.168",
       "cross-spawn": "7.0.6",
       "hono": "4.10.7",
@@ -127,7 +129,6 @@
     "@aws-sdk/client-s3": "3.933.0",
     "@opencode-ai/plugin": "workspace:*",
     "@opencode-ai/script": "workspace:*",
-    "@opencode-ai/sdk": "1.18.5",
     "heap-snapshot-toolkit": "1.1.3",
     "typescript": "catalog:"
   },
@@ -173,7 +174,7 @@
     "@ai-sdk/google@3.0.73": "patches/@ai-sdk%2Fgoogle@3.0.73.patch",
     "@pierre/trees@1.0.0-beta.4": "patches/@pierre%2Ftrees@1.0.0-beta.4.patch",
     "@modelcontextprotocol/sdk@1.29.0": "patches/@modelcontextprotocol%2Fsdk@1.29.0.patch",
-    "@tanstack/virtual-core@3.17.3": "patches/@tanstack%2Fvirtual-core@3.17.3.patch",
-    "@ff-labs/fff-bun@0.10.1": "patches/@ff-labs%2Ffff-bun@0.10.1.patch"
+    "@tanstack/virtual-core@3.17.8": "patches/@tanstack%2Fvirtual-core@3.17.8.patch",
+    "@ff-labs/fff-bun@0.10.5": "patches/@ff-labs%2Ffff-bun@0.10.5.patch"
   }
 }

packages/ai/README.md

@@ -309,7 +309,7 @@ Included providers: OpenAI, Anthropic, Google (Gemini), Google Vertex Gemini and
 
 ### Package-like entrypoints
 
-Native catalog integrations load provider behavior through package-like entrypoints. These are export paths from the same `@opencode-ai/ai` npm package, not independently published packages. Each entrypoint exports the same `model(modelID, settings)` contract, and `settings` contains serializable provider configuration plus common `headers`, `body`, and `limits` overlays.
+Native catalog integrations load provider behavior through package-like entrypoints. These are export paths from the same `@opencode-ai/ai` npm package, not independently published packages. Each entrypoint exports the same `model(modelID, settings)` contract, and `settings` contains serializable provider configuration plus common `headers` and `body` overlays.
 
 ```ts
 import { model } from "@opencode-ai/ai/providers/openai/responses"
@@ -317,7 +317,6 @@ import { model } from "@opencode-ai/ai/providers/openai/responses"
 const selected = model("gpt-5", {
   apiKey: process.env.OPENAI_API_KEY,
   headers: { "x-application": "opencode" },
-  limits: { context: 200_000, output: 64_000 },
 })
 ```
 

packages/ai/script/setup-recording-env.ts

@@ -370,7 +370,7 @@ const responseError = Effect.fn("RecordingEnv.responseError")(function* (
   response: HttpClientResponse.HttpClientResponse,
 ) {
   if (response.status >= 200 && response.status < 300) return undefined
-  const body = yield* response.text.pipe(Effect.catch(() => Effect.succeed("")))
+  const body = yield* response.text.pipe(Effect.orElseSucceed(() => ""))
   return `${response.status}${body ? `: ${body.slice(0, 180)}` : ""}`
 })
 

packages/ai/src/protocols/anthropic-messages.ts

@@ -31,6 +31,19 @@ import { ToolStream } from "./utils/tool-stream.js"
 const ADAPTER = "anthropic-messages"
 export const DEFAULT_BASE_URL = "https://api.anthropic.com/v1"
 export const PATH = "/messages"
+export const DEFAULT_MAX_TOKENS = 32_000
+
+const SSE_EVENTS = new Set([
+  "message",
+  "message_start",
+  "message_delta",
+  "message_stop",
+  "content_block_start",
+  "content_block_delta",
+  "content_block_stop",
+  "error",
+])
+export const framing = Framing.sseEvents(SSE_EVENTS)
 
 export type ThinkingInput =
   | {
@@ -233,7 +246,7 @@ export type AnthropicMessagesBody = Schema.Schema.Type<typeof AnthropicMessagesB
 
 const AnthropicUsage = Schema.StructWithRest(
   Schema.Struct({
-    input_tokens: Schema.optional(Schema.Number),
+    input_tokens: optionalNull(Schema.Number),
     output_tokens: Schema.optional(Schema.Number),
     cache_creation_input_tokens: optionalNull(Schema.Number),
     cache_read_input_tokens: optionalNull(Schema.Number),
@@ -361,16 +374,18 @@ const lowerToolChoice = (toolChoice: NonNullable<LLMRequest["toolChoice"]>) =>
     tool: (name) => ({ type: "tool" as const, name }),
   })
 
+const scrubToolCallID = (id: string) => id.replace(/[^a-zA-Z0-9_-]/g, "_")
+
 const lowerToolCall = (part: ToolCallPart): AnthropicToolUseBlock => ({
   type: "tool_use",
-  id: part.id,
+  id: scrubToolCallID(part.id),
   name: part.name,
   input: part.input,
 })
 
 const lowerServerToolCall = (part: ToolCallPart): AnthropicServerToolUseBlock => ({
   type: "server_tool_use",
-  id: part.id,
+  id: scrubToolCallID(part.id),
   name: part.name,
   input: part.input,
 })
@@ -392,7 +407,7 @@ const lowerServerToolResult = Effect.fn("AnthropicMessages.lowerServerToolResult
   // Prefer the provider-owned replay payload; fall back to the result value for
   // histories constructed directly from provider events.
   const payload = part.providerMetadata?.anthropic?.["result"] ?? part.result.value
-  return { type: wireType, tool_use_id: part.id, content: payload } satisfies AnthropicServerToolResultBlock
+  return { type: wireType, tool_use_id: scrubToolCallID(part.id), content: payload } satisfies AnthropicServerToolResultBlock
 })
 
 const lowerMedia = Effect.fn("AnthropicMessages.lowerMedia")(function* (part: MediaPart) {
@@ -574,7 +589,7 @@ const lowerMessages = Effect.fn("AnthropicMessages.lowerMessages")(function* (
         return yield* ProviderShared.unsupportedContent("Anthropic Messages", "tool", ["tool-result"])
       content.push({
         type: "tool_result",
-        tool_use_id: part.id,
+        tool_use_id: scrubToolCallID(part.id),
         content: yield* lowerToolResultContent(part),
         is_error: part.result.type === "error" ? true : undefined,
         cache_control: cacheControl(breakpoints, part.cache),
@@ -624,7 +639,6 @@ const resolveThinking = Effect.fn("AnthropicMessages.resolveThinking")(function*
 const fromRequest = Effect.fn("AnthropicMessages.fromRequest")(function* (request: LLMRequest) {
   const generation = request.generation
   const toolSchemaCompatibility = request.model.compatibility?.toolSchema
-  const outputLimit = request.model.defaults?.limits?.output ?? request.model.route.defaults.limits?.output ?? 4096
   // Allocate the 4-breakpoint budget in invalidation order: tools → system →
   // messages. Tools live highest in the cache hierarchy, so when callers
   // over-mark we keep their tool hints and shed the message-tail ones first.
@@ -663,7 +677,7 @@ const fromRequest = Effect.fn("AnthropicMessages.fromRequest")(function* (reques
     tools,
     tool_choice: toolChoice,
     stream: true as const,
-    max_tokens: generation?.maxTokens ?? outputLimit,
+    max_tokens: generation?.maxTokens ?? DEFAULT_MAX_TOKENS,
     temperature: generation?.temperature,
     top_p: generation?.topP,
     top_k: generation?.topK,
@@ -692,7 +706,7 @@ const mapFinishReason = (reason: string | null | undefined): FinishReason => {
 // expose that subset through `output_tokens_details.thinking_tokens`.
 const mapUsage = (usage: AnthropicUsage | undefined): Usage | undefined => {
   if (!usage) return undefined
-  const nonCached = usage.input_tokens
+  const nonCached = usage.input_tokens ?? undefined
   const cacheRead = usage.cache_read_input_tokens ?? undefined
   const cacheWrite = usage.cache_creation_input_tokens ?? undefined
   const inputTokens = ProviderShared.sumTokens(nonCached, cacheRead, cacheWrite)
@@ -1039,7 +1053,7 @@ export const route = Route.make({
   protocol,
   endpoint: Endpoint.path(PATH, { baseURL: DEFAULT_BASE_URL }),
   auth: Auth.none,
-  framing: Framing.sse,
+  framing,
   headers: () => ({ "anthropic-version": "2023-06-01" }),
 })
 

packages/ai/src/protocols/gemini.ts

@@ -37,6 +37,10 @@ const requiresThoughtSignatureFallback = (modelID: string) => {
   return !/(^|\/)gemini-robotics-er-1\.5(?:[.-]|$)/i.test(modelID)
 }
 
+// Gemini 3 accepts media nested inside function responses; matched Gemini 2.5 variants reject it,
+// so their tool-result attachments lower as a separate user turn instead.
+const routesLegacyToolMedia = (modelID: string) => /gemini-2[.-]5(?:[.-]|$)/i.test(modelID)
+
 export interface OptionsInput {
   readonly [key: string]: unknown
   readonly cachedContent?: string
@@ -208,11 +212,11 @@ type GeminiEvent = Schema.Schema.Type<typeof GeminiEvent>
 interface ParserState {
   readonly finishReason?: string
   readonly hasToolCalls: boolean
-  readonly nextToolCallId: number
   readonly promptFeedback?: GeminiPromptFeedback
   readonly usage?: Usage
   readonly lifecycle: Lifecycle.State
   readonly reasoningSignature?: string
+  readonly textSignature?: string
 }
 
 // =============================================================================
@@ -284,12 +288,22 @@ const lowerToolCall = (part: ToolCallPart) => ({
 
 const lowerMessages = Effect.fn("Gemini.lowerMessages")(function* (request: LLMRequest) {
   const contents: GeminiContent[] = []
+  const legacyToolMedia = routesLegacyToolMedia(request.model.id)
+  let pendingMedia: GeminiInlineDataPart[] | undefined
+  const flushMedia = () => {
+    if (!pendingMedia) return
+    contents.push({ role: "user", parts: [{ text: "Attached media from tool result:" }, ...pendingMedia] })
+    pendingMedia = undefined
+  }
 
   for (const message of request.messages) {
+    if (message.role !== "tool") flushMedia()
     if (message.role === "system") {
       const part = yield* ProviderShared.wrappedSystemUpdate("Gemini", message)
       const previous = contents.at(-1)
-      if (previous?.role === "user")
+      // Gemini rejects a continuation whose function-response turn carries extra
+      // parts, so an update after a tool result starts its own user turn.
+      if (previous?.role === "user" && !previous.parts.some((item) => "functionResponse" in item))
         contents[contents.length - 1] = { role: "user", parts: [...previous.parts, { text: part.text }] }
       else contents.push({ role: "user", parts: [{ text: part.text }] })
       continue
@@ -314,7 +328,7 @@ const lowerMessages = Effect.fn("Gemini.lowerMessages")(function* (request: LLMR
         if (!ProviderShared.supportsContent(part, ["text", "reasoning", "tool-call"]))
           return yield* ProviderShared.unsupportedContent("Gemini", "assistant", ["text", "reasoning", "tool-call"])
         if (part.type === "text") {
-          parts.push({ text: part.text })
+          parts.push({ text: part.text, thoughtSignature: thoughtSignature(part.providerMetadata) })
           continue
         }
         if (part.type === "reasoning") {
@@ -365,6 +379,7 @@ const lowerMessages = Effect.fn("Gemini.lowerMessages")(function* (request: LLMR
         const value = ProviderShared.normalizeToolFile(item)
         media.push({ inlineData: { mimeType: value.mime, data: value.base64 } })
       }
+      if (legacyToolMedia && media.length > 0) (pendingMedia ??= []).push(...media)
       parts.push({
         functionResponse: {
           id: functionCallId(part.providerMetadata),
@@ -373,13 +388,19 @@ const lowerMessages = Effect.fn("Gemini.lowerMessages")(function* (request: LLMR
             name: part.name,
             content: text.join("\n"),
           },
-          parts: media.length > 0 ? media : undefined,
+          parts: legacyToolMedia || media.length === 0 ? undefined : media,
         },
       })
     }
-    contents.push({ role: "user", parts })
+    // Gemini requires every response to a parallel call batch in one user turn,
+    // so consecutive tool results join the open function-response turn.
+    const previous = contents.at(-1)
+    if (previous?.role === "user" && previous.parts.some((item) => "functionResponse" in item))
+      contents[contents.length - 1] = { role: "user", parts: [...previous.parts, ...parts] }
+    else contents.push({ role: "user", parts })
   }
 
+  flushMedia()
   return contents
 })
 
@@ -519,14 +540,16 @@ const finish = (state: ParserState): ReadonlyArray<LLMEvent> => {
   if (finishReason === undefined && state.usage === undefined) return []
 
   const events: LLMEvent[] = []
-  const lifecycle = state.reasoningSignature
-    ? Lifecycle.reasoningEnd(
-        state.lifecycle,
-        events,
-        "reasoning-0",
-        googleMetadata({ thoughtSignature: state.reasoningSignature }),
-      )
-    : state.lifecycle
+  let lifecycle = state.lifecycle
+  if (state.reasoningSignature !== undefined)
+    lifecycle = Lifecycle.reasoningEnd(
+      lifecycle,
+      events,
+      "reasoning-0",
+      googleMetadata({ thoughtSignature: state.reasoningSignature }),
+    )
+  if (state.textSignature !== undefined)
+    lifecycle = Lifecycle.textEnd(lifecycle, events, "text-0", googleMetadata({ thoughtSignature: state.textSignature }))
   Lifecycle.finish(lifecycle, events, {
     reason: {
       normalized:
@@ -556,12 +579,15 @@ const step = (state: ParserState, event: GeminiEvent) => {
   const events: LLMEvent[] = []
   let hasToolCalls = nextState.hasToolCalls
   let lifecycle = nextState.lifecycle
-  let nextToolCallId = nextState.nextToolCallId
   let reasoningSignature = nextState.reasoningSignature
+  let textSignature = nextState.textSignature
 
   for (const part of candidate.content.parts) {
-    if ("thoughtSignature" in part && part.thoughtSignature && "thought" in part && part.thought)
-      reasoningSignature = part.thoughtSignature
+    const signature = "thoughtSignature" in part && part.thoughtSignature ? part.thoughtSignature : undefined
+    // Gemini attaches replay signatures to thought parts, visible text, or function calls;
+    // each block kind must retain the signature attached to its own parts.
+    if (signature !== undefined && "thought" in part && part.thought) reasoningSignature = signature
+    else if (signature !== undefined && "text" in part) textSignature = signature
     if ("text" in part && part.text.length > 0) {
       if (part.thought) {
         lifecycle = Lifecycle.reasoningDelta(
@@ -569,7 +595,7 @@ const step = (state: ParserState, event: GeminiEvent) => {
           events,
           "reasoning-0",
           part.text,
-          part.thoughtSignature ? googleMetadata({ thoughtSignature: part.thoughtSignature }) : undefined,
+          signature ? googleMetadata({ thoughtSignature: signature }) : undefined,
         )
         continue
       }
@@ -579,13 +605,22 @@ const step = (state: ParserState, event: GeminiEvent) => {
         "reasoning-0",
         reasoningSignature ? googleMetadata({ thoughtSignature: reasoningSignature }) : undefined,
       )
-      lifecycle = Lifecycle.textDelta(lifecycle, events, "text-0", part.text)
+      lifecycle = Lifecycle.textDelta(
+        lifecycle,
+        events,
+        "text-0",
+        part.text,
+        textSignature ? googleMetadata({ thoughtSignature: textSignature }) : undefined,
+      )
+      textSignature = undefined
       continue
     }
 
     if ("functionCall" in part) {
       const input = part.functionCall.args === undefined ? {} : part.functionCall.args
-      const id = `tool_${nextToolCallId++}`
+      // Gemini 2.0+ and Vertex supply a unique function call ID on the part; when omitted (e.g. Gemini 1.5),
+      // generate a globally unique ID rather than a per-request counter to prevent cross-request collisions in downstream registries.
+      const id = part.functionCall.id ?? `tool_${crypto.randomUUID().replaceAll("-", "")}`
       const metadata = {
         ...(part.functionCall.id === undefined ? {} : { functionCallId: part.functionCall.id }),
         ...(part.thoughtSignature === undefined ? {} : { thoughtSignature: part.thoughtSignature }),
@@ -614,8 +649,8 @@ const step = (state: ParserState, event: GeminiEvent) => {
       ...nextState,
       hasToolCalls,
       lifecycle,
-      nextToolCallId,
       reasoningSignature,
+      textSignature,
       finishReason: candidate.finishReason ?? nextState.finishReason,
     },
     events,
@@ -637,7 +672,7 @@ export const protocol = Protocol.make({
   },
   stream: {
     event: Protocol.jsonEvent(GeminiEvent),
-    initial: () => ({ hasToolCalls: false, nextToolCallId: 0, lifecycle: Lifecycle.initial() }),
+    initial: () => ({ hasToolCalls: false, lifecycle: Lifecycle.initial() }),
     step,
     onHalt: finish,
   },

packages/ai/src/protocols/index.ts

@@ -8,3 +8,4 @@ export * as OpenAICompatibleResponses from "./openai-compatible-responses.js"
 export * as OpenAIResponses from "./openai-responses.js"
 export * as OpenResponses from "./open-responses.js"
 export * as OpenResponsesChannel from "./open-responses-channel.js"
+export * as XAIResponses from "./xai-responses.js"

packages/ai/src/protocols/open-responses-channel.ts

@@ -10,6 +10,7 @@ import {
 } from "../route/transport/index.js"
 import * as ProviderShared from "./shared.js"
 import { OpenResponses } from "./open-responses.js"
+import { OpenResponsesContinuation } from "./open-responses-continuation.js"
 
 const WebSocketResponseCreate = Schema.StructWithRest(Schema.Struct({ type: Schema.tag("response.create") }), [
   Schema.Record(Schema.String, Schema.Unknown),
@@ -22,12 +23,9 @@ export interface Options {
   readonly id: string
   readonly name: string
   readonly rotateAfterMs?: number
+  readonly enabled?: (url: string) => boolean
+  readonly url?: (url: string) => string
   readonly headers?: (headers: Headers.Headers) => Headers.Headers
-  readonly driver?: (input: {
-    readonly request: Readonly<Record<string, unknown>>
-    readonly message: string
-    readonly base: WebSocketChannelDriver
-  }) => WebSocketChannelDriver
 }
 
 export interface Prepared {
@@ -147,18 +145,25 @@ export const transport = <Body>(options: Options): Transport<Body, Prepared, str
       Effect.gen(function* () {
         const parts = yield* HttpTransport.jsonRequestParts(input)
         const headers = Headers.remove(options.headers?.(parts.headers) ?? parts.headers, "content-length")
-        const channel = input.webSocket
-          ? yield* Effect.gen(function* () {
-              const create = yield* message(parts.jsonBody)
-              const base = driver(options, create.message)
-              return {
-                url: yield* WebSocketTransport.toWebSocketUrl(parts.url),
-                headers,
-                rotateAfterMs: options.rotateAfterMs,
-                driver: options.driver?.({ request: create.request, message: create.message, base }) ?? base,
-              }
-            })
-          : undefined
+        const channel =
+          input.webSocket && (options.enabled?.(parts.url) ?? true)
+            ? yield* Effect.gen(function* () {
+                const create = yield* message(parts.jsonBody)
+                const base = driver(options, create.message)
+                return {
+                  url: yield* WebSocketTransport.toWebSocketUrl(options.url?.(parts.url) ?? parts.url),
+                  headers,
+                  rotateAfterMs: options.rotateAfterMs,
+                  driver: OpenResponsesContinuation.driver({
+                    id: options.id,
+                    name: options.name,
+                    request: create.request,
+                    message: create.message,
+                    base,
+                  }),
+                }
+              })
+            : undefined
         return {
           http: {
             request: ProviderShared.jsonPost({ url: parts.url, body: parts.bodyText, headers: parts.headers }),

packages/ai/src/protocols/open-responses-continuation.ts

@@ -4,7 +4,7 @@ import { Effect, Option, Schema } from "effect"
 import * as ProviderShared from "./shared.js"
 import { OpenResponses } from "./open-responses.js"
 
-const PROTOCOL = "openai-responses.websocket.v1"
+const PROTOCOL = "open-responses.websocket.v1"
 const VERSION = 1
 const decodeEvent = Schema.decodeUnknownEffect(OpenResponses.protocol.stream.event)
 
@@ -161,4 +161,4 @@ export const driver = (input: DriverInput): WebSocketChannelDriver => {
   }
 }
 
-export const OpenAIResponsesChannel = { driver } as const
+export const OpenResponsesContinuation = { driver } as const

packages/ai/src/protocols/open-responses.ts

@@ -42,8 +42,12 @@ const OpenResponsesInputImage = Schema.Struct({
 const OpenResponsesInputFile = Schema.Struct({
   type: Schema.tag("input_file"),
   filename: Schema.String,
-  file_data: Schema.String,
-  mime_type: Schema.optional(Schema.String),
+  file_data: Schema.optional(Schema.String),
+  file_url: Schema.optional(Schema.String),
+})
+const OpenResponsesInputVideo = Schema.Struct({
+  type: Schema.tag("input_video"),
+  video_url: Schema.String,
 })
 const MediaInput = Schema.Union([OpenResponsesInputImage, OpenResponsesInputFile])
 export type MediaInput = Schema.Schema.Type<typeof MediaInput>
@@ -54,9 +58,14 @@ const OpenResponsesOutputText = Schema.Struct({
   text: Schema.String,
 })
 
-export const MessagePhase = Schema.Literals(["commentary", "final_answer"])
+export const MessagePhase = Schema.NullOr(Schema.Literals(["commentary", "final_answer"]))
 type MessagePhase = Schema.Schema.Type<typeof MessagePhase>
 
+const messagePhase = (value: unknown): MessagePhase | undefined => {
+  if (value === null || value === "commentary" || value === "final_answer") return value
+  return undefined
+}
+
 const OpenResponsesReasoningSummaryText = Schema.Struct({
   type: Schema.tag("summary_text"),
   text: Schema.String,
@@ -81,6 +90,7 @@ const OpenResponsesFunctionCallOutputContent = Schema.Union([
   OpenResponsesInputText,
   OpenResponsesInputImage,
   OpenResponsesInputFile,
+  OpenResponsesInputVideo,
 ])
 
 const OpenResponsesFunctionCallOutput = Schema.Union([
@@ -93,6 +103,8 @@ export const InputItem = Schema.Union([
   Schema.Struct({ role: Schema.tag("developer"), content: Schema.String }),
   Schema.Struct({ role: Schema.tag("user"), content: Schema.Array(OpenResponsesInputContent) }),
   Schema.Struct({
+    type: Schema.tag("message"),
+    id: Schema.optionalKey(Schema.String),
     role: Schema.tag("assistant"),
     content: Schema.Array(OpenResponsesOutputText),
     phase: Schema.optionalKey(MessagePhase),
@@ -101,6 +113,7 @@ export const InputItem = Schema.Union([
   OpenResponsesItemReference,
   Schema.Struct({
     type: Schema.tag("function_call"),
+    id: Schema.optionalKey(Schema.String),
     call_id: Schema.String,
     name: Schema.String,
     arguments: Schema.String,
@@ -115,6 +128,8 @@ type OpenResponsesInputItem = Schema.Schema.Type<typeof InputItem>
 type LoweredInputItem =
   | OpenResponsesInputItem
   | {
+      readonly type: "message"
+      readonly id?: string
       readonly role: "assistant"
       readonly content: ReadonlyArray<{ readonly type: "output_text"; readonly text: string }>
       readonly phase?: MessagePhase | null
@@ -128,8 +143,6 @@ type OpenResponsesReasoningInput = {
   summary: Array<{ type: "summary_text"; text: string }>
   encrypted_content?: string | null
 }
-type OpenResponsesReasoningReplay = Omit<OpenResponsesReasoningInput, "id">
-
 export const Tool = Schema.Struct({
   type: Schema.tag("function"),
   name: Schema.String,
@@ -159,6 +172,14 @@ export const coreFields = {
   tools: optionalArray(Tool),
   tool_choice: Schema.optional(ToolChoice),
   store: Schema.optional(Schema.Boolean),
+  metadata: Schema.optional(Schema.Record(Schema.String, Schema.String)),
+  safety_identifier: Schema.optional(Schema.String),
+  stream_options: Schema.optional(
+    Schema.Struct({
+      include_obfuscation: Schema.optional(Schema.Boolean),
+    }),
+  ),
+  top_logprobs: Schema.optional(Schema.Int.check(Schema.isBetween({ minimum: 0, maximum: 20 }))),
   truncation: Schema.optional(OpenResponsesOptions.TruncationSchema),
   service_tier: Schema.optional(OpenResponsesOptions.ServiceTierSchema),
   prompt_cache_key: Schema.optional(Schema.String),
@@ -179,6 +200,8 @@ export const coreFields = {
   parallel_tool_calls: Schema.optional(Schema.Boolean),
   temperature: Schema.optional(Schema.Number),
   top_p: Schema.optional(Schema.Number),
+  presence_penalty: Schema.optional(Schema.Number),
+  frequency_penalty: Schema.optional(Schema.Number),
 }
 
 const OpenResponsesBody = Schema.Struct({
@@ -224,6 +247,7 @@ const OpenResponsesErrorPayload = Schema.Struct({
   message: optionalNull(Schema.String),
   param: optionalNull(Schema.String),
 })
+type OpenResponsesErrorPayload = Schema.Schema.Type<typeof OpenResponsesErrorPayload>
 
 const WebSocketErrorHeader = Schema.Union([Schema.String, Schema.Number, Schema.Boolean])
 export const WebSocketErrorEvent = Schema.StructWithRest(
@@ -296,7 +320,6 @@ export interface Extension {
     readonly media: ProviderShared.NormalizedMedia
     readonly request: LLMRequest
   }) => MediaInput | undefined
-  readonly messagePhase?: (value: unknown) => MessagePhase | null | undefined
 }
 
 const BASE: Extension = { id: ADAPTER, name: NAME }
@@ -309,7 +332,6 @@ export interface ParserState {
   readonly hasFunctionCall: boolean
   readonly lifecycle: Lifecycle.State
   readonly messageItems: ReadonlySet<string>
-  readonly messagePhase: (value: unknown) => MessagePhase | null | undefined
   readonly messagePhases: Readonly<Record<string, MessagePhase | null>>
   readonly reasoningItems: Readonly<Record<string, ReasoningStreamItem>>
   readonly store: boolean | undefined
@@ -353,52 +375,63 @@ export const lowerToolChoice = (protocolName: string, toolChoice: NonNullable<LL
     tool: (toolName) => ({ type: "function" as const, name: toolName }),
   })
 
-const lowerToolCall = (part: ToolCallPart): OpenResponsesInputItem => ({
-  type: "function_call",
-  call_id: part.id,
-  name: part.name,
-  arguments: ProviderShared.encodeJson(part.input),
-})
+const itemID = (providerMetadata: ProviderMetadata | undefined, providerMetadataKey: string) => {
+  const metadata = providerMetadata?.[providerMetadataKey]
+  return ProviderShared.isRecord(metadata) && typeof metadata.itemId === "string" && metadata.itemId.length > 0
+    ? metadata.itemId
+    : undefined
+}
+
+const lowerToolCall = (part: ToolCallPart, providerMetadataKey: string): OpenResponsesInputItem => {
+  const id = itemID(part.providerMetadata, providerMetadataKey)
+  return {
+    type: "function_call",
+    ...(id ? { id } : {}),
+    call_id: part.id,
+    name: part.name,
+    arguments: ProviderShared.encodeJson(part.input),
+  }
+}
 
 const lowerReasoning = (part: ReasoningPart, providerMetadataKey: string): OpenResponsesReasoningInput | undefined => {
   const metadata = part.providerMetadata?.[providerMetadataKey]
-  if (!ProviderShared.isRecord(metadata) || typeof metadata.itemId !== "string" || metadata.itemId.length === 0)
-    return undefined
+  const id = itemID(part.providerMetadata, providerMetadataKey)
+  if (!ProviderShared.isRecord(metadata) || !id) return undefined
   const encryptedContent =
     typeof metadata.reasoningEncryptedContent === "string" || metadata.reasoningEncryptedContent === null
       ? metadata.reasoningEncryptedContent
       : undefined
   return {
     type: "reasoning",
-    id: metadata.itemId,
+    id,
     summary: part.text.length > 0 ? [{ type: "summary_text", text: part.text }] : [],
     encrypted_content: encryptedContent,
   }
 }
 
-const hostedToolItemID = (part: ToolResultPart, providerMetadataKey: string) => {
-  const metadata = part.providerMetadata?.[providerMetadataKey]
-  return ProviderShared.isRecord(metadata) && typeof metadata.itemId === "string" && metadata.itemId.length > 0
-    ? metadata.itemId
-    : undefined
-}
-
 const lowerMedia = Effect.fn("OpenResponses.lowerMedia")(function* (
   part: MediaPart,
   request: LLMRequest,
   extension: Extension,
+  target: "message" | "tool-result",
 ) {
   const media = ProviderShared.normalizeMedia(part)
   const extended = extension.lowerMedia?.({ part, media, request })
   if (extended) return extended
+  const url =
+    typeof part.data === "string" && (part.data.startsWith("https://") || part.data.startsWith("http://"))
+      ? part.data
+      : undefined
   if (!media.mime.startsWith("image/")) {
+    if (target === "tool-result" && media.mime.startsWith("video/"))
+      return { type: "input_video" as const, video_url: url ?? media.dataUrl }
     return {
       type: "input_file" as const,
       filename: part.filename ?? (media.mime === "application/pdf" ? "document.pdf" : "file"),
-      file_data: media.dataUrl,
+      ...(url ? { file_url: url } : { file_data: media.base64 }),
     }
   }
-  return { type: "input_image" as const, image_url: media.dataUrl }
+  return { type: "input_image" as const, image_url: url ?? media.dataUrl }
 })
 
 const lowerUserContent = Effect.fnUntraced(function* (
@@ -407,10 +440,17 @@ const lowerUserContent = Effect.fnUntraced(function* (
   extension: Extension,
 ) {
   if (part.type === "text") return { type: "input_text" as const, text: part.text }
-  if (part.type === "media") return yield* lowerMedia(part, request, extension)
+  if (part.type === "media") return yield* lowerMessageMedia(part, request, extension)
   return yield* ProviderShared.unsupportedContent(extension.name, "user", ["text", "media"])
 })
 
+const lowerMessageMedia = Effect.fnUntraced(function* (part: MediaPart, request: LLMRequest, extension: Extension) {
+  const lowered = yield* lowerMedia(part, request, extension, "message")
+  if (lowered.type === "input_video")
+    return yield* ProviderShared.invalidRequest(`${extension.name} user messages do not support input_video`)
+  return lowered
+})
+
 // Tool results may carry structured text, images, and files. Keep media as provider-native
 // content instead of JSON-stringifying base64 into a prompt string.
 const lowerToolResultContentItem = Effect.fnUntraced(function* (
@@ -423,6 +463,20 @@ const lowerToolResultContentItem = Effect.fnUntraced(function* (
     { type: "media", mediaType: item.mime, data: item.uri, filename: item.name },
     request,
     extension,
+    "tool-result",
+  )
+})
+
+const lowerHostedToolResultContentItem = Effect.fnUntraced(function* (
+  item: Content,
+  request: LLMRequest,
+  extension: Extension,
+) {
+  if (item.type === "text") return { type: "input_text" as const, text: item.text }
+  return yield* lowerMessageMedia(
+    { type: "media", mediaType: item.mime, data: item.uri, filename: item.name },
+    request,
+    extension,
   )
 })
 
@@ -465,24 +519,26 @@ const lowerMessages = Effect.fn("OpenResponses.lowerMessages")(function* (reques
 
     if (message.role === "assistant") {
       const content: TextPart[] = []
-      const reasoningItems: Record<string, OpenResponsesReasoningReplay> = {}
+      const reasoningItems: Record<string, OpenResponsesReasoningInput> = {}
       const reasoningReferences = new Set<string>()
       const hostedToolReferences = new Set<string>()
       const flushText = () => {
         if (content.length === 0) return
-        const groups = content.reduce<Array<{ phase: MessagePhase | null | undefined; parts: TextPart[] }>>(
-          (groups, part) => {
-            const metadata = part.providerMetadata?.[providerMetadataKey]
-            const phase = ProviderShared.isRecord(metadata) ? messagePhase(metadata.phase, extension) : undefined
-            const group = groups.at(-1)
-            if (group && group.phase === phase) group.parts.push(part)
-            else groups.push({ phase, parts: [part] })
-            return groups
-          },
-          [],
-        )
+        const groups = content.reduce<
+          Array<{ id: string | undefined; phase: MessagePhase | null | undefined; parts: TextPart[] }>
+        >((groups, part) => {
+          const metadata = part.providerMetadata?.[providerMetadataKey]
+          const id = itemID(part.providerMetadata, providerMetadataKey)
+          const phase = ProviderShared.isRecord(metadata) ? messagePhase(metadata.phase) : undefined
+          const group = groups.at(-1)
+          if (group && group.id === id && group.phase === phase) group.parts.push(part)
+          else groups.push({ id, phase, parts: [part] })
+          return groups
+        }, [])
         input.push(
           ...groups.map((group) => ({
+            type: "message" as const,
+            ...(group.id === undefined ? {} : { id: group.id }),
             role: "assistant" as const,
             content: group.parts.map((part) => ({ type: "output_text" as const, text: part.text })),
             ...(group.phase === undefined ? {} : { phase: group.phase }),
@@ -511,34 +567,31 @@ const lowerMessages = Effect.fn("OpenResponses.lowerMessages")(function* (reques
               existing.encrypted_content = reasoning.encrypted_content
             continue
           }
-          const replay = {
-            type: reasoning.type,
-            summary: reasoning.summary,
-            encrypted_content: reasoning.encrypted_content,
-          }
-          reasoningItems[reasoning.id] = replay
-          input.push(replay)
+          reasoningItems[reasoning.id] = reasoning
+          input.push(reasoning)
           continue
         }
         if (part.type === "tool-call") {
           flushText()
           if (part.providerExecuted === true) continue
-          input.push(lowerToolCall(part))
+          input.push(lowerToolCall(part, providerMetadataKey))
           continue
         }
         if (part.type === "tool-result" && part.providerExecuted === true) {
           flushText()
-          const itemID = hostedToolItemID(part, providerMetadataKey)
-          if (store !== false && itemID && !hostedToolReferences.has(itemID))
-            input.push({ type: "item_reference", id: itemID })
+          const id = itemID(part.providerMetadata, providerMetadataKey)
+          if (store !== false && id && !hostedToolReferences.has(id))
+            input.push({ type: "item_reference", id })
           if (store === false && part.result.type === "content") {
             const content: ReadonlyArray<Content> = part.result.value
             input.push({
               role: "user",
-              content: yield* Effect.forEach(content, (item) => lowerToolResultContentItem(item, request, extension)),
+              content: yield* Effect.forEach(content, (item) =>
+                lowerHostedToolResultContentItem(item, request, extension),
+              ),
             })
           }
-          if (itemID) hostedToolReferences.add(itemID)
+          if (id) hostedToolReferences.add(id)
           continue
         }
         return yield* ProviderShared.unsupportedContent(extension.name, "assistant", [
@@ -578,6 +631,12 @@ const lowerOptions = (request: LLMRequest) => {
   return {
     ...(options.instructions ? { instructions: options.instructions } : {}),
     ...(options.store !== undefined ? { store: options.store } : {}),
+    ...(options.metadata ? { metadata: options.metadata } : {}),
+    ...(options.safetyIdentifier ? { safety_identifier: options.safetyIdentifier } : {}),
+    ...(options.streamOptions?.includeObfuscation !== undefined
+      ? { stream_options: { include_obfuscation: options.streamOptions.includeObfuscation } }
+      : {}),
+    ...(options.topLogprobs !== undefined ? { top_logprobs: options.topLogprobs } : {}),
     ...(request.promptCacheKey ? { prompt_cache_key: request.promptCacheKey } : {}),
     ...(options.include ? { include: options.include } : {}),
     ...(options.reasoningEffort || options.reasoningSummary
@@ -627,6 +686,8 @@ export const fromRequestWithExtension = Effect.fn("OpenResponses.fromRequestWith
     max_output_tokens: generation?.maxTokens,
     temperature: generation?.temperature,
     top_p: generation?.topP,
+    presence_penalty: generation?.presencePenalty,
+    frequency_penalty: generation?.frequencyPenalty,
     ...lowerOptions(request),
   }
 })
@@ -695,7 +756,7 @@ const onOutputTextDelta = (state: ParserState, event: Event, id: string): StepRe
   if (!event.delta) return [state, NO_EVENTS]
   const events: LLMEvent[] = []
   const phase = state.messagePhases[id]
-  const metadata = phase === undefined ? undefined : providerMetadata(state, { phase })
+  const metadata = providerMetadata(state, { itemId: id, ...(phase === undefined ? {} : { phase }) })
   const lifecycle = Lifecycle.textStart(state.lifecycle, events, id, metadata)
   return [{ ...state, lifecycle: Lifecycle.textDelta(lifecycle, events, id, event.delta) }, events]
 }
@@ -742,18 +803,17 @@ const reasoningMetadata = (state: ParserState, item: StreamItem & { id: string }
 // best-effort, not guaranteed.
 const onOutputItemAdded = (state: ParserState, event: Event): StepResult => {
   const item = event.item
-  if (item?.type === "message" && item.id)
+  if (item?.type === "message" && item.id) {
+    const phase = messagePhase(item.phase)
     return [
       {
         ...state,
         messageItems: new Set([...state.messageItems, item.id]),
-        messagePhases: (() => {
-          const phase = state.messagePhase(item.phase)
-          return phase === undefined ? state.messagePhases : { ...state.messagePhases, [item.id]: phase }
-        })(),
+        messagePhases: phase === undefined ? state.messagePhases : { ...state.messagePhases, [item.id]: phase },
       },
       NO_EVENTS,
     ]
+  }
   if (item && isReasoningItem(item)) {
     const events: LLMEvent[] = []
     return [
@@ -911,7 +971,7 @@ const onOutputItemDone = Effect.fn("OpenResponses.onOutputItemDone")(function* (
   if (!item) return [state, NO_EVENTS] satisfies StepResult
 
   if (item.type === "message" && item.id) {
-    const itemPhase = state.messagePhase(item.phase)
+    const itemPhase = messagePhase(item.phase)
     const phase = itemPhase === undefined ? state.messagePhases[item.id] : itemPhase
     const events: LLMEvent[] = []
     const messageItems = new Set(state.messageItems)
@@ -924,7 +984,7 @@ const onOutputItemDone = Effect.fn("OpenResponses.onOutputItemDone")(function* (
           state.lifecycle,
           events,
           item.id,
-          phase === undefined ? undefined : providerMetadata(state, { phase }),
+          providerMetadata(state, { itemId: item.id, ...(phase === undefined ? {} : { phase }) }),
         ),
         messageItems,
         messagePhases,
@@ -937,7 +997,11 @@ const onOutputItemDone = Effect.fn("OpenResponses.onOutputItemDone")(function* (
     if (!item.id || !item.call_id || !item.name) return [state, NO_EVENTS] satisfies StepResult
     const tools = state.tools[item.id]
       ? state.tools
-      : ToolStream.start(state.tools, item.id, { id: item.call_id, name: item.name })
+      : ToolStream.start(state.tools, item.id, {
+          id: item.call_id,
+          name: item.name,
+          providerMetadata: providerMetadata(state, { itemId: item.id }),
+        })
     const result =
       item.arguments === undefined
         ? yield* ToolStream.finish(state.id, tools, item.id)
@@ -988,11 +1052,19 @@ const onOutputItemDone = Effect.fn("OpenResponses.onOutputItemDone")(function* (
   return [state, NO_EVENTS] satisfies StepResult
 })
 
-const onResponseFinish = (state: ParserState, event: Event): StepResult => {
-  const events: LLMEvent[] = []
+const onResponseFinish = Effect.fn("OpenResponses.onResponseFinish")(function* (state: ParserState, event: Event) {
+  // Some compatible providers omit output_item.done even after completing the response.
+  const pending =
+    event.type === "response.completed"
+      ? yield* ToolStream.finishAll(state.id, state.tools)
+      : { tools: state.tools, events: NO_EVENTS }
+  const events: LLMEvent[] = [...pending.events]
+  const hasFunctionCall =
+    pending.events.some((event) => LLMEvent.is.toolCall(event) || LLMEvent.is.toolInputError(event)) ||
+    state.hasFunctionCall
   const lifecycle = Lifecycle.finish(state.lifecycle, events, {
     reason: {
-      normalized: mapFinishReason(event, state.hasFunctionCall),
+      normalized: mapFinishReason(event, hasFunctionCall),
       raw: event.response?.incomplete_details?.reason,
     },
     usage: mapUsage(event.response?.usage, state.providerMetadataKey),
@@ -1004,25 +1076,29 @@ const onResponseFinish = (state: ParserState, event: Event): StepResult => {
           })
         : undefined,
   })
-  return [{ ...state, lifecycle }, events]
-}
+  return [{ ...state, lifecycle, hasFunctionCall, tools: pending.tools }, events] satisfies StepResult
+})
 
-// Build a single human-readable message from whatever the provider supplied.
+// Build the prettiest summary available from whatever the provider supplied.
 // When both code and message are present, prefix the code so consumers see
 // the failure mode (e.g. `rate_limit_exceeded: Slow down`) instead of just
 // the bare message — production rate limits and context-length failures used
-// to be indistinguishable from generic stream drops.
-const providerErrorMessage = (event: Event, fallback: string): string => {
-  const nested = event.error ?? event.response?.error ?? undefined
+// to be indistinguishable from generic stream drops. Returns undefined when
+// the payload carries no usable summary.
+const providerErrorMessage = (event: Event, nested: OpenResponsesErrorPayload | undefined): string | undefined => {
   const message = event.message || nested?.message || undefined
   const code = event.code || nested?.code || undefined
   if (message && code) return `${code}: ${message}`
-  return message || code || fallback
+  return message || code
 }
 
 export const providerFailure = (id: string, event: Event, fallback: string) => {
-  const code = event.code || event.error?.code || event.response?.error?.code || undefined
-  const message = providerErrorMessage(event, fallback)
+  const nested = event.error ?? event.response?.error ?? undefined
+  const code = event.code || nested?.code || undefined
+  // Keep the full raw payload on the error even when the message is a summary.
+  const body = JSON.stringify(nested ?? event) ?? ""
+  const summary = providerErrorMessage(event, nested)
+  const message = summary ?? (body === "{}" ? fallback : body)
   const status =
     typeof event.status === "number"
       ? event.status
@@ -1032,7 +1108,8 @@ export const providerFailure = (id: string, event: Event, fallback: string) => {
   return new AIError({
     module: id,
     method: "stream",
-    reason: classifyProviderFailure({ message, code, status }),
+    body,
+    reason: classifyProviderFailure({ message, code, status, rawBody: body }),
   })
 }
 
@@ -1047,14 +1124,19 @@ export const step = (state: ParserState, event: Event) => {
         : onOutputTextDone(state, event, event.item_id),
     )
   }
+  if (event.type === "response.refusal.delta" || event.type === "response.refusal.done") {
+    const value = event.type === "response.refusal.delta" ? event.delta : event.refusal
+    if (!event.item_id || typeof value !== "string") return ProviderShared.eventError(state.id, `${event.type} is malformed`)
+    return Effect.succeed(
+      event.type === "response.refusal.delta"
+        ? onOutputTextDelta(state, event, event.item_id)
+        : onOutputTextDone(state, { ...event, text: value }, event.item_id),
+    )
+  }
   if (event.type === "response.reasoning.delta" || event.type === "response.reasoning_summary_text.delta") {
     if (!event.item_id) return ProviderShared.eventError(state.id, `${event.type} is missing item_id`)
     return Effect.succeed(onReasoningDelta(state, event, event.item_id))
   }
-  if (event.type === "response.reasoning.done" || event.type === "response.reasoning_summary_text.done") {
-    if (!event.item_id) return ProviderShared.eventError(state.id, `${event.type} is missing item_id`)
-    return Effect.succeed(onReasoningDone(state, event))
-  }
   if (event.type === "response.reasoning_summary_part.added")
     return event.item_id
       ? Effect.succeed(onReasoningSummaryPartAdded(state, event))
@@ -1074,8 +1156,7 @@ export const step = (state: ParserState, event: Event) => {
       return ProviderShared.eventError(state.id, `${event.type} message is missing id`)
     return onOutputItemDone(state, event)
   }
-  if (event.type === "response.completed" || event.type === "response.incomplete")
-    return Effect.succeed(onResponseFinish(state, event))
+  if (event.type === "response.completed" || event.type === "response.incomplete") return onResponseFinish(state, event)
   if (event.type === "response.failed") return providerError(state, event, `${state.name} response failed`)
   if (event.type === "error")
     return decodeKnownErrorEvent(event).pipe(
@@ -1100,17 +1181,11 @@ export const initial = (request: LLMRequest, extension: Extension = BASE): Parse
   tools: ToolStream.empty<string>(),
   lifecycle: Lifecycle.initial(),
   messageItems: new Set<string>(),
-  messagePhase: (value) => messagePhase(value, extension),
   messagePhases: {},
   reasoningItems: {},
   store: OpenResponsesOptions.resolve(request).store,
 })
 
-const messagePhase = (value: unknown, extension: Extension): MessagePhase | null | undefined => {
-  if (value === "commentary" || value === "final_answer") return value
-  return extension.messagePhase?.(value)
-}
-
 export const protocol = Protocol.make({
   id: ADAPTER,
   body: {

packages/ai/src/protocols/openai-chat.ts

@@ -28,7 +28,7 @@ import { ToolSchemaProjection } from "./utils/tool-schema.js"
 import { ToolStream } from "./utils/tool-stream.js"
 
 const ADAPTER = "openai-chat"
-const RESERVED_REASONING_FIELDS = new Set(["role", "content", "tool_calls"])
+const RESERVED_REASONING_FIELDS = new Set(["role", "content", "refusal", "tool_calls"])
 export const DEFAULT_BASE_URL = "https://api.openai.com/v1"
 export const PATH = "/chat/completions"
 
@@ -194,6 +194,7 @@ type OpenAIChatToolCallDelta = Schema.Schema.Type<typeof OpenAIChatToolCallDelta
 const OpenAIChatDelta = Schema.StructWithRest(
   Schema.Struct({
     content: optionalNull(Schema.String),
+    refusal: optionalNull(Schema.String),
     reasoning_content: optionalNull(Schema.String),
     reasoning: optionalNull(Schema.String),
     reasoning_text: optionalNull(Schema.String),
@@ -709,6 +710,7 @@ const step = (state: ParserState, event: OpenAIChatEvent) =>
     const reasoning = reasoningDelta(delta, state.reasoningField)
     const hasLateContent =
       Boolean(delta?.content) ||
+      Boolean(delta?.refusal) ||
       reasoning !== undefined ||
       (Array.isArray(delta?.reasoning_details) && delta.reasoning_details.length > 0) ||
       toolDeltas.some((tool) => Boolean(tool.id) || Boolean(tool.function?.name) || Boolean(tool.function?.arguments))
@@ -728,7 +730,7 @@ const step = (state: ParserState, event: OpenAIChatEvent) =>
     else if (
       reasoningDetailsObserved &&
       !lifecycle.reasoning.has("reasoning-0") &&
-      (Boolean(delta?.content) || toolDeltas.length > 0)
+      (Boolean(delta?.content) || Boolean(delta?.refusal) || toolDeltas.length > 0)
     )
       lifecycle = Lifecycle.reasoningStart(lifecycle, events, "reasoning-0", deltaMetadata)
     const reasoningEmitted = state.reasoningEmitted || lifecycle.reasoning.has("reasoning-0")
@@ -743,6 +745,16 @@ const step = (state: ParserState, event: OpenAIChatEvent) =>
       lifecycle = Lifecycle.textDelta(lifecycle, events, "text-0", delta.content)
     }
 
+    if (delta?.refusal) {
+      lifecycle = Lifecycle.reasoningEnd(
+        lifecycle,
+        events,
+        "reasoning-0",
+        reasoningMetadata(reasoningField, reasoningDetailsObserved ? state.reasoningDetails : undefined),
+      )
+      lifecycle = Lifecycle.textDelta(lifecycle, events, "text-0", delta.refusal)
+    }
+
     // Compatible providers may omit indexes. Prefer durable identity, then use
     // batch position for parallel deltas or the latest call for sparse chunks.
     for (const [position, tool] of toolDeltas.entries()) {

packages/ai/src/protocols/openai-images.ts

@@ -110,7 +110,7 @@ export const model = (input: ModelInput) => {
       const multipartImages = yield* Effect.forEach(sourceImages, (image) => {
         if (image.type === "bytes") return Effect.succeed({ data: image.data, mediaType: image.mediaType })
         if (image.type === "url") return ImageInputs.decodeDataUrl(image.url, ADAPTER)
-        return Effect.succeed(undefined)
+        return Effect.undefined
       })
       const multipartMask =
         mask === undefined

packages/ai/src/protocols/openai-responses.ts

@@ -5,14 +5,13 @@ import { Auth } from "../route/auth.js"
 import { Endpoint } from "../route/endpoint.js"
 import { Protocol } from "../route/protocol.js"
 import { HttpTransport } from "../route/transport/index.js"
-import { LLMEvent, LLMRequest, type JsonSchema, type ToolDefinition } from "../schema/index.js"
+import { LLMRequest, type JsonSchema, type ToolDefinition } from "../schema/index.js"
 import { OpenResponses } from "./open-responses.js"
 import { optionalArray, ProviderShared } from "./shared.js"
-import { Lifecycle } from "./utils/lifecycle.js"
 import { OpenAIImage } from "./utils/openai-image.js"
+import { ResponsesHostedTools } from "./utils/responses-hosted-tools.js"
 import { ToolSchemaProjection } from "./utils/tool-schema.js"
 import { OpenResponsesChannel } from "./open-responses-channel.js"
-import { OpenAIResponsesChannel } from "./openai-responses-channel.js"
 
 const ADAPTER = "openai-responses"
 const NAME = "OpenAI Responses"
@@ -40,18 +39,8 @@ const OpenAIResponsesToolChoice = Schema.Union([
   Schema.Struct({ type: Schema.tag("image_generation") }),
 ])
 
-const OpenAIResponsesInputItem = Schema.Union([
-  Schema.Struct({
-    role: Schema.tag("assistant"),
-    content: Schema.Array(Schema.Struct({ type: Schema.tag("output_text"), text: Schema.String })),
-    phase: Schema.optionalKey(Schema.NullOr(OpenResponses.MessagePhase)),
-  }),
-  OpenResponses.InputItem,
-])
-
 const OpenAIResponsesCoreFields = {
   ...OpenResponses.coreFields,
-  input: Schema.Array(OpenAIResponsesInputItem),
   tools: optionalArray(OpenAIResponsesTools),
   tool_choice: Schema.optional(OpenAIResponsesToolChoice),
 }
@@ -65,16 +54,6 @@ export type OpenAIResponsesBody = Schema.Schema.Type<typeof OpenAIResponsesBody>
 const extension = {
   id: ADAPTER,
   name: NAME,
-  messagePhase: (value: unknown) => (value === null ? null : undefined),
-  lowerMedia: ({ part, media, request }) => {
-    if (request.model.provider !== "xai" || media.mime !== "application/pdf") return undefined
-    return {
-      type: "input_file",
-      filename: part.filename ?? "document.pdf",
-      file_data: media.base64,
-      mime_type: media.mime,
-    }
-  },
 } satisfies OpenResponses.Extension
 
 const nativeImageToolInput = (tool: ToolDefinition) => {
@@ -126,46 +105,7 @@ const fromRequest = Effect.fn("OpenAIResponses.fromRequest")(function* (request:
   } satisfies OpenAIResponsesBody
 })
 
-type HostedToolData = OpenResponses.StreamItem & {
-  readonly id: string
-  readonly status?: string
-  readonly action?: unknown
-  readonly queries?: unknown
-  readonly results?: unknown
-  readonly code?: string
-  readonly container_id?: string
-  readonly outputs?: unknown
-  readonly server_label?: string
-  readonly output?: unknown
-  readonly result?: string
-  readonly output_format?: "png" | "jpeg" | "webp"
-  readonly error?: unknown
-}
-
-const HOSTED_TOOLS = {
-  web_search_call: { name: "web_search", input: (item) => item.action ?? {} },
-  web_search_preview_call: { name: "web_search_preview", input: (item) => item.action ?? {} },
-  file_search_call: { name: "file_search", input: (item) => ({ queries: item.queries ?? [] }) },
-  code_interpreter_call: {
-    name: "code_interpreter",
-    input: (item) => ({ code: item.code, container_id: item.container_id }),
-  },
-  computer_use_call: { name: "computer_use", input: (item) => item.action ?? {} },
-  image_generation_call: { name: "image_generation", input: () => ({}) },
-  mcp_call: {
-    name: "mcp",
-    input: (item) => ({ server_label: item.server_label, name: item.name, arguments: item.arguments }),
-  },
-  local_shell_call: { name: "local_shell", input: (item) => item.action ?? {} },
-} as const satisfies Record<string, { readonly name: string; readonly input: (item: HostedToolData) => unknown }>
-
-type HostedToolType = keyof typeof HOSTED_TOOLS
-type HostedToolItem = HostedToolData & { readonly type: HostedToolType }
-
-const isHostedToolItem = (item: OpenResponses.StreamItem): item is HostedToolItem =>
-  item.type in HOSTED_TOOLS && typeof item.id === "string" && item.id.length > 0
-
-const hostedToolResult = Effect.fn("OpenAIResponses.hostedToolResult")(function* (item: HostedToolItem) {
+const hostedToolResult = Effect.fn("OpenAIResponses.hostedToolResult")(function* (item: ResponsesHostedTools.Item) {
   const isError = item.error !== undefined && item.error !== null
   if (item.type === "image_generation_call" && item.result) {
     yield* Effect.fromResult(Encoding.decodeBase64(item.result)).pipe(
@@ -186,32 +126,22 @@ const hostedToolResult = Effect.fn("OpenAIResponses.hostedToolResult")(function*
   return isError ? { type: "error" as const, value: item.error } : { type: "json" as const, value: item }
 })
 
-const onHostedToolDone = Effect.fn("OpenAIResponses.onHostedToolDone")(function* (
-  state: OpenResponses.ParserState,
-  item: HostedToolItem,
-) {
-  const tool = HOSTED_TOOLS[item.type]
-  const providerMetadata = OpenResponses.providerMetadata(state, { itemId: item.id })
-  const events: LLMEvent[] = []
-  const lifecycle = Lifecycle.stepStart(state.lifecycle, events)
-  events.push(
-    LLMEvent.toolCall({
-      id: item.id,
-      name: tool.name,
-      input: tool.input(item),
-      providerExecuted: true,
-      providerMetadata,
-    }),
-    LLMEvent.toolResult({
-      id: item.id,
-      name: tool.name,
-      result: yield* hostedToolResult(item),
-      providerExecuted: true,
-      providerMetadata,
-    }),
-  )
-  return [{ ...state, lifecycle }, events] satisfies OpenResponses.StepResult
-})
+const HOSTED_TOOLS = {
+  web_search_call: { name: "web_search", input: (item) => item.action ?? {} },
+  web_search_preview_call: { name: "web_search_preview", input: (item) => item.action ?? {} },
+  file_search_call: { name: "file_search", input: (item) => ({ queries: item.queries ?? [] }) },
+  code_interpreter_call: {
+    name: "code_interpreter",
+    input: (item) => ({ code: item.code, container_id: item.container_id }),
+  },
+  computer_use_call: { name: "computer_use", input: (item) => item.action ?? {} },
+  image_generation_call: { name: "image_generation", input: () => ({}), result: hostedToolResult },
+  mcp_call: {
+    name: "mcp",
+    input: (item) => ({ server_label: item.server_label, name: item.name, arguments: item.arguments }),
+  },
+  local_shell_call: { name: "local_shell", input: (item) => item.action ?? {} },
+} as const satisfies ResponsesHostedTools.Definitions
 
 const step = (state: OpenResponses.ParserState, event: OpenResponses.Event) => {
   if (event.type === "response.reasoning_text.delta" || event.type === "response.reasoning_summary.delta")
@@ -222,8 +152,8 @@ const step = (state: OpenResponses.ParserState, event: OpenResponses.Event) => {
     return event.item_id
       ? Effect.succeed(OpenResponses.onReasoningDone(state, event))
       : ProviderShared.eventError(ADAPTER, `${event.type} is missing item_id`)
-  if (event.type === "response.output_item.done" && event.item && isHostedToolItem(event.item))
-    return onHostedToolDone(state, event.item)
+  if (event.type === "response.output_item.done" && event.item && ResponsesHostedTools.isItem(event.item, HOSTED_TOOLS))
+    return ResponsesHostedTools.onDone(state, event.item, HOSTED_TOOLS)
   return OpenResponses.step(state, event)
 }
 
@@ -245,12 +175,12 @@ const endpoint = Endpoint.path<OpenAIResponsesBody>(PATH, { baseURL: DEFAULT_BAS
 const auth = Auth.none
 
 export const httpTransport = HttpTransport.sseJson.with<OpenAIResponsesBody>()
-export const transport = OpenResponsesChannel.transport<OpenAIResponsesBody>({
+export const channelTransport = OpenResponsesChannel.transport<OpenAIResponsesBody>
+export const transport = channelTransport({
   id: ADAPTER,
   name: NAME,
   rotateAfterMs: WEBSOCKET_ROTATE_AFTER_MS,
   headers: (headers) => Headers.set(headers, "openai-beta", headers["openai-beta"] ?? WEBSOCKET_PROTOCOL_HEADER),
-  driver: (input) => OpenAIResponsesChannel.driver({ id: ADAPTER, name: NAME, ...input }),
 })
 
 export const route = Route.make({

packages/ai/src/protocols/shared.ts

@@ -197,19 +197,28 @@ export const errorText = (error: unknown) => {
 
 /**
  * `framing` step for Server-Sent Events. Decodes UTF-8, runs the SSE channel
- * decoder, and drops empty / `[DONE]` keep-alive events so the protocol event
- * schema sees one JSON string per element. The SSE channel emits a
+ * decoder, optionally filters named events, and drops empty / `[DONE]`
+ * keep-alive events so the protocol event schema sees one JSON string per
+ * element. The SSE channel emits a
  * `Retry` control event on its error channel; we drop it here (we don't
  * implement client-driven retries). Decoder failures become provider output
  * errors so the public error channel stays `AIError`.
  */
-export const sseFraming = (bytes: Stream.Stream<Uint8Array, AIError>): Stream.Stream<string, AIError> =>
+export const sseFraming = (
+  bytes: Stream.Stream<Uint8Array, AIError>,
+  events?: ReadonlySet<string>,
+): Stream.Stream<string, AIError> =>
   bytes.pipe(
     Stream.decodeText(),
     Stream.pipeThroughChannel(Sse.decode()),
     Stream.catchTag("Retry", () => Stream.empty),
     Stream.catchTag("SseError", (error) => Stream.fail(eventError("sse", error.message))),
-    Stream.filter((event) => event.data.length > 0 && event.data !== "[DONE]"),
+    Stream.filter(
+      (event) =>
+        (events === undefined || events.has(event.event)) &&
+        event.data.length > 0 &&
+        (event.data !== "[DONE]" || (events !== undefined && event.event !== "message")),
+    ),
     Stream.map((event) => event.data),
   )
 

packages/ai/src/protocols/utils/image-input.ts

@@ -16,7 +16,7 @@ export const decodeDataUrl = (
   url: string,
   module: string,
 ): Effect.Effect<{ readonly mediaType: string; readonly data: Uint8Array } | undefined, AIError> => {
-  if (!url.startsWith("data:")) return Effect.succeed(undefined)
+  if (!url.startsWith("data:")) return Effect.undefined
   const match = /^data:([^;,]+);base64,(.*)$/s.exec(url)
   if (!match) return Effect.fail(invalid(module, "Image data URLs must contain a MIME type and base64 data"))
   return Effect.fromResult(Encoding.decodeBase64(match[2])).pipe(

packages/ai/src/protocols/utils/lifecycle.ts

@@ -21,9 +21,15 @@ export const textStart = (state: State, events: LLMEvent[], id: string, provider
   return { ...stepped, text: new Set([...stepped.text, id]) }
 }
 
-export const textDelta = (state: State, events: LLMEvent[], id: string, text: string): State => {
+export const textDelta = (
+  state: State,
+  events: LLMEvent[],
+  id: string,
+  text: string,
+  providerMetadata?: ProviderMetadata,
+): State => {
   const started = textStart(state, events, id)
-  events.push(LLMEvent.textDelta({ id, text }))
+  events.push(LLMEvent.textDelta({ id, text, providerMetadata }))
   return started
 }
 

packages/ai/src/protocols/utils/open-responses-options.ts

@@ -47,9 +47,17 @@ export const AllowedTools = Schema.Struct({
 })
 export type AllowedTools = typeof AllowedTools.Type
 
+export const StreamOptions = Schema.Struct({
+  includeObfuscation: Schema.optional(Schema.Boolean),
+})
+
 export const Options = Schema.Struct({
   instructions: Schema.optional(Schema.String),
   store: Schema.optional(Schema.Boolean),
+  metadata: Schema.optional(Schema.Record(Schema.String, Schema.String)),
+  safetyIdentifier: Schema.optional(Schema.String),
+  streamOptions: Schema.optional(StreamOptions),
+  topLogprobs: Schema.optional(Schema.Int.check(Schema.isBetween({ minimum: 0, maximum: 20 }))),
   reasoningEffort: Schema.optional(ReasoningEffort),
   reasoningSummary: Schema.optional(Schema.Literals(["auto", "concise", "detailed"])),
   include: Schema.optional(Schema.Array(ResponseIncludableSchema)),

packages/ai/src/protocols/utils/responses-hosted-tools.ts

@@ -0,0 +1,70 @@
+import { Effect } from "effect"
+import { LLMEvent, type AIError, type ToolResultPart } from "../../schema/index.js"
+import { OpenResponses } from "../open-responses.js"
+import { Lifecycle } from "./lifecycle.js"
+
+export type Item = OpenResponses.StreamItem & {
+  readonly id: string
+  readonly status?: string
+  readonly action?: unknown
+  readonly queries?: unknown
+  readonly results?: unknown
+  readonly code?: string
+  readonly container_id?: string
+  readonly outputs?: unknown
+  readonly server_label?: string
+  readonly output?: unknown
+  readonly result?: string
+  readonly output_format?: "png" | "jpeg" | "webp"
+  readonly error?: unknown
+}
+
+export interface Definition {
+  readonly name: string
+  readonly input: (item: Item) => unknown
+  readonly result?: (item: Item) => Effect.Effect<ToolResultPart["result"], AIError>
+}
+
+export type Definitions = Readonly<Record<string, Definition>>
+
+export const isItem = <Tools extends Definitions>(item: OpenResponses.StreamItem, tools: Tools): item is Item =>
+  item.type in tools && typeof item.id === "string" && item.id.length > 0
+
+export const onDone: (
+  state: OpenResponses.ParserState,
+  item: Item,
+  tools: Definitions,
+) => Effect.Effect<OpenResponses.StepResult, AIError> = Effect.fn("ResponsesHostedTools.onDone")(function* (
+  state,
+  item,
+  tools,
+) {
+  const tool = tools[item.type]
+  if (!tool) return [state, []] satisfies OpenResponses.StepResult
+  const providerMetadata = OpenResponses.providerMetadata(state, { itemId: item.id })
+  const events: LLMEvent[] = []
+  const lifecycle = Lifecycle.stepStart(state.lifecycle, events)
+  events.push(
+    LLMEvent.toolCall({
+      id: item.id,
+      name: tool.name,
+      input: tool.input(item),
+      providerExecuted: true,
+      providerMetadata,
+    }),
+    LLMEvent.toolResult({
+      id: item.id,
+      name: tool.name,
+      result: tool.result
+        ? yield* tool.result(item)
+        : item.error !== undefined && item.error !== null
+          ? { type: "error", value: item.error }
+          : { type: "json", value: item },
+      providerExecuted: true,
+      providerMetadata,
+    }),
+  )
+  return [{ ...state, lifecycle }, events] satisfies OpenResponses.StepResult
+})
+
+export * as ResponsesHostedTools from "./responses-hosted-tools.js"

packages/ai/src/protocols/xai-responses.ts

@@ -0,0 +1,55 @@
+import { Effect } from "effect"
+import { Protocol } from "../route/protocol.js"
+import { OpenResponses } from "./open-responses.js"
+import { ProviderShared } from "./shared.js"
+import { ResponsesHostedTools } from "./utils/responses-hosted-tools.js"
+
+const ADAPTER = "xai-responses"
+const NAME = "xAI Responses"
+
+const extension = {
+  id: ADAPTER,
+  name: NAME,
+} satisfies OpenResponses.Extension
+
+const HOSTED_TOOLS = {
+  web_search_call: { name: "web_search", input: (item) => item.action ?? {} },
+  x_search_call: { name: "x_search", input: (item) => item.action ?? {} },
+  file_search_call: { name: "file_search", input: (item) => ({ queries: item.queries ?? [] }) },
+  code_interpreter_call: {
+    name: "code_interpreter",
+    input: (item) => ({ code: item.code, container_id: item.container_id }),
+  },
+  image_generation_call: { name: "image_generation", input: () => ({}) },
+  mcp_call: {
+    name: "mcp",
+    input: (item) => ({ server_label: item.server_label, name: item.name, arguments: item.arguments }),
+  },
+} as const satisfies ResponsesHostedTools.Definitions
+
+const step = (state: OpenResponses.ParserState, event: OpenResponses.Event) => {
+  if (event.type === "response.reasoning_text.delta" || event.type === "response.reasoning_summary.delta")
+    return event.item_id
+      ? Effect.succeed(OpenResponses.onReasoningDelta(state, event, event.item_id))
+      : ProviderShared.eventError(ADAPTER, `${event.type} is missing item_id`)
+  if (event.type === "response.reasoning_text.done" || event.type === "response.reasoning_summary.done")
+    return event.item_id
+      ? Effect.succeed(OpenResponses.onReasoningDone(state, event))
+      : ProviderShared.eventError(ADAPTER, `${event.type} is missing item_id`)
+  if (event.type === "response.output_item.done" && event.item && ResponsesHostedTools.isItem(event.item, HOSTED_TOOLS))
+    return ResponsesHostedTools.onDone(state, event.item, HOSTED_TOOLS)
+  return OpenResponses.step(state, event)
+}
+
+export const protocol = Protocol.make({
+  id: ADAPTER,
+  body: OpenResponses.protocol.body,
+  stream: {
+    event: OpenResponses.protocol.stream.event,
+    initial: (request) => OpenResponses.initial(request, extension),
+    step,
+    terminal: OpenResponses.terminal,
+  },
+})
+
+export * as XAIResponses from "./xai-responses.js"

packages/ai/src/provider-error.ts

@@ -74,11 +74,15 @@ const INVALID_REQUEST_CODES = new Set(["invalid_prompt", "invalid_request_error"
 const RATE_LIMIT_TEXT = /rate increased too quickly|rate[-_\s]?limit|too[_\s]?many[_\s]?requests/i
 const QUOTA_TEXT = /insufficient[-_\s]?quota|quota[-_\s]?exceeded/i
 const CONTENT_POLICY_TEXT = /content[-_\s]?policy|content_filter|safety/i
+const NETWORK_ERROR_TEXT = /network[-_\s]error/i
 
 export interface ProviderFailure {
   readonly message: string
   readonly status?: number | undefined
   readonly code?: string | undefined
+  // Raw wire payload, scanned for failure signals (codes, overflow phrases)
+  // that the summary message does not carry. Not shown to users.
+  readonly rawBody?: string | undefined
   readonly retryAfterMs?: number | undefined
   readonly rateLimit?: HttpRateLimitDetails | undefined
   readonly http?: HttpContext | undefined
@@ -88,11 +92,13 @@ export interface ProviderFailure {
 // Keep HTTP failures and provider-reported stream failures on one typed path so
 // session retry policy never needs provider-specific string matching.
 export function classifyProviderFailure(input: ProviderFailure): AIError["reason"] {
-  const body = input.http?.body ?? ""
+  const body = input.http?.body ?? input.rawBody ?? ""
   const codes = [input.code, ...providerCodes(body), ...providerCodes(input.message)]
     .filter((code): code is string => code !== undefined)
     .map((code) => code.toLowerCase())
-  const text = body || input.message
+  // Scan the raw payload too so signals missing from the summary message
+  // (e.g. overflow phrases nested in a JSON error body) still classify.
+  const text = [input.message, body].filter((value) => value.length > 0).join("\n")
   const common = { message: input.message, providerMetadata: input.providerMetadata, http: input.http }
   const clientScoped = input.status === undefined || (input.status >= 400 && input.status < 500)
 
@@ -127,6 +133,7 @@ export function classifyProviderFailure(input: ProviderFailure): AIError["reason
       retryAfterMs: input.retryAfterMs,
       rateLimit: input.rateLimit,
     })
+  if (NETWORK_ERROR_TEXT.test(text)) return new ProviderInternalReason({ ...common, status: input.status })
   if (codes.some((code) => SERVER_CODES.has(code) || code.includes("exhausted") || code.includes("unavailable")))
     return new ProviderInternalReason({
       ...common,

packages/ai/src/provider-package.ts

@@ -4,11 +4,6 @@ export interface Settings extends Readonly<Record<string, unknown>> {
   readonly baseURL?: string
   readonly headers?: Readonly<Record<string, string>>
   readonly body?: Readonly<Record<string, unknown>>
-  readonly limits?: {
-    readonly context: number
-    readonly input?: number
-    readonly output: number
-  }
 }
 
 export interface Definition<

packages/ai/src/providers/amazon-bedrock-mantle.ts

@@ -90,7 +90,6 @@ const config = (settings: Settings): Config => {
     credentials: settings.credentials,
     headers: settings.headers === undefined ? undefined : { ...settings.headers },
     http: settings.body === undefined ? undefined : { body: { ...settings.body } },
-    limits: settings.limits,
     providerOptions: settings.providerOptions,
     region: settings.region,
   }

packages/ai/src/providers/amazon-bedrock.ts

@@ -62,7 +62,6 @@ export const model: ProviderPackage.Definition<Settings>["model"] = (modelID, se
     generation: settings.topP === undefined ? undefined : { topP: settings.topP },
     headers: settings.headers === undefined ? undefined : { ...settings.headers },
     http: settings.body === undefined ? undefined : { body: { ...settings.body } },
-    limits: settings.limits,
     region: settings.region,
   }).model(modelID)
 }

packages/ai/src/providers/anthropic-compatible.ts

@@ -68,7 +68,6 @@ export const model: ProviderPackage.Definition<Settings, AnthropicMessages.Provi
     baseURL: settings.baseURL,
     headers: settings.headers === undefined ? undefined : { ...settings.headers },
     http: settings.body === undefined ? undefined : { body: { ...settings.body } },
-    limits: settings.limits,
     provider: settings.provider,
     providerOptions: settings.providerOptions,
   }).model(modelID)

packages/ai/src/providers/anthropic.ts

@@ -63,7 +63,6 @@ export const model: ProviderPackage.Definition<Settings, AnthropicMessages.Provi
     baseURL: settings.baseURL,
     headers: settings.headers === undefined ? undefined : { ...settings.headers },
     http: settings.body === undefined ? undefined : { body: { ...settings.body } },
-    limits: settings.limits,
     providerOptions: settings.providerOptions,
   }).model(modelID)
 }

packages/ai/src/providers/azure.ts

@@ -1,3 +1,4 @@
+import { Headers } from "effect/unstable/http"
 import { Auth } from "../route/auth.js"
 import { type AtLeastOne, type ProviderAuthOption } from "../route/auth-options.js"
 import type { Route as RouteDef, RouteDefaultsInput } from "../route/client.js"
@@ -10,6 +11,7 @@ import { withOpenAIOptions, type OpenAIProviderOptionsInput } from "./openai-opt
 
 export const id = ProviderID.make("azure")
 const routeAuth = Auth.remove("authorization")
+const RESPONSES_WEBSOCKET_ROTATE_AFTER_MS = 55 * 60 * 1000
 
 // Azure needs the customer's resource URL; supply either `resourceName`
 // (helper builds the URL) or `baseURL` directly.
@@ -40,6 +42,30 @@ const responsesRoute = OpenAIResponses.route.with({
   id: "azure-openai-responses",
   provider: id,
   auth: routeAuth,
+  transport: OpenAIResponses.channelTransport({
+    id: "azure-openai-responses",
+    name: "Azure OpenAI Responses",
+    rotateAfterMs: RESPONSES_WEBSOCKET_ROTATE_AFTER_MS,
+    enabled: (value) => {
+      const url = new URL(value)
+      return (
+        url.protocol === "https:" &&
+        url.hostname.endsWith(".openai.azure.com") &&
+        url.pathname.endsWith("/openai/v1/responses") &&
+        url.searchParams.get("api-version") === "v1"
+      )
+    },
+    url: (value) => {
+      const url = new URL(value)
+      url.searchParams.delete("api-version")
+      return url.toString()
+    },
+    headers: (headers) => {
+      const apiKey = headers["api-key"]
+      if (!apiKey) return headers
+      return Headers.remove(Headers.set(headers, "authorization", `Bearer ${apiKey}`), "api-key")
+    },
+  }),
 })
 
 const chatRoute = OpenAIChat.route.with({
@@ -126,7 +152,6 @@ const config = (settings: Settings): Config => {
     apiVersion: settings.apiVersion,
     headers: settings.headers === undefined ? undefined : { ...settings.headers },
     http: settings.body === undefined ? undefined : { body: { ...settings.body } },
-    limits: settings.limits,
     providerOptions: settings.providerOptions,
     queryParams: settings.queryParams === undefined ? undefined : { ...settings.queryParams },
     useDeploymentBasedUrls: settings.useDeploymentBasedUrls,

packages/ai/src/providers/google-vertex-chat.ts

@@ -75,7 +75,6 @@ export const model: ProviderPackage.Definition<Settings, OpenAIProviderOptionsIn
     baseURL: settings.baseURL,
     headers: settings.headers === undefined ? undefined : { ...settings.headers },
     http: settings.body === undefined ? undefined : { body: { ...settings.body } },
-    limits: settings.limits,
     location: settings.location,
     project: settings.project,
     providerOptions: settings.providerOptions,

packages/ai/src/providers/google-vertex-messages.ts

@@ -4,7 +4,6 @@ import { AnthropicMessages } from "../protocols/anthropic-messages.js"
 import { Auth } from "../route/auth.js"
 import { Route, type RouteDefaultsInput } from "../route/client.js"
 import { Endpoint } from "../route/endpoint.js"
-import { Framing } from "../route/framing.js"
 import { Protocol } from "../route/protocol.js"
 import { ProviderID, type ModelID } from "../schema/index.js"
 import { GoogleVertexShared } from "./google-vertex-shared.js"
@@ -14,6 +13,7 @@ export type AnthropicProviderOptionsInput = AnthropicMessages.ProviderOptionsInp
 export type AnthropicThinkingInput = AnthropicMessages.ThinkingInput
 
 const VERSION = "vertex-2023-10-16" as const
+const HEADER_VERSION = "2023-06-01" as const
 
 export const id = ProviderID.make("google-vertex")
 
@@ -57,7 +57,8 @@ const route = Route.make({
   }),
   endpoint: Endpoint.path(({ request }) => `/${request.model.id}:streamRawPredict`),
   auth: Auth.none,
-  framing: Framing.sse,
+  framing: AnthropicMessages.framing,
+  headers: () => ({ "anthropic-version": HEADER_VERSION }),
 })
 
 export const routes = [route]
@@ -110,7 +111,6 @@ export const model: ProviderPackage.Definition<Settings, AnthropicMessages.Provi
     baseURL: settings.baseURL,
     headers: settings.headers === undefined ? undefined : { ...settings.headers },
     http: settings.body === undefined ? undefined : { body: { ...settings.body } },
-    limits: settings.limits,
     location: settings.location,
     project: settings.project,
     providerOptions: settings.providerOptions,

packages/ai/src/providers/google-vertex-responses.ts

@@ -80,7 +80,6 @@ export const model: ProviderPackage.Definition<Settings, OpenResponsesProviderOp
     baseURL: settings.baseURL,
     headers: settings.headers === undefined ? undefined : { ...settings.headers },
     http: settings.body === undefined ? undefined : { body: { ...settings.body } },
-    limits: settings.limits,
     location: settings.location,
     project: settings.project,
     providerOptions: settings.providerOptions,

packages/ai/src/providers/google-vertex.ts

@@ -119,7 +119,6 @@ export const model: ProviderPackage.Definition<Settings, GeminiProviderOptionsIn
     baseURL: settings.baseURL,
     headers: settings.headers === undefined ? undefined : { ...settings.headers },
     http: settings.body === undefined ? undefined : { body: { ...settings.body } },
-    limits: settings.limits,
     location: settings.location,
     project: settings.project,
     providerOptions: settings.providerOptions,

packages/ai/src/providers/google.ts

@@ -63,7 +63,6 @@ export const model: ProviderPackage.Definition<Settings, Gemini.ProviderOptionsI
     baseURL: settings.baseURL,
     headers: settings.headers === undefined ? undefined : { ...settings.headers },
     http: settings.body === undefined ? undefined : { body: { ...settings.body } },
-    limits: settings.limits,
     providerOptions: settings.providerOptions,
   }).model(modelID)
 

packages/ai/src/providers/openai-compatible-responses.ts

@@ -55,7 +55,6 @@ export const model: ProviderPackage.Definition<Settings, OpenResponsesProviderOp
     baseURL: settings.baseURL,
     headers: settings.headers === undefined ? undefined : { ...settings.headers },
     http: settings.body === undefined ? undefined : { body: { ...settings.body } },
-    limits: settings.limits,
     provider: settings.provider,
     providerOptions: settings.providerOptions,
   }).model(modelID)

packages/ai/src/providers/openai-compatible.ts

@@ -74,7 +74,6 @@ export const model: ProviderPackage.Definition<Settings, OpenAIProviderOptionsIn
     baseURL: settings.baseURL,
     headers: settings.headers === undefined ? undefined : { ...settings.headers },
     http: settings.body === undefined ? undefined : { body: { ...settings.body } },
-    limits: settings.limits,
     provider: settings.provider,
     providerOptions: settings.providerOptions,
   }).model(modelID)

packages/ai/src/providers/openai.ts

@@ -124,7 +124,6 @@ const config = (settings: Settings): Config => {
     baseURL: settings.baseURL,
     headers: Object.keys(headers).length === 0 ? undefined : headers,
     http: settings.body === undefined ? undefined : { body: { ...settings.body } },
-    limits: settings.limits,
     providerOptions: settings.providerOptions,
     queryParams: settings.queryParams === undefined ? undefined : { ...settings.queryParams },
   }

packages/ai/src/providers/openrouter.ts

@@ -200,6 +200,5 @@ export const model: ProviderPackage.Definition<Settings, OpenRouterProviderOptio
     baseURL: settings.baseURL,
     headers: settings.headers,
     http: settings.body === undefined ? undefined : { body: { ...settings.body } },
-    limits: settings.limits,
     providerOptions: settings.providerOptions,
   }).model(modelID)

packages/ai/src/providers/xai.ts

@@ -5,7 +5,8 @@ import { HttpOptions, ProviderID, type ModelID } from "../schema/index.js"
 import * as OpenAICompatibleProfiles from "./openai-compatible-profile.js"
 import * as OpenAICompatibleChat from "../protocols/openai-compatible-chat.js"
 import * as OpenAIChat from "../protocols/openai-chat.js"
-import * as OpenAIResponses from "../protocols/openai-responses.js"
+import { OpenResponsesChannel } from "../protocols/open-responses-channel.js"
+import { XAIResponses } from "../protocols/xai-responses.js"
 import { XAIImages } from "../protocols/xai-images.js"
 import type { OpenAIOptionsInput } from "./openai-options.js"
 import type { ProviderPackage } from "../provider-package.js"
@@ -28,13 +29,19 @@ export interface Settings extends ProviderPackage.Settings {
 
 export type { XAIImageOptions } from "../protocols/xai-images.js"
 
+const RESPONSES_WEBSOCKET_ROTATE_AFTER_MS = 24 * 60 * 1000
+
 const responsesRoute = Route.make({
   id: "openai-responses",
   provider: id,
   providerMetadataKey: "xai",
-  protocol: OpenAIResponses.protocol,
+  protocol: XAIResponses.protocol,
   endpoint: Endpoint.path("/responses", { baseURL: OpenAICompatibleProfiles.profiles.xai.baseURL }),
-  transport: OpenAIResponses.httpTransport,
+  transport: OpenResponsesChannel.transport({
+    id: "openai-responses",
+    name: "xAI Responses",
+    rotateAfterMs: RESPONSES_WEBSOCKET_ROTATE_AFTER_MS,
+  }),
   defaults: { providerOptions: { store: false } },
 })
 
@@ -101,7 +108,6 @@ export const model: ProviderPackage.Definition<Settings, XAIProviderOptionsInput
     baseURL: settings.baseURL,
     headers: settings.headers,
     http: settings.body === undefined ? undefined : { body: { ...settings.body } },
-    limits: settings.limits,
     providerOptions: settings.providerOptions,
   }).model(modelID)
 export const responses = provider.responses

packages/ai/src/route/client.ts

@@ -16,7 +16,6 @@ import {
   LLMRequest,
   LLMResponse,
   LanguageModel,
-  LanguageModelLimits,
   LLMEvent,
   InvalidProviderOutputReason,
   ProviderID,
@@ -74,7 +73,6 @@ export type RouteRoutedLanguageModelInput = Omit<LanguageModel.Input, "route">
 
 export interface RouteDefaults {
   readonly headers?: Record<string, string>
-  readonly limits?: LanguageModelLimits
   readonly generation?: GenerationOptions
   readonly providerOptions?: ProviderOptions
   readonly http?: HttpOptions
@@ -82,7 +80,6 @@ export interface RouteDefaults {
 
 export interface RouteDefaultsInput {
   readonly headers?: Record<string, string>
-  readonly limits?: LanguageModelLimits.Input
   readonly generation?: GenerationOptions.Input
   readonly providerOptions?: ProviderOptions
   readonly http?: HttpOptions.Input
@@ -119,7 +116,6 @@ const mergeRouteDefaults = (base: RouteDefaults | undefined, patch: RouteDefault
     ...base,
     ...patch,
     headers,
-    limits: patch.limits === undefined ? base?.limits : LanguageModelLimits.make(patch.limits),
     generation: mergeGenerationOptions(generationOptions(base?.generation), generationOptions(patch.generation)),
     providerOptions: mergeProviderOptions(base?.providerOptions, patch.providerOptions),
     http: mergeHttpOptions(

packages/ai/src/route/framing.ts

@@ -24,4 +24,10 @@ export interface Definition<Frame> {
 /** Server-Sent Events framing. Used by every JSON-streaming HTTP provider. */
 export const sse: Definition<string> = { id: "sse", frame: ProviderShared.sseFraming }
 
+/** SSE framing restricted to protocol-recognized event names. */
+export const sseEvents = (events: ReadonlySet<string>): Definition<string> => ({
+  id: "sse",
+  frame: (bytes) => ProviderShared.sseFraming(bytes, events),
+})
+
 export * as Framing from "./framing.js"

packages/ai/src/schema/errors.ts

@@ -153,6 +153,9 @@ export class AIError extends Schema.TaggedError<AIError>()("AI.Error", {
   module: Schema.String,
   method: Schema.String,
   reason: AIErrorReason,
+  // Raw provider payload as a string, so classified failures never lose the
+  // original error detail even when the pretty message is a summary.
+  body: Schema.optional(Schema.String),
 }) {
   override readonly cause = this.reason
 

packages/ai/src/schema/options.ts

@@ -114,22 +114,7 @@ export const mergeGenerationOptions = (...items: ReadonlyArray<GenerationOptions
   return Object.values(result).some((value) => value !== undefined) ? result : undefined
 }
 
-export class LanguageModelLimits extends Schema.Class<LanguageModelLimits>("LLM.LanguageModelLimits")({
-  context: Schema.optional(Schema.Number),
-  input: Schema.optional(Schema.Number),
-  output: Schema.optional(Schema.Number),
-}) {}
-
-export namespace LanguageModelLimits {
-  export type Input = LanguageModelLimits | ConstructorParameters<typeof LanguageModelLimits>[0]
-
-  /** Normalize model limit input into the canonical `LanguageModelLimits` class. */
-  export const make = (input: Input | undefined) =>
-    input instanceof LanguageModelLimits ? input : new LanguageModelLimits(input ?? {})
-}
-
 export class LanguageModelDefaults extends Schema.Class<LanguageModelDefaults>("LLM.LanguageModelDefaults")({
-  limits: Schema.optional(LanguageModelLimits),
   generation: Schema.optional(GenerationOptions),
   providerOptions: Schema.optional(ProviderOptions),
   http: Schema.optional(HttpOptions),
@@ -139,7 +124,6 @@ export namespace LanguageModelDefaults {
   export type Input =
     | LanguageModelDefaults
     | {
-        readonly limits?: LanguageModelLimits.Input
         readonly generation?: GenerationOptions.Input
         readonly providerOptions?: ProviderOptions
         readonly http?: HttpOptions.Input
@@ -149,7 +133,6 @@ export namespace LanguageModelDefaults {
   export const make = (input: Input) => {
     if (input instanceof LanguageModelDefaults) return input
     return new LanguageModelDefaults({
-      limits: input.limits === undefined ? undefined : LanguageModelLimits.make(input.limits),
       generation: input.generation === undefined ? undefined : GenerationOptions.make(input.generation),
       providerOptions: input.providerOptions,
       http: input.http === undefined ? undefined : HttpOptions.make(input.http),

packages/ai/test/compile.test.ts

@@ -270,21 +270,20 @@ describe("request option precedence", () => {
     ),
   )
 
-  it.effect("uses model output limits after route limits and before call maxTokens", () =>
+  it.effect("uses the Anthropic default before call maxTokens", () =>
     Effect.gen(function* () {
       const route = AnthropicMessages.route.with({
         endpoint: { baseURL: "https://api.anthropic.test/v1/" },
         auth: Auth.header("x-api-key", "test"),
-        limits: { output: 128 },
       })
-      const model = route.model({ id: "claude-sonnet-4-5", defaults: { limits: { output: 64 } } })
+      const model = route.model({ id: "claude-sonnet-4-5" })
       const withoutMaxTokens = yield* compileRequest(LLM.request({ model, prompt: "Say hello.", cache: "none" }))
       const withMaxTokens = yield* compileRequest(
-        LLM.request({ model, prompt: "Say hello.", cache: "none", generation: { maxTokens: 32 } }),
+        LLM.request({ model, prompt: "Say hello.", cache: "none", generation: { maxTokens: 8_000 } }),
       )
 
-      expect(withoutMaxTokens.body.max_tokens).toBe(64)
-      expect(withMaxTokens.body.max_tokens).toBe(32)
+      expect(withoutMaxTokens.body.max_tokens).toBe(32_000)
+      expect(withMaxTokens.body.max_tokens).toBe(8_000)
     }),
   )
 })

packages/ai/test/fixtures/recordings/openai-responses-websocket/continues-a-tool-call-over-one-socket.json

@@ -0,0 +1,100 @@
+{
+  "version": 1,
+  "metadata": {
+    "provider": "openai",
+    "protocol": "openai-responses",
+    "transport": "websocket",
+    "model": "gpt-5.5",
+    "tags": [
+      "prefix:openai-responses-websocket",
+      "provider:openai",
+      "protocol:openai-responses",
+      "transport:websocket",
+      "tool",
+      "continuation"
+    ],
+    "name": "openai-responses-websocket/continues-a-tool-call-over-one-socket",
+    "recordedAt": "2026-08-20T00:00:00.000Z"
+  },
+  "interactions": [
+    {
+      "transport": "websocket",
+      "connection": {
+        "sequence": 0,
+        "url": "wss://api.openai.com/v1/responses",
+        "protocols": [],
+        "close": {
+          "code": 1000,
+          "reason": ""
+        }
+      },
+      "events": [
+        {
+          "direction": "client",
+          "kind": "text",
+          "body": "{\"type\":\"response.create\",\"model\":\"gpt-5.5\",\"input\":[{\"role\":\"system\",\"content\":\"Call get_weather once, then reply exactly: Paris is sunny.\"},{\"role\":\"user\",\"content\":[{\"type\":\"input_text\",\"text\":\"What is the weather in Paris?\"}]}],\"tools\":[{\"type\":\"function\",\"name\":\"get_weather\",\"description\":\"Get current weather for a city.\",\"parameters\":{\"type\":\"object\",\"properties\":{\"city\":{\"type\":\"string\"}},\"required\":[\"city\"],\"additionalProperties\":false},\"strict\":false}],\"store\":false,\"max_output_tokens\":50,\"include\":[\"reasoning.encrypted_content\"],\"reasoning\":{\"effort\":\"medium\",\"summary\":\"auto\"},\"text\":{\"verbosity\":\"low\"}}"
+        },
+        {
+          "direction": "server",
+          "kind": "text",
+          "body": "{\"type\":\"response.created\",\"response\":{\"id\":\"resp_ws_tool_1\"}}"
+        },
+        {
+          "direction": "server",
+          "kind": "text",
+          "body": "{\"type\":\"response.output_item.added\",\"item\":{\"type\":\"function_call\",\"id\":\"fc_ws_weather\",\"call_id\":\"call_ws_weather\",\"name\":\"get_weather\",\"arguments\":\"\"}}"
+        },
+        {
+          "direction": "server",
+          "kind": "text",
+          "body": "{\"type\":\"response.function_call_arguments.delta\",\"item_id\":\"fc_ws_weather\",\"delta\":\"{\\\"city\\\":\\\"Paris\\\"}\"}"
+        },
+        {
+          "direction": "server",
+          "kind": "text",
+          "body": "{\"type\":\"response.output_item.done\",\"item\":{\"type\":\"function_call\",\"id\":\"fc_ws_weather\",\"call_id\":\"call_ws_weather\",\"name\":\"get_weather\",\"arguments\":\"{\\\"city\\\":\\\"Paris\\\"}\"}}"
+        },
+        {
+          "direction": "server",
+          "kind": "text",
+          "body": "{\"type\":\"response.completed\",\"response\":{\"id\":\"resp_ws_tool_1\"}}"
+        },
+        {
+          "direction": "client",
+          "kind": "text",
+          "body": "{\"type\":\"response.create\",\"model\":\"gpt-5.5\",\"input\":[{\"type\":\"function_call_output\",\"call_id\":\"call_ws_weather\",\"output\":\"{\\\"temperature\\\":22,\\\"condition\\\":\\\"sunny\\\"}\"}],\"tools\":[{\"type\":\"function\",\"name\":\"get_weather\",\"description\":\"Get current weather for a city.\",\"parameters\":{\"type\":\"object\",\"properties\":{\"city\":{\"type\":\"string\"}},\"required\":[\"city\"],\"additionalProperties\":false},\"strict\":false}],\"store\":false,\"max_output_tokens\":50,\"previous_response_id\":\"resp_ws_tool_1\",\"include\":[\"reasoning.encrypted_content\"],\"reasoning\":{\"effort\":\"medium\",\"summary\":\"auto\"},\"text\":{\"verbosity\":\"low\"}}"
+        },
+        {
+          "direction": "server",
+          "kind": "text",
+          "body": "{\"type\":\"response.created\",\"response\":{\"id\":\"resp_ws_tool_2\"}}"
+        },
+        {
+          "direction": "server",
+          "kind": "text",
+          "body": "{\"type\":\"response.output_item.added\",\"item\":{\"type\":\"message\",\"id\":\"msg_ws_tool_2\",\"role\":\"assistant\",\"content\":[]}}"
+        },
+        {
+          "direction": "server",
+          "kind": "text",
+          "body": "{\"type\":\"response.output_text.delta\",\"item_id\":\"msg_ws_tool_2\",\"delta\":\"Paris is sunny.\"}"
+        },
+        {
+          "direction": "server",
+          "kind": "text",
+          "body": "{\"type\":\"response.output_text.done\",\"item_id\":\"msg_ws_tool_2\",\"text\":\"Paris is sunny.\"}"
+        },
+        {
+          "direction": "server",
+          "kind": "text",
+          "body": "{\"type\":\"response.output_item.done\",\"item\":{\"type\":\"message\",\"id\":\"msg_ws_tool_2\",\"role\":\"assistant\",\"content\":[{\"type\":\"output_text\",\"text\":\"Paris is sunny.\"}]}}"
+        },
+        {
+          "direction": "server",
+          "kind": "text",
+          "body": "{\"type\":\"response.completed\",\"response\":{\"id\":\"resp_ws_tool_2\"}}"
+        }
+      ]
+    }
+  ]
+}

packages/ai/test/fixtures/recordings/openai-responses-websocket/reconstructs-full-context-after-reconnect.json

@@ -0,0 +1,119 @@
+{
+  "version": 1,
+  "metadata": {
+    "provider": "openai",
+    "protocol": "openai-responses",
+    "transport": "websocket",
+    "model": "gpt-5.5",
+    "tags": [
+      "prefix:openai-responses-websocket",
+      "provider:openai",
+      "protocol:openai-responses",
+      "transport:websocket",
+      "reconnect",
+      "full-context"
+    ],
+    "name": "openai-responses-websocket/reconstructs-full-context-after-reconnect",
+    "recordedAt": "2026-08-20T00:00:00.000Z"
+  },
+  "interactions": [
+    {
+      "transport": "websocket",
+      "connection": {
+        "sequence": 0,
+        "url": "wss://api.openai.com/v1/responses",
+        "protocols": [],
+        "close": {
+          "code": 1000,
+          "reason": ""
+        }
+      },
+      "events": [
+        {
+          "direction": "client",
+          "kind": "text",
+          "body": "{\"type\":\"response.create\",\"model\":\"gpt-5.5\",\"input\":[{\"role\":\"system\",\"content\":\"Follow the user's exact reply instruction.\"},{\"role\":\"user\",\"content\":[{\"type\":\"input_text\",\"text\":\"Reply exactly: Alpha.\"}]}],\"store\":false,\"max_output_tokens\":30,\"include\":[\"reasoning.encrypted_content\"],\"reasoning\":{\"effort\":\"medium\",\"summary\":\"auto\"},\"text\":{\"verbosity\":\"low\"}}"
+        },
+        {
+          "direction": "server",
+          "kind": "text",
+          "body": "{\"type\":\"response.created\",\"response\":{\"id\":\"resp_ws_reconnect_1\"}}"
+        },
+        {
+          "direction": "server",
+          "kind": "text",
+          "body": "{\"type\":\"response.output_item.added\",\"item\":{\"type\":\"message\",\"id\":\"msg_ws_reconnect_1\",\"role\":\"assistant\",\"content\":[]}}"
+        },
+        {
+          "direction": "server",
+          "kind": "text",
+          "body": "{\"type\":\"response.output_text.delta\",\"item_id\":\"msg_ws_reconnect_1\",\"delta\":\"Alpha.\"}"
+        },
+        {
+          "direction": "server",
+          "kind": "text",
+          "body": "{\"type\":\"response.output_text.done\",\"item_id\":\"msg_ws_reconnect_1\",\"text\":\"Alpha.\"}"
+        },
+        {
+          "direction": "server",
+          "kind": "text",
+          "body": "{\"type\":\"response.output_item.done\",\"item\":{\"type\":\"message\",\"id\":\"msg_ws_reconnect_1\",\"role\":\"assistant\",\"content\":[{\"type\":\"output_text\",\"text\":\"Alpha.\"}]}}"
+        },
+        {
+          "direction": "server",
+          "kind": "text",
+          "body": "{\"type\":\"response.completed\",\"response\":{\"id\":\"resp_ws_reconnect_1\"}}"
+        }
+      ]
+    },
+    {
+      "transport": "websocket",
+      "connection": {
+        "sequence": 1,
+        "url": "wss://api.openai.com/v1/responses",
+        "protocols": [],
+        "close": {
+          "code": 1000,
+          "reason": ""
+        }
+      },
+      "events": [
+        {
+          "direction": "client",
+          "kind": "text",
+          "body": "{\"type\":\"response.create\",\"model\":\"gpt-5.5\",\"input\":[{\"role\":\"system\",\"content\":\"Follow the user's exact reply instruction.\"},{\"role\":\"user\",\"content\":[{\"type\":\"input_text\",\"text\":\"Reply exactly: Alpha.\"}]},{\"type\":\"message\",\"id\":\"msg_ws_reconnect_1\",\"role\":\"assistant\",\"content\":[{\"type\":\"output_text\",\"text\":\"Alpha.\"}]},{\"role\":\"user\",\"content\":[{\"type\":\"input_text\",\"text\":\"Reply exactly: Beta.\"}]}],\"store\":false,\"max_output_tokens\":30,\"include\":[\"reasoning.encrypted_content\"],\"reasoning\":{\"effort\":\"medium\",\"summary\":\"auto\"},\"text\":{\"verbosity\":\"low\"}}"
+        },
+        {
+          "direction": "server",
+          "kind": "text",
+          "body": "{\"type\":\"response.created\",\"response\":{\"id\":\"resp_ws_reconnect_2\"}}"
+        },
+        {
+          "direction": "server",
+          "kind": "text",
+          "body": "{\"type\":\"response.output_item.added\",\"item\":{\"type\":\"message\",\"id\":\"msg_ws_reconnect_2\",\"role\":\"assistant\",\"content\":[]}}"
+        },
+        {
+          "direction": "server",
+          "kind": "text",
+          "body": "{\"type\":\"response.output_text.delta\",\"item_id\":\"msg_ws_reconnect_2\",\"delta\":\"Beta.\"}"
+        },
+        {
+          "direction": "server",
+          "kind": "text",
+          "body": "{\"type\":\"response.output_text.done\",\"item_id\":\"msg_ws_reconnect_2\",\"text\":\"Beta.\"}"
+        },
+        {
+          "direction": "server",
+          "kind": "text",
+          "body": "{\"type\":\"response.output_item.done\",\"item\":{\"type\":\"message\",\"id\":\"msg_ws_reconnect_2\",\"role\":\"assistant\",\"content\":[{\"type\":\"output_text\",\"text\":\"Beta.\"}]}}"
+        },
+        {
+          "direction": "server",
+          "kind": "text",
+          "body": "{\"type\":\"response.completed\",\"response\":{\"id\":\"resp_ws_reconnect_2\"}}"
+        }
+      ]
+    }
+  ]
+}

packages/ai/test/fixtures/recordings/openai-responses-websocket/recovers-from-explicit-continuation-rejection.json

@@ -0,0 +1,129 @@
+{
+  "version": 1,
+  "metadata": {
+    "provider": "openai",
+    "protocol": "openai-responses",
+    "transport": "websocket",
+    "model": "gpt-5.5",
+    "tags": [
+      "prefix:openai-responses-websocket",
+      "provider:openai",
+      "protocol:openai-responses",
+      "transport:websocket",
+      "continuation",
+      "recovery"
+    ],
+    "name": "openai-responses-websocket/recovers-from-explicit-continuation-rejection",
+    "recordedAt": "2026-08-20T00:00:00.000Z"
+  },
+  "interactions": [
+    {
+      "transport": "websocket",
+      "connection": {
+        "sequence": 0,
+        "url": "wss://api.openai.com/v1/responses",
+        "protocols": [],
+        "close": {
+          "code": 1000,
+          "reason": ""
+        }
+      },
+      "events": [
+        {
+          "direction": "client",
+          "kind": "text",
+          "body": "{\"type\":\"response.create\",\"model\":\"gpt-5.5\",\"input\":[{\"role\":\"system\",\"content\":\"Follow the user's exact reply instruction.\"},{\"role\":\"user\",\"content\":[{\"type\":\"input_text\",\"text\":\"Reply exactly: Ready.\"}]}],\"store\":false,\"max_output_tokens\":30,\"include\":[\"reasoning.encrypted_content\"],\"reasoning\":{\"effort\":\"medium\",\"summary\":\"auto\"},\"text\":{\"verbosity\":\"low\"}}"
+        },
+        {
+          "direction": "server",
+          "kind": "text",
+          "body": "{\"type\":\"response.created\",\"response\":{\"id\":\"resp_ws_rejection_1\"}}"
+        },
+        {
+          "direction": "server",
+          "kind": "text",
+          "body": "{\"type\":\"response.output_item.added\",\"item\":{\"type\":\"message\",\"id\":\"msg_ws_rejection_1\",\"role\":\"assistant\",\"content\":[]}}"
+        },
+        {
+          "direction": "server",
+          "kind": "text",
+          "body": "{\"type\":\"response.output_text.delta\",\"item_id\":\"msg_ws_rejection_1\",\"delta\":\"Ready.\"}"
+        },
+        {
+          "direction": "server",
+          "kind": "text",
+          "body": "{\"type\":\"response.output_text.done\",\"item_id\":\"msg_ws_rejection_1\",\"text\":\"Ready.\"}"
+        },
+        {
+          "direction": "server",
+          "kind": "text",
+          "body": "{\"type\":\"response.output_item.done\",\"item\":{\"type\":\"message\",\"id\":\"msg_ws_rejection_1\",\"role\":\"assistant\",\"content\":[{\"type\":\"output_text\",\"text\":\"Ready.\"}]}}"
+        },
+        {
+          "direction": "server",
+          "kind": "text",
+          "body": "{\"type\":\"response.completed\",\"response\":{\"id\":\"resp_ws_rejection_1\"}}"
+        }
+      ]
+    },
+    {
+      "transport": "websocket",
+      "connection": {
+        "sequence": 1,
+        "url": "wss://api.openai.com/v1/responses",
+        "protocols": [],
+        "close": {
+          "code": 1000,
+          "reason": ""
+        }
+      },
+      "events": [
+        {
+          "direction": "client",
+          "kind": "text",
+          "body": "{\"type\":\"response.create\",\"model\":\"gpt-5.5\",\"input\":[{\"role\":\"user\",\"content\":[{\"type\":\"input_text\",\"text\":\"Reply exactly: Recovered.\"}]}],\"store\":false,\"max_output_tokens\":30,\"previous_response_id\":\"resp_ws_rejection_1\",\"include\":[\"reasoning.encrypted_content\"],\"reasoning\":{\"effort\":\"medium\",\"summary\":\"auto\"},\"text\":{\"verbosity\":\"low\"}}"
+        },
+        {
+          "direction": "server",
+          "kind": "text",
+          "body": "{\"type\":\"error\",\"error\":{\"code\":\"previous_response_not_found\",\"message\":\"Previous response not found\"}}"
+        },
+        {
+          "direction": "client",
+          "kind": "text",
+          "body": "{\"type\":\"response.create\",\"model\":\"gpt-5.5\",\"input\":[{\"role\":\"system\",\"content\":\"Follow the user's exact reply instruction.\"},{\"role\":\"user\",\"content\":[{\"type\":\"input_text\",\"text\":\"Reply exactly: Ready.\"}]},{\"type\":\"message\",\"id\":\"msg_ws_rejection_1\",\"role\":\"assistant\",\"content\":[{\"type\":\"output_text\",\"text\":\"Ready.\"}]},{\"role\":\"user\",\"content\":[{\"type\":\"input_text\",\"text\":\"Reply exactly: Recovered.\"}]}],\"store\":false,\"max_output_tokens\":30,\"include\":[\"reasoning.encrypted_content\"],\"reasoning\":{\"effort\":\"medium\",\"summary\":\"auto\"},\"text\":{\"verbosity\":\"low\"}}"
+        },
+        {
+          "direction": "server",
+          "kind": "text",
+          "body": "{\"type\":\"response.created\",\"response\":{\"id\":\"resp_ws_rejection_2\"}}"
+        },
+        {
+          "direction": "server",
+          "kind": "text",
+          "body": "{\"type\":\"response.output_item.added\",\"item\":{\"type\":\"message\",\"id\":\"msg_ws_rejection_2\",\"role\":\"assistant\",\"content\":[]}}"
+        },
+        {
+          "direction": "server",
+          "kind": "text",
+          "body": "{\"type\":\"response.output_text.delta\",\"item_id\":\"msg_ws_rejection_2\",\"delta\":\"Recovered.\"}"
+        },
+        {
+          "direction": "server",
+          "kind": "text",
+          "body": "{\"type\":\"response.output_text.done\",\"item_id\":\"msg_ws_rejection_2\",\"text\":\"Recovered.\"}"
+        },
+        {
+          "direction": "server",
+          "kind": "text",
+          "body": "{\"type\":\"response.output_item.done\",\"item\":{\"type\":\"message\",\"id\":\"msg_ws_rejection_2\",\"role\":\"assistant\",\"content\":[{\"type\":\"output_text\",\"text\":\"Recovered.\"}]}}"
+        },
+        {
+          "direction": "server",
+          "kind": "text",
+          "body": "{\"type\":\"response.completed\",\"response\":{\"id\":\"resp_ws_rejection_2\"}}"
+        }
+      ]
+    }
+  ]
+}

packages/ai/test/fixtures/recordings/pdf/openai-tool-result.json

@@ -1,7 +1,14 @@
 {
   "version": 1,
   "metadata": {
-    "tags": ["prefix:pdf", "pdf", "provider:openai", "protocol:openai-responses", "tool", "tool-result"],
+    "tags": [
+      "prefix:pdf",
+      "pdf",
+      "provider:openai",
+      "protocol:openai-responses",
+      "tool",
+      "tool-result"
+    ],
     "name": "pdf/openai-tool-result",
     "recordedAt": "2026-07-22T18:15:36.438Z"
   },
@@ -14,7 +21,7 @@
         "headers": {
           "content-type": "application/json"
         },
-        "body": "{\"model\":\"gpt-4o-mini\",\"input\":[{\"role\":\"system\",\"content\":\"Read the PDF returned by the tool and follow the user's response format exactly.\"},{\"role\":\"user\",\"content\":[{\"type\":\"input_text\",\"text\":\"Return only the verification code from the PDF.\"}]},{\"type\":\"function_call\",\"call_id\":\"call_pdf_1\",\"name\":\"read_pdf\",\"arguments\":\"{}\"},{\"type\":\"function_call_output\",\"call_id\":\"call_pdf_1\",\"output\":[{\"type\":\"input_text\",\"text\":\"PDF read successfully\"},{\"type\":\"input_file\",\"filename\":\"verification.pdf\",\"file_data\":\"data:application/pdf;base64,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\"}]}],\"tools\":[{\"type\":\"function\",\"name\":\"read_pdf\",\"description\":\"Read the attached PDF.\",\"parameters\":{\"type\":\"object\",\"properties\":{},\"additionalProperties\":false},\"strict\":false}],\"store\":false,\"max_output_tokens\":40,\"temperature\":0,\"stream\":true}"
+        "body": "{\"model\":\"gpt-4o-mini\",\"input\":[{\"role\":\"system\",\"content\":\"Read the PDF returned by the tool and follow the user's response format exactly.\"},{\"role\":\"user\",\"content\":[{\"type\":\"input_text\",\"text\":\"Return only the verification code from the PDF.\"}]},{\"type\":\"function_call\",\"call_id\":\"call_pdf_1\",\"name\":\"read_pdf\",\"arguments\":\"{}\"},{\"type\":\"function_call_output\",\"call_id\":\"call_pdf_1\",\"output\":[{\"type\":\"input_text\",\"text\":\"PDF read successfully\"},{\"type\":\"input_file\",\"filename\":\"verification.pdf\",\"file_data\":\"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\"}]}],\"tools\":[{\"type\":\"function\",\"name\":\"read_pdf\",\"description\":\"Read the attached PDF.\",\"parameters\":{\"type\":\"object\",\"properties\":{},\"additionalProperties\":false},\"strict\":false}],\"store\":false,\"max_output_tokens\":40,\"temperature\":0,\"stream\":true}"
       },
       "response": {
         "status": 200,

packages/ai/test/fixtures/recordings/pdf/openai-user-input.json

@@ -1,7 +1,13 @@
 {
   "version": 1,
   "metadata": {
-    "tags": ["prefix:pdf", "pdf", "provider:openai", "protocol:openai-responses", "user-input"],
+    "tags": [
+      "prefix:pdf",
+      "pdf",
+      "provider:openai",
+      "protocol:openai-responses",
+      "user-input"
+    ],
     "name": "pdf/openai-user-input",
     "recordedAt": "2026-07-22T18:15:34.867Z"
   },
@@ -14,7 +20,7 @@
         "headers": {
           "content-type": "application/json"
         },
-        "body": "{\"model\":\"gpt-4o-mini\",\"input\":[{\"role\":\"user\",\"content\":[{\"type\":\"input_file\",\"filename\":\"verification.pdf\",\"file_data\":\"data:application/pdf;base64,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\"},{\"type\":\"input_text\",\"text\":\"Return only the verification code from the PDF.\"}]}],\"store\":false,\"max_output_tokens\":40,\"temperature\":0,\"stream\":true}"
+        "body": "{\"model\":\"gpt-4o-mini\",\"input\":[{\"role\":\"user\",\"content\":[{\"type\":\"input_file\",\"filename\":\"verification.pdf\",\"file_data\":\"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\"},{\"type\":\"input_text\",\"text\":\"Return only the verification code from the PDF.\"}]}],\"store\":false,\"max_output_tokens\":40,\"temperature\":0,\"stream\":true}"
       },
       "response": {
         "status": 200,

packages/ai/test/fixtures/recordings/pdf/xai-tool-result.json

@@ -1,9 +1,17 @@
 {
   "version": 1,
   "metadata": {
-    "tags": ["prefix:pdf", "pdf", "provider:xai", "protocol:openai-responses", "tool", "tool-result"],
+    "tags": [
+      "prefix:pdf",
+      "pdf",
+      "provider:xai",
+      "protocol:xai-responses",
+      "tool",
+      "tool-result"
+    ],
     "name": "pdf/xai-tool-result",
-    "recordedAt": "2026-07-22T18:15:43.608Z"
+    "recordedAt": "2026-07-22T18:15:43.608Z",
+    "protocol": "xai-responses"
   },
   "interactions": [
     {
@@ -14,7 +22,7 @@
         "headers": {
           "content-type": "application/json"
         },
-        "body": "{\"model\":\"grok-4.5\",\"input\":[{\"role\":\"system\",\"content\":\"Read the PDF returned by the tool and follow the user's response format exactly.\"},{\"role\":\"user\",\"content\":[{\"type\":\"input_text\",\"text\":\"Return only the verification code from the PDF.\"}]},{\"type\":\"function_call\",\"call_id\":\"call_pdf_1\",\"name\":\"read_pdf\",\"arguments\":\"{}\"},{\"type\":\"function_call_output\",\"call_id\":\"call_pdf_1\",\"output\":[{\"type\":\"input_text\",\"text\":\"PDF read successfully\"},{\"type\":\"input_file\",\"filename\":\"verification.pdf\",\"file_data\":\"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\",\"mime_type\":\"application/pdf\"}]}],\"tools\":[{\"type\":\"function\",\"name\":\"read_pdf\",\"description\":\"Read the attached PDF.\",\"parameters\":{\"type\":\"object\",\"properties\":{},\"additionalProperties\":false},\"strict\":false}],\"store\":false,\"max_output_tokens\":40,\"temperature\":0,\"stream\":true}"
+        "body": "{\"model\":\"grok-4.5\",\"input\":[{\"role\":\"system\",\"content\":\"Read the PDF returned by the tool and follow the user's response format exactly.\"},{\"role\":\"user\",\"content\":[{\"type\":\"input_text\",\"text\":\"Return only the verification code from the PDF.\"}]},{\"type\":\"function_call\",\"call_id\":\"call_pdf_1\",\"name\":\"read_pdf\",\"arguments\":\"{}\"},{\"type\":\"function_call_output\",\"call_id\":\"call_pdf_1\",\"output\":[{\"type\":\"input_text\",\"text\":\"PDF read successfully\"},{\"type\":\"input_file\",\"filename\":\"verification.pdf\",\"file_data\":\"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\"}]}],\"tools\":[{\"type\":\"function\",\"name\":\"read_pdf\",\"description\":\"Read the attached PDF.\",\"parameters\":{\"type\":\"object\",\"properties\":{},\"additionalProperties\":false},\"strict\":false}],\"store\":false,\"max_output_tokens\":40,\"temperature\":0,\"stream\":true}"
       },
       "response": {
         "status": 200,

packages/ai/test/fixtures/recordings/pdf/xai-user-input.json

@@ -1,9 +1,16 @@
 {
   "version": 1,
   "metadata": {
-    "tags": ["prefix:pdf", "pdf", "provider:xai", "protocol:openai-responses", "user-input"],
+    "tags": [
+      "prefix:pdf",
+      "pdf",
+      "provider:xai",
+      "protocol:xai-responses",
+      "user-input"
+    ],
     "name": "pdf/xai-user-input",
-    "recordedAt": "2026-07-22T18:15:42.429Z"
+    "recordedAt": "2026-07-22T18:15:42.429Z",
+    "protocol": "xai-responses"
   },
   "interactions": [
     {
@@ -14,7 +21,7 @@
         "headers": {
           "content-type": "application/json"
         },
-        "body": "{\"model\":\"grok-4.5\",\"input\":[{\"role\":\"user\",\"content\":[{\"type\":\"input_file\",\"filename\":\"verification.pdf\",\"file_data\":\"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\",\"mime_type\":\"application/pdf\"},{\"type\":\"input_text\",\"text\":\"Return only the verification code from the PDF.\"}]}],\"store\":false,\"max_output_tokens\":40,\"temperature\":0,\"stream\":true}"
+        "body": "{\"model\":\"grok-4.5\",\"input\":[{\"role\":\"user\",\"content\":[{\"type\":\"input_file\",\"filename\":\"verification.pdf\",\"file_data\":\"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\"},{\"type\":\"input_text\",\"text\":\"Return only the verification code from the PDF.\"}]}],\"store\":false,\"max_output_tokens\":40,\"temperature\":0,\"stream\":true}"
       },
       "response": {
         "status": 200,

packages/ai/test/lib/sse.ts

@@ -10,6 +10,9 @@ export const sseEvents = (...chunks: ReadonlyArray<unknown>): string =>
 
 const formatChunk = (chunk: unknown) => `data: ${typeof chunk === "string" ? chunk : JSON.stringify(chunk)}\n\n`
 
+export const sseNamedEvent = (event: string, data: unknown): string =>
+  `event: ${event}\ndata: ${typeof data === "string" ? data : JSON.stringify(data)}`
+
 /**
  * Build an SSE body from already-serialized strings (used when the chunk shape
  * itself is part of what's being tested, e.g. malformed chunks).

packages/ai/test/llm.test.ts

@@ -121,7 +121,6 @@ describe("llm constructors", () => {
     const model = chatRoute.model({
       id: "kimi-k2",
       defaults: {
-        limits: { context: 128_000, output: 8_192 },
         generation: { maxTokens: 1_024, stop: ["END"] },
         providerOptions: { parallelToolCalls: false },
         http: { body: { extra_body: true } },
@@ -130,7 +129,6 @@ describe("llm constructors", () => {
     })
     const request = LLM.request({ model, prompt: "Say hello." })
 
-    expect(request.model.defaults?.limits).toEqual({ context: 128_000, output: 8_192 })
     expect(request.model.defaults?.generation).toEqual({ maxTokens: 1_024, stop: ["END"] })
     expect(request.model.defaults?.providerOptions).toEqual({ parallelToolCalls: false })
     expect(request.model.defaults?.http).toEqual({ body: { extra_body: true } })

packages/ai/test/provider-error.test.ts

@@ -82,6 +82,14 @@ describe("provider error classification", () => {
     ])
   })
 
+  test("classifies network error text as provider internal", () => {
+    expect(
+      ["network error", "network-error", "network_error"].map(
+        (message) => classifyProviderFailure({ message })._tag,
+      ),
+    ).toEqual(["ProviderInternal", "ProviderInternal", "ProviderInternal"])
+  })
+
   test("classifies nested provider codes when a top-level code is also present", () => {
     expect(
       [
@@ -98,3 +106,20 @@ describe("provider error classification", () => {
     expect(classifyProviderFailure({ message: "not-json" })._tag).toBe("UnknownProvider")
   })
 })
+
+describe("provider error rawBody classification", () => {
+  test("classifies overflow signals buried in the raw payload when the summary is vague", () => {
+    const reason = classifyProviderFailure({
+      message: "Request failed",
+      rawBody: '{"error":{"message":"This model\'s maximum context length is 40960 tokens"}}',
+    })
+    expect(reason._tag).toBe("InvalidRequest")
+    expect(reason).toMatchObject({ classification: "context-overflow" })
+  })
+
+  test("extracts nested codes from the raw payload", () => {
+    expect(
+      classifyProviderFailure({ message: "Request failed", rawBody: '{"error":{"code":"insufficient_quota"}}' })._tag,
+    ).toBe("QuotaExceeded")
+  })
+})

packages/ai/test/provider-package.test.ts

@@ -43,7 +43,6 @@ describe("provider package entrypoints", () => {
       baseURL: "https://provider.example.test/v1",
       headers: { "x-application": "opencode" },
       body: { service_tier: "priority" },
-      limits: { context: 200_000, output: 64_000 },
     }
     const openrouter = OpenRouter.model("anthropic/claude-sonnet-4", {
       ...settings,
@@ -58,7 +57,6 @@ describe("provider package entrypoints", () => {
       expect(selected.route.endpoint.baseURL).toBe(settings.baseURL)
       expect(selected.route.defaults.headers).toEqual(settings.headers)
       expect(selected.route.defaults.http?.body).toEqual(settings.body)
-      expect(selected.route.defaults.limits).toEqual(settings.limits)
     }
     expect(openrouter.route.defaults.providerOptions).toEqual({ usage: true })
     expect(xai.route.defaults.providerOptions).toMatchObject({ reasoningEffort: "high", store: false })
@@ -70,14 +68,12 @@ describe("provider package entrypoints", () => {
       baseURL: "https://api.openai.test/v1",
       headers: { "x-application": "opencode" },
       body: { service_tier: "priority" },
-      limits: { context: 200_000, output: 64_000 },
       unrelatedInheritedSetting: true,
     })
 
     expect(selected.route.id).toBe("openai-responses")
     expect(selected.route.defaults.headers).toEqual({ "x-application": "opencode" })
     expect(selected.route.defaults.http?.body).toEqual({ service_tier: "priority" })
-    expect(selected.route.defaults.limits).toEqual({ context: 200_000, output: 64_000 })
   })
 
   test("maps OpenAI-compatible Responses settings onto the executable model", async () => {
@@ -88,7 +84,6 @@ describe("provider package entrypoints", () => {
       provider: "example",
       headers: { "x-application": "opencode" },
       body: { service_tier: "priority" },
-      limits: { context: 200_000, output: 64_000 },
       providerOptions: { reasoningEffort: "low", store: true },
     })
 
@@ -100,7 +95,6 @@ describe("provider package entrypoints", () => {
     })
     expect(selected.route.defaults.headers).toEqual({ "x-application": "opencode" })
     expect(selected.route.defaults.http?.body).toEqual({ service_tier: "priority" })
-    expect(selected.route.defaults.limits).toEqual({ context: 200_000, output: 64_000 })
     expect(selected.route.defaults.providerOptions).toEqual({ reasoningEffort: "low", store: true })
   })
 
@@ -112,7 +106,6 @@ describe("provider package entrypoints", () => {
       provider: "example",
       headers: { "x-application": "opencode" },
       body: { metadata: { user_id: "user_1" } },
-      limits: { context: 200_000, output: 64_000 },
       providerOptions: { effort: "low" },
     })
 
@@ -124,7 +117,6 @@ describe("provider package entrypoints", () => {
     })
     expect(selected.route.defaults.headers).toEqual({ "x-application": "opencode" })
     expect(selected.route.defaults.http?.body).toEqual({ metadata: { user_id: "user_1" } })
-    expect(selected.route.defaults.limits).toEqual({ context: 200_000, output: 64_000 })
     expect(selected.route.defaults.providerOptions).toEqual({ effort: "low" })
   })
 
@@ -185,7 +177,6 @@ describe("provider package entrypoints", () => {
       resourceName: "opencode-test",
       headers: { "x-application": "opencode" },
       body: { service_tier: "priority" },
-      limits: { context: 200_000, output: 64_000 },
     }
 
     const responses = AzureResponses.model("deployment", settings)
@@ -196,7 +187,6 @@ describe("provider package entrypoints", () => {
     expect(responses.route.endpoint.baseURL).toBe("https://opencode-test.openai.azure.com/openai/v1")
     expect(responses.route.defaults.headers).toEqual({ "x-application": "opencode" })
     expect(responses.route.defaults.http?.body).toEqual({ service_tier: "priority" })
-    expect(responses.route.defaults.limits).toEqual({ context: 200_000, output: 64_000 })
     expect(chat.route.id).toBe("azure-openai-chat")
   })
 
@@ -228,7 +218,6 @@ describe("provider package entrypoints", () => {
       baseURL: "https://generativelanguage.test/v1beta",
       headers: { "x-application": "opencode" },
       body: { safetySettings: [] },
-      limits: { context: 1_000_000, output: 65_536 },
       providerOptions: { thinkingConfig: { thinkingBudget: 1_024 } },
     })
 
@@ -236,7 +225,6 @@ describe("provider package entrypoints", () => {
     expect(selected.route.endpoint.baseURL).toBe("https://generativelanguage.test/v1beta")
     expect(selected.route.defaults.headers).toEqual({ "x-application": "opencode" })
     expect(selected.route.defaults.http?.body).toEqual({ safetySettings: [] })
-    expect(selected.route.defaults.limits).toEqual({ context: 1_000_000, output: 65_536 })
     expect(selected.route.defaults.providerOptions).toEqual({ thinkingConfig: { thinkingBudget: 1_024 } })
   })
 
@@ -250,7 +238,6 @@ describe("provider package entrypoints", () => {
       apiKey: "fixture",
       headers: { "x-application": "opencode" },
       body: { safetySettings: [] },
-      limits: { context: 1_000_000, output: 65_536 },
     })
     const messages = GoogleVertexMessages.model("claude-sonnet-4-6", {
       accessToken: "fixture",
@@ -274,7 +261,6 @@ describe("provider package entrypoints", () => {
     expect(gemini.route.endpoint.baseURL).toBe("https://aiplatform.googleapis.com/v1/publishers/google")
     expect(gemini.route.defaults.headers).toEqual({ "x-application": "opencode" })
     expect(gemini.route.defaults.http?.body).toEqual({ safetySettings: [] })
-    expect(gemini.route.defaults.limits).toEqual({ context: 1_000_000, output: 65_536 })
     expect(
       GoogleVertex.model("gemini-3.5-flash", {
         accessToken: "fixture",

packages/ai/test/provider/anthropic-messages.test.ts

@@ -8,7 +8,7 @@ import * as AnthropicMessages from "../../src/protocols/anthropic-messages.js"
 import { continuationRequest, nativeAnthropicMessagesContinuation } from "../continuation-scenarios.js"
 import { it } from "../lib/effect.js"
 import { dynamicResponse, fixedResponse } from "../lib/http.js"
-import { sseEvents } from "../lib/sse.js"
+import { sseEvents, sseNamedEvent, sseRaw } from "../lib/sse.js"
 
 const model = AnthropicMessages.route
   .with({ endpoint: { baseURL: "https://api.anthropic.test/v1/" }, auth: Auth.header("x-api-key", "test") })
@@ -322,11 +322,34 @@ describe("Anthropic Messages route", () => {
           { role: "user", content: [{ type: "tool_result", tool_use_id: "call_1", content: '{"forecast":"sunny"}' }] },
         ],
         stream: true,
-        max_tokens: 4096,
+        max_tokens: 32_000,
       })
     }),
   )
 
+  it.effect("scrubs outbound tool call IDs without truncating them", () =>
+    Effect.gen(function* () {
+      const id = `functions.lookup:1|${"x".repeat(64)}`
+      const scrubbed = `functions_lookup_1_${"x".repeat(64)}`
+      const prepared = yield* compileRequest(
+        LLM.request({
+          model,
+          messages: [
+            Message.assistant([ToolCallPart.make({ id, name: "lookup", input: {} })]),
+            Message.tool({ id, name: "lookup", result: "done" }),
+          ],
+          cache: "none",
+        }),
+      )
+
+      expect(prepared.body.messages).toMatchObject([
+        { role: "assistant", content: [{ type: "tool_use", id: scrubbed, name: "lookup", input: {} }] },
+        { role: "user", content: [{ type: "tool_result", tool_use_id: scrubbed }] },
+      ])
+      expect(scrubbed.length).toBeGreaterThan(64)
+    }),
+  )
+
   it.effect("batches parallel tool results into one Anthropic user message", () =>
     Effect.gen(function* () {
       const prepared = yield* compileRequest(
@@ -640,7 +663,60 @@ describe("Anthropic Messages route", () => {
     }),
   )
 
-  it.effect("maps thinking tokens and preserves unknown Anthropic usage fields", () =>
+  it.effect("ignores unknown named SSE events", () =>
+    Effect.gen(function* () {
+      const response = yield* LLMClient.generate(request).pipe(
+        Effect.provide(
+          fixedResponse(
+            sseRaw(
+              sseNamedEvent("message_start", {
+                type: "message_start",
+                message: { usage: { input_tokens: 5 } },
+              }),
+              sseNamedEvent("proxy.stats", "not json"),
+              sseNamedEvent("content_block_start", {
+                type: "content_block_start",
+                index: 0,
+                content_block: { type: "text", text: "" },
+              }),
+              sseNamedEvent("content_block_delta", {
+                type: "content_block_delta",
+                index: 0,
+                delta: { type: "text_delta", text: "Hello" },
+              }),
+              sseNamedEvent("content_block_stop", { type: "content_block_stop", index: 0 }),
+              sseNamedEvent("message_delta", {
+                type: "message_delta",
+                delta: { stop_reason: "end_turn" },
+                usage: { output_tokens: 1 },
+              }),
+              sseNamedEvent("message_stop", { type: "message_stop" }),
+              sseNamedEvent("proxy.done", "still not json"),
+            ),
+          ),
+        ),
+      )
+
+      expect(response.message.content).toEqual([{ type: "text", text: "Hello" }])
+      expect(response.finishReason).toEqual({ normalized: "stop", raw: "end_turn" })
+    }),
+  )
+
+  it.effect("rejects malformed recognized SSE events", () =>
+    Effect.gen(function* () {
+      const error = yield* LLMClient.generate(request).pipe(
+        Effect.provide(fixedResponse(sseRaw(sseNamedEvent("message_start", "[DONE]")))),
+        Effect.flip,
+      )
+
+      expect(error.reason).toMatchObject({
+        _tag: "InvalidProviderOutput",
+        message: "Invalid anthropic/anthropic-messages stream event",
+      })
+    }),
+  )
+
+  it.effect("maps nullable input tokens and preserves unknown Anthropic usage fields", () =>
     Effect.gen(function* () {
       const response = yield* LLMClient.generate(request).pipe(
         Effect.provide(
@@ -663,6 +739,7 @@ describe("Anthropic Messages route", () => {
                 type: "message_delta",
                 delta: { stop_reason: "end_turn" },
                 usage: {
+                  input_tokens: null,
                   output_tokens: 8,
                   server_tool_use: { web_search_requests: 2, terminal_counter: 3 },
                   output_tokens_details: { terminal_detail: "preserved" },
@@ -682,7 +759,7 @@ describe("Anthropic Messages route", () => {
         totalTokens: 15,
         providerMetadata: {
           anthropic: {
-            input_tokens: 5,
+            input_tokens: null,
             cache_read_input_tokens: 2,
             service_tier: "standard",
             cache_creation: { ephemeral_5m_input_tokens: 1 },
@@ -1339,14 +1416,14 @@ describe("Anthropic Messages route", () => {
             Message.assistant([
               {
                 type: "tool-call",
-                id: "srvtoolu_abc",
+                id: "srvtoolu.abc",
                 name: "web_search",
                 input: { query: "effect 4" },
                 providerExecuted: true,
               },
               {
                 type: "tool-result",
-                id: "srvtoolu_abc",
+                id: "srvtoolu.abc",
                 name: "web_search",
                 result: { type: "json", value: [{ url: "https://example.com" }] },
                 providerExecuted: true,

packages/ai/test/provider/gemini.test.ts

@@ -139,6 +139,95 @@ describe("Gemini route", () => {
     }),
   )
 
+  it.effect("keeps system updates separate from function responses", () =>
+    Effect.gen(function* () {
+      const prepared = yield* compileRequest(
+        LLM.request({
+          model,
+          messages: [
+            Message.assistant([ToolCallPart.make({ id: "call_1", name: "lookup", input: { query: "weather" } })]),
+            Message.tool({ id: "call_1", name: "lookup", result: "done", resultType: "text" }),
+            Message.system("Update."),
+            Message.system("Later update."),
+          ],
+        }),
+      )
+
+      expect(prepared.body.contents).toEqual([
+        {
+          role: "model",
+          parts: [{ functionCall: { id: undefined, name: "lookup", args: { query: "weather" } } }],
+        },
+        {
+          role: "user",
+          parts: [
+            {
+              functionResponse: {
+                id: undefined,
+                name: "lookup",
+                response: { name: "lookup", content: "done" },
+              },
+            },
+          ],
+        },
+        {
+          role: "user",
+          parts: [
+            { text: "<system-update>\nUpdate.\n</system-update>" },
+            { text: "<system-update>\nLater update.\n</system-update>" },
+          ],
+        },
+      ])
+    }),
+  )
+
+  it.effect("merges parallel tool results into one function-response turn", () =>
+    Effect.gen(function* () {
+      const prepared = yield* compileRequest(
+        LLM.request({
+          model,
+          messages: [
+            Message.assistant([
+              ToolCallPart.make({ id: "call_1", name: "lookup", input: { query: "weather" } }),
+              ToolCallPart.make({ id: "call_2", name: "lookup", input: { query: "time" } }),
+            ]),
+            Message.tool({ id: "call_1", name: "lookup", result: "sunny", resultType: "text" }),
+            Message.tool({ id: "call_2", name: "lookup", result: "noon", resultType: "text" }),
+          ],
+        }),
+      )
+
+      expect(prepared.body.contents).toEqual([
+        {
+          role: "model",
+          parts: [
+            { functionCall: { id: undefined, name: "lookup", args: { query: "weather" } } },
+            { functionCall: { id: undefined, name: "lookup", args: { query: "time" } } },
+          ],
+        },
+        {
+          role: "user",
+          parts: [
+            {
+              functionResponse: {
+                id: undefined,
+                name: "lookup",
+                response: { name: "lookup", content: "sunny" },
+              },
+            },
+            {
+              functionResponse: {
+                id: undefined,
+                name: "lookup",
+                response: { name: "lookup", content: "noon" },
+              },
+            },
+          ],
+        },
+      ])
+    }),
+  )
+
   it.effect("prepares multimodal user input and tool history", () =>
     Effect.gen(function* () {
       const prepared = yield* compileRequest(
@@ -239,14 +328,18 @@ describe("Gemini route", () => {
               functionResponse: {
                 name: "read",
                 response: { name: "read", content: "Image read successfully" },
-                parts: [
-                  { inlineData: { mimeType: "image/png", data: "AAECAw==" } },
-                  { inlineData: { mimeType: "application/pdf", data: "JVBERi0xLjQ=" } },
-                ],
               },
             },
           ],
         },
+        {
+          role: "user",
+          parts: [
+            { text: "Attached media from tool result:" },
+            { inlineData: { mimeType: "image/png", data: "AAECAw==" } },
+            { inlineData: { mimeType: "application/pdf", data: "JVBERi0xLjQ=" } },
+          ],
+        },
       ])
       expect(JSON.stringify(prepared.body.contents)).not.toContain('"content":"AAECAw=="')
     }),
@@ -279,11 +372,161 @@ describe("Gemini route", () => {
               functionResponse: {
                 name: "read",
                 response: { name: "read", content: "" },
-                parts: [{ inlineData: { mimeType: "image/jpeg", data: "/9j/" } }],
               },
             },
           ],
         },
+        {
+          role: "user",
+          parts: [
+            { text: "Attached media from tool result:" },
+            { inlineData: { mimeType: "image/jpeg", data: "/9j/" } },
+          ],
+        },
+      ])
+    }),
+  )
+
+  it.effect("nests media inside function responses for gemini 3", () =>
+    Effect.gen(function* () {
+      const prepared = yield* compileRequest(
+        LLM.request({
+          model: gemini3,
+          messages: [
+            Message.assistant([
+              ToolCallPart.make({
+                id: "call_image",
+                name: "read",
+                input: { path: "pixel.png" },
+                providerMetadata: { google: { thoughtSignature: "sig_1" } },
+              }),
+            ]),
+            Message.tool({
+              id: "call_image",
+              name: "read",
+              result: {
+                type: "content",
+                value: [
+                  { type: "text", text: "Image read successfully" },
+                  { type: "file", uri: "data:image/png;base64,AAECAw==", mime: "image/png", name: "pixel.png" },
+                ],
+              },
+            }),
+          ],
+        }),
+      )
+
+      expect(prepared.body.contents).toEqual([
+        {
+          role: "model",
+          parts: [{ functionCall: { name: "read", args: { path: "pixel.png" } }, thoughtSignature: "sig_1" }],
+        },
+        {
+          role: "user",
+          parts: [
+            {
+              functionResponse: {
+                name: "read",
+                response: { name: "read", content: "Image read successfully" },
+                parts: [{ inlineData: { mimeType: "image/png", data: "AAECAw==" } }],
+              },
+            },
+          ],
+        },
+      ])
+    }),
+  )
+
+  it.effect("flushes pending media before system update text", () =>
+    Effect.gen(function* () {
+      const prepared = yield* compileRequest(
+        LLM.request({
+          model,
+          messages: [
+            Message.assistant([ToolCallPart.make({ id: "call_1", name: "shot", input: {} })]),
+            Message.tool({
+              id: "call_1",
+              name: "shot",
+              result: {
+                type: "content",
+                value: [{ type: "file", uri: "data:image/png;base64,AAEC", mime: "image/png" }],
+              },
+            }),
+            Message.system("Update."),
+          ],
+        }),
+      )
+
+      expect(prepared.body.contents).toEqual([
+        { role: "model", parts: [{ functionCall: { name: "shot", args: {} } }] },
+        {
+          role: "user",
+          parts: [{ functionResponse: { name: "shot", response: { name: "shot", content: "" } } }],
+        },
+        {
+          role: "user",
+          parts: [
+            { text: "Attached media from tool result:" },
+            { inlineData: { mimeType: "image/png", data: "AAEC" } },
+            { text: "<system-update>\nUpdate.\n</system-update>" },
+          ],
+        },
+      ])
+    }),
+  )
+
+  it.effect("collects legacy tool media into one turn after merged responses", () =>
+    Effect.gen(function* () {
+      const prepared = yield* compileRequest(
+        LLM.request({
+          model,
+          messages: [
+            Message.assistant([
+              ToolCallPart.make({ id: "call_1", name: "shot", input: {} }),
+              ToolCallPart.make({ id: "call_2", name: "shot", input: {} }),
+            ]),
+            Message.tool({
+              id: "call_1",
+              name: "shot",
+              result: {
+                type: "content",
+                value: [{ type: "file", uri: "data:image/png;base64,AAEC", mime: "image/png" }],
+              },
+            }),
+            Message.tool({
+              id: "call_2",
+              name: "shot",
+              result: {
+                type: "content",
+                value: [{ type: "text", text: "no image here" }],
+              },
+            }),
+          ],
+        }),
+      )
+
+      expect(prepared.body.contents).toEqual([
+        {
+          role: "model",
+          parts: [
+            { functionCall: { name: "shot", args: {} } },
+            { functionCall: { name: "shot", args: {} } },
+          ],
+        },
+        {
+          role: "user",
+          parts: [
+            { functionResponse: { name: "shot", response: { name: "shot", content: "" } } },
+            { functionResponse: { name: "shot", response: { name: "shot", content: "no image here" } } },
+          ],
+        },
+        {
+          role: "user",
+          parts: [
+            { text: "Attached media from tool result:" },
+            { inlineData: { mimeType: "image/png", data: "AAEC" } },
+          ],
+        },
       ])
     }),
   )
@@ -605,7 +848,7 @@ describe("Gemini route", () => {
         providerMetadata: { google: { thoughtSignature: "thought_sig" } },
       })
       expect(toolCall).toMatchObject({
-        id: "tool_0",
+        id: "provider_call",
         providerMetadata: { google: { functionCallId: "provider_call", thoughtSignature: "tool_sig" } },
       })
       expect(response.events.findIndex((event) => event.type === "reasoning-end")).toBeLessThan(
@@ -619,14 +862,14 @@ describe("Gemini route", () => {
             Message.assistant([
               { type: "reasoning", text: "thinking", providerMetadata: reasoningEnd?.providerMetadata },
               ToolCallPart.make({
-                id: "tool_0",
+                id: "provider_call",
                 name: "lookup",
                 input: { query: "weather" },
                 providerMetadata: toolCall?.providerMetadata,
               }),
             ]),
             Message.tool({
-              id: "tool_0",
+              id: "provider_call",
               name: "lookup",
               result: "done",
               resultType: "text",
@@ -662,6 +905,61 @@ describe("Gemini route", () => {
     }),
   )
 
+  it.effect("preserves thoughtSignature on visible text parts", () =>
+    Effect.gen(function* () {
+      const body = sseEvents({
+        candidates: [
+          {
+            content: { role: "model", parts: [{ text: "All done.", thoughtSignature: "text_sig" }] },
+            finishReason: "STOP",
+          },
+        ],
+      })
+      const response = yield* LLMClient.generate(request).pipe(Effect.provide(fixedResponse(body)))
+      const delta = response.events.find((event) => event.type === "text-delta")
+      expect(delta).toMatchObject({
+        id: "text-0",
+        text: "All done.",
+        providerMetadata: { google: { thoughtSignature: "text_sig" } },
+      })
+
+      const prepared = yield* compileRequest(
+        LLM.request({
+          model,
+          messages: [Message.assistant([{ type: "text", text: "All done.", providerMetadata: delta?.providerMetadata }])],
+        }),
+      )
+      expect(prepared.body.contents).toEqual([
+        { role: "model", parts: [{ text: "All done.", thoughtSignature: "text_sig" }] },
+      ])
+    }),
+  )
+
+  it.effect("flushes a trailing empty signed text part at block close", () =>
+    Effect.gen(function* () {
+      const body = sseEvents({
+        candidates: [
+          {
+            content: {
+              role: "model",
+              parts: [{ text: "Working." }, { text: "", thoughtSignature: "tail_sig" }],
+            },
+            finishReason: "STOP",
+          },
+        ],
+      })
+      const response = yield* LLMClient.generate(request).pipe(Effect.provide(fixedResponse(body)))
+      const delta = response.events.find((event) => event.type === "text-delta")
+      const end = response.events.find((event) => event.type === "text-end")
+
+      expect(delta).toMatchObject({ id: "text-0", text: "Working.", providerMetadata: undefined })
+      expect(end).toMatchObject({
+        id: "text-0",
+        providerMetadata: { google: { thoughtSignature: "tail_sig" } },
+      })
+    }),
+  )
+
   it.effect("replays unsigned Gemini 3 tool calls with the validator bypass sentinel", () =>
     Effect.gen(function* () {
       const prepared = yield* compileRequest(
@@ -803,21 +1101,17 @@ describe("Gemini route", () => {
         providerMetadata: { google: { promptTokenCount: 5, candidatesTokenCount: 1 } },
       })
 
-      expect(response.toolCalls).toEqual([
-        {
-          type: "tool-call",
-          id: "tool_0",
-          name: "lookup",
-          input: { query: "weather" },
-          providerExecuted: undefined,
-          providerMetadata: undefined,
-        },
-      ])
+      expect(response.toolCalls[0].id).toMatch(/^tool_[0-9a-zA-Z]+$/)
+      expect(response.toolCalls[0]).toMatchObject({
+        type: "tool-call",
+        name: "lookup",
+        input: { query: "weather" },
+      })
       expect(response.events).toEqual([
         { type: "step-start", index: 0 },
         {
           type: "tool-call",
-          id: "tool_0",
+          id: response.toolCalls[0].id,
           name: "lookup",
           input: { query: "weather" },
           providerExecuted: undefined,
@@ -860,7 +1154,8 @@ describe("Gemini route", () => {
         ),
       )
 
-      expect(response.toolCalls).toEqual([{ type: "tool-call", id: "tool_0", name: "ping", input: {} }])
+      expect(response.toolCalls[0].id).toMatch(/^tool_[0-9a-zA-Z]+$/)
+      expect(response.toolCalls).toMatchObject([{ type: "tool-call", name: "ping", input: {} }])
     }),
   )
 
@@ -900,7 +1195,7 @@ describe("Gemini route", () => {
             content: {
               role: "model",
               parts: [
-                { functionCall: { id: "tool_0", name: "lookup", args: { query: "weather" } } },
+                { functionCall: { id: "call_0", name: "lookup", args: { query: "weather" } } },
                 { functionCall: { name: "lookup", args: { query: "news" } } },
               ],
             },
@@ -914,16 +1209,20 @@ describe("Gemini route", () => {
         }),
       ).pipe(Effect.provide(fixedResponse(body)))
 
-      expect(response.toolCalls).toEqual([
-        {
-          type: "tool-call",
-          id: "tool_0",
-          name: "lookup",
-          input: { query: "weather" },
-          providerMetadata: { google: { functionCallId: "tool_0" } },
-        },
-        { type: "tool-call", id: "tool_1", name: "lookup", input: { query: "news" } },
-      ])
+      expect(response.toolCalls[0]).toMatchObject({
+        type: "tool-call",
+        id: "call_0",
+        name: "lookup",
+        input: { query: "weather" },
+        providerMetadata: { google: { functionCallId: "call_0" } },
+      })
+      expect(response.toolCalls[1]).toMatchObject({
+        type: "tool-call",
+        name: "lookup",
+        input: { query: "news" },
+      })
+      expect(response.toolCalls[1].id).toMatch(/^tool_[0-9a-zA-Z]+$/)
+      expect(response.toolCalls[0].id).not.toBe(response.toolCalls[1].id)
       expect(response.events.at(-1)).toMatchObject({
         type: "finish",
         reason: { normalized: "tool-calls", raw: "STOP" },
@@ -931,6 +1230,31 @@ describe("Gemini route", () => {
     }),
   )
 
+  it.effect("assigns distinct unique fallback ids across separate requests", () =>
+    Effect.gen(function* () {
+      const body = sseEvents({
+        candidates: [
+          {
+            content: {
+              role: "model",
+              parts: [{ functionCall: { name: "lookup", args: { query: "weather" } } }],
+            },
+            finishReason: "STOP",
+          },
+        ],
+      })
+      const req = LLMRequest.update(request, {
+        tools: [ToolDefinition.make({ name: "lookup", description: "Lookup data", inputSchema: { type: "object" } })],
+      })
+      const first = yield* LLMClient.generate(req).pipe(Effect.provide(fixedResponse(body)))
+      const second = yield* LLMClient.generate(req).pipe(Effect.provide(fixedResponse(body)))
+
+      expect(first.toolCalls[0].id).toMatch(/^tool_[0-9a-zA-Z]+$/)
+      expect(second.toolCalls[0].id).toMatch(/^tool_[0-9a-zA-Z]+$/)
+      expect(first.toolCalls[0].id).not.toBe(second.toolCalls[0].id)
+    }),
+  )
+
   it.effect("maps length and content-filter finish reasons", () =>
     Effect.gen(function* () {
       const length = yield* LLMClient.generate(request).pipe(

packages/ai/test/provider/google-vertex.test.ts

@@ -96,7 +96,7 @@ describe("Google Vertex providers", () => {
                 "https://aiplatform.eu.rep.googleapis.com/v1/projects/vertex-project/locations/eu/publishers/anthropic/models/claude-sonnet-4-6:streamRawPredict",
               )
               expect(request.headers.get("authorization")).toBe("Bearer vertex-token")
-              expect(request.headers.get("anthropic-version")).toBeNull()
+              expect(request.headers.get("anthropic-version")).toBe("2023-06-01")
               const body = yield* Effect.promise(() => request.json())
               expect(body).toMatchObject({
                 anthropic_version: "vertex-2023-10-16",

packages/ai/test/provider/openai-chat.test.ts

@@ -664,6 +664,74 @@ describe("OpenAI Chat route", () => {
     }),
   )
 
+  it.effect("preserves streamed refusals as ordinary assistant text", () =>
+    Effect.gen(function* () {
+      const response = yield* LLMClient.generate(request).pipe(
+        Effect.provide(
+          fixedResponse(
+            sseEvents(
+              deltaChunk({ role: "assistant", refusal: "I can't" }),
+              deltaChunk({ refusal: " help with that." }),
+              deltaChunk({}, "stop"),
+            ),
+          ),
+        ),
+      )
+
+      expect(response.text).toBe("I can't help with that.")
+      expect(response.finishReason).toEqual({ normalized: "stop", raw: "stop" })
+      expect(response.message.content).toEqual([{ type: "text", text: "I can't help with that." }])
+
+      const replay = yield* compileRequest(LLM.request({ model, messages: [response.message] }))
+      expect(replay.body.messages).toEqual([{ role: "assistant", content: "I can't help with that." }])
+    }),
+  )
+
+  it.effect("orders metadata-only reasoning before refusal output", () =>
+    Effect.gen(function* () {
+      const response = yield* LLMClient.generate(request).pipe(
+        Effect.provide(
+          fixedResponse(
+            sseEvents(
+              { choices: [{ delta: { reasoning_details: [] } }] },
+              deltaChunk({ refusal: "I can't help with that." }),
+              deltaChunk({}, "stop"),
+            ),
+          ),
+        ),
+      )
+
+      expect(response.message.content).toEqual([
+        { type: "reasoning", text: "", providerMetadata: { openai: { reasoningDetails: [] } } },
+        {
+          type: "text",
+          text: "I can't help with that.",
+        },
+      ])
+    }),
+  )
+
+  it.effect("joins content and refusal deltas into ordinary assistant text", () =>
+    Effect.gen(function* () {
+      const response = yield* LLMClient.generate(request).pipe(
+        Effect.provide(
+          fixedResponse(
+            sseEvents(
+              deltaChunk({ refusal: "No." }),
+              deltaChunk({ content: " Alternative." }),
+              deltaChunk({ refusal: " Still no." }),
+              deltaChunk({}, "stop"),
+            ),
+          ),
+        ),
+      )
+
+      expect(response.text).toBe("No. Alternative. Still no.")
+      expect(response.events.filter(LLMEvent.is.textStart).map((event) => event.id)).toEqual(["text-0"])
+      expect(response.events.filter(LLMEvent.is.textEnd).map((event) => event.id)).toEqual(["text-0"])
+    }),
+  )
+
   it.effect("parses and replays OpenAI-compatible reasoning fields", () =>
     Effect.gen(function* () {
       const fields = ["reasoning_content", "reasoning", "reasoning_text"] as const

packages/ai/test/provider/openai-compatible-responses.test.ts

@@ -73,7 +73,7 @@ describe("Open Responses-compatible route", () => {
       expect(prepared.body.input).toEqual([
         { role: "user", content: [{ type: "input_text", text: "Before." }] },
         { role: "developer", content: "Operator update." },
-        { role: "assistant", content: [{ type: "output_text", text: "After." }] },
+        { type: "message", role: "assistant", content: [{ type: "output_text", text: "After." }] },
       ])
     }),
   )
@@ -93,7 +93,7 @@ describe("Open Responses-compatible route", () => {
     }),
   )
 
-  it.effect("omits OpenAI-only nullable phases from the Open Responses baseline", () =>
+  it.effect("preserves nullable phases in the forgiving Open Responses baseline", () =>
     Effect.gen(function* () {
       const model = configure({
         apiKey: "test-key",
@@ -113,11 +113,69 @@ describe("Open Responses-compatible route", () => {
       )
 
       expect(prepared.body).toMatchObject({
-        input: [{ role: "assistant", content: [{ type: "output_text", text: "Unclassified." }] }],
+        input: [
+          { type: "message", role: "assistant", content: [{ type: "output_text", text: "Unclassified." }], phase: null },
+        ],
       })
     }),
   )
 
+  it.effect("preserves standard refusal content as ordinary assistant text", () =>
+    Effect.gen(function* () {
+      const model = configure({
+        apiKey: "test-key",
+        baseURL: "https://responses.example.test/v1",
+        provider: "example",
+      }).model("example-model")
+      const response = yield* LLMClient.generate(LLM.request({ model, prompt: "Unsafe request" })).pipe(
+        Effect.provide(
+          fixedResponse(
+            sseEvents(
+              {
+                type: "response.output_item.added",
+                output_index: 0,
+                item: { type: "message", id: "msg_refusal", content: [] },
+              },
+              {
+                type: "response.refusal.done",
+                item_id: "msg_refusal",
+                refusal: "I can't help with that.",
+              },
+              {
+                type: "response.output_item.done",
+                output_index: 0,
+                item: {
+                  type: "message",
+                  id: "msg_refusal",
+                  content: [{ type: "refusal", refusal: "I can't help with that." }],
+                },
+              },
+              { type: "response.completed", response: { id: "resp_1" } },
+            ),
+          ),
+        ),
+      )
+
+      expect(response.message.content).toEqual([
+        {
+          type: "text",
+          text: "I can't help with that.",
+          providerMetadata: { openresponses: { itemId: "msg_refusal" } },
+        },
+      ])
+
+      const prepared = yield* compileRequest(LLM.request({ model, messages: [response.message] }))
+      expect(prepared.body.input).toEqual([
+        {
+          type: "message",
+          id: "msg_refusal",
+          role: "assistant",
+          content: [{ type: "output_text", text: "I can't help with that." }],
+        },
+      ])
+    }),
+  )
+
   it.effect("reads standard Open Responses options", () =>
     Effect.gen(function* () {
       const model = configure({
@@ -126,6 +184,10 @@ describe("Open Responses-compatible route", () => {
         providerOptions: {
           reasoningEffort: "low",
           store: true,
+          metadata: { environment: "test" },
+          safetyIdentifier: "user_123",
+          streamOptions: { includeObfuscation: false },
+          topLogprobs: 3,
           truncation: "auto",
           allowedTools: { toolNames: ["lookup"] },
           maxToolCalls: 2,
@@ -136,6 +198,7 @@ describe("Open Responses-compatible route", () => {
         LLM.request({
           model,
           prompt: "Think.",
+          generation: { presencePenalty: 0.2, frequencyPenalty: -0.1 },
           tools: [ToolDefinition.make({ name: "lookup", description: "Lookup data", inputSchema: { type: "object" } })],
         }),
       )
@@ -143,6 +206,12 @@ describe("Open Responses-compatible route", () => {
       expect(prepared.body).toMatchObject({
         reasoning: { effort: "low" },
         store: true,
+        metadata: { environment: "test" },
+        safety_identifier: "user_123",
+        stream_options: { include_obfuscation: false },
+        top_logprobs: 3,
+        presence_penalty: 0.2,
+        frequency_penalty: -0.1,
         truncation: "auto",
         tool_choice: {
           type: "allowed_tools",

packages/ai/test/provider/openai-responses-websocket.recorded.test.ts

@@ -0,0 +1,216 @@
+import { describe, expect } from "bun:test"
+import { Effect, Stream } from "effect"
+import { Socket } from "effect/unstable/socket"
+import { LLM, LLMRequest, Message, ToolRuntime } from "../../src/index.js"
+import {
+  LLMClient,
+  WebSocketTransport,
+  type ChannelCheckpoint,
+  type ChannelObservation,
+  type WebSocketChannelExchange,
+  type WebSocketChannelExecutor,
+  type WebSocketConnection,
+} from "../../src/route.js"
+import { configure } from "../../src/providers/openai.js"
+import { decodeJson } from "../../src/protocols/shared.js"
+import { weatherRuntimeTool, weatherTool, weatherToolName } from "../recorded-scenarios.js"
+import { recordedTests } from "../recorded-test.js"
+
+const model = configure({ apiKey: process.env.OPENAI_API_KEY ?? "fixture" }).responses("gpt-5.5")
+const recorded = recordedTests({
+  prefix: "openai-responses-websocket",
+  provider: "openai",
+  protocol: "openai-responses",
+  requires: ["OPENAI_API_KEY"],
+  tags: ["transport:websocket"],
+  metadata: { transport: "websocket", model: model.id },
+})
+
+const observationFrame = (observation: ChannelObservation) => {
+  if (observation.type === "frame" || observation.type === "completed" || observation.type === "incomplete")
+    return Effect.succeed(observation.frame)
+  return Effect.fail(observation.error)
+}
+
+const terminal = (observation: ChannelObservation) => observation.type !== "frame"
+
+// This deliberately models only sequential test traffic. Core owns production connection pooling and recovery.
+const makeChannel = Effect.gen(function* () {
+  const constructor = yield* Socket.WebSocketConstructor
+  let connection: WebSocketConnection | undefined
+  let checkpoint: ChannelCheckpoint | undefined
+  let pending: ChannelCheckpoint | undefined
+  let opens = 0
+  const sent: unknown[] = []
+
+  const close = Effect.suspend(() => {
+    const current = connection
+    connection = undefined
+    return current ? current.close : Effect.void
+  })
+  yield* Effect.addFinalizer(() => close)
+
+  const executor: WebSocketChannelExecutor = {
+    execute: (exchange: WebSocketChannelExchange) =>
+      Effect.gen(function* () {
+        if (!connection) {
+          connection = yield* WebSocketTransport.open(exchange.connect).pipe(
+            Effect.provideService(Socket.WebSocketConstructor, constructor),
+          )
+          opens += 1
+        }
+        const current = connection
+        const create = yield* exchange.driver.create(checkpoint)
+        if (create.mode === "full") checkpoint = undefined
+        pending = undefined
+        sent.push(decodeJson(create.message))
+        yield* current.sendText(create.message)
+        const decoder = new TextDecoder()
+        return {
+          frames: current.messages.pipe(
+            Stream.map((message) => WebSocketTransport.messageText(message, decoder)),
+            Stream.mapEffect((frame) => exchange.driver.observe(create, frame)),
+            Stream.tap((observation) =>
+              Effect.sync(() => {
+                if (!terminal(observation)) return
+                pending = observation.type === "completed" ? observation.checkpoint : undefined
+                if (observation.type !== "completed") checkpoint = undefined
+              }),
+            ),
+            Stream.takeUntil(terminal),
+            Stream.mapEffect(observationFrame),
+          ),
+          complete: Effect.sync(() => {
+            checkpoint = pending
+            pending = undefined
+          }),
+        }
+      }),
+  }
+
+  return {
+    executor,
+    sent,
+    opens: () => opens,
+    reconnect: (preserveCheckpoint = false) =>
+      close.pipe(
+        Effect.andThen(
+          Effect.sync(() => {
+            pending = undefined
+            if (!preserveCheckpoint) checkpoint = undefined
+          }),
+        ),
+      ),
+  }
+})
+
+describe("OpenAI Responses WebSocket recorded", () => {
+  recorded.effect.with("continues a tool call over one socket", { tags: ["tool", "continuation"] }, () =>
+    Effect.gen(function* () {
+      const channel = yield* makeChannel
+      const request = LLM.request({
+        id: "recorded_openai_responses_websocket_tool",
+        model,
+        system: "Call get_weather once, then reply exactly: Paris is sunny.",
+        prompt: "What is the weather in Paris?",
+        tools: [weatherTool],
+        generation: { maxTokens: 50 },
+        cache: "none",
+      })
+      const first = yield* LLMClient.generate(request, { webSocket: channel.executor })
+      const call = first.toolCalls[0]
+      if (!call) yield* Effect.die("Expected get_weather tool call")
+      const result = yield* ToolRuntime.dispatch({ [weatherToolName]: weatherRuntimeTool }, call)
+      const second = yield* LLMClient.generate(
+        LLMRequest.update(request, {
+          messages: [
+            ...request.messages,
+            first.message,
+            Message.tool({ id: call.id, name: call.name, result: result.result }),
+          ],
+        }),
+        { webSocket: channel.executor },
+      )
+
+      expect(second.text).toBe("Paris is sunny.")
+      expect(channel.opens()).toBe(1)
+      expect(channel.sent).toHaveLength(2)
+      expect(channel.sent[1]).toMatchObject({
+        previous_response_id: expect.any(String),
+        input: [{ type: "function_call_output", call_id: call.id, output: expect.any(String) }],
+      })
+    }),
+  )
+
+  recorded.effect.with("reconstructs full context after reconnect", { tags: ["reconnect", "full-context"] }, () =>
+    Effect.gen(function* () {
+      const channel = yield* makeChannel
+      const request = LLM.request({
+        id: "recorded_openai_responses_websocket_reconnect",
+        model,
+        system: "Follow the user's exact reply instruction.",
+        prompt: "Reply exactly: Alpha.",
+        generation: { maxTokens: 30 },
+        cache: "none",
+      })
+      const first = yield* LLMClient.generate(request, { webSocket: channel.executor })
+      yield* channel.reconnect()
+      const second = yield* LLMClient.generate(
+        LLMRequest.update(request, {
+          messages: [...request.messages, first.message, Message.user("Reply exactly: Beta.")],
+        }),
+        { webSocket: channel.executor },
+      )
+
+      expect(first.text).toBe("Alpha.")
+      expect(second.text).toBe("Beta.")
+      expect(channel.opens()).toBe(2)
+      expect(channel.sent[1]).not.toHaveProperty("previous_response_id")
+      expect(channel.sent[1]).toMatchObject({
+        input: [
+          { role: "system", content: "Follow the user's exact reply instruction." },
+          { role: "user", content: [{ type: "input_text", text: "Reply exactly: Alpha." }] },
+          { role: "assistant", content: [{ type: "output_text", text: "Alpha." }] },
+          { role: "user", content: [{ type: "input_text", text: "Reply exactly: Beta." }] },
+        ],
+      })
+    }),
+  )
+
+  recorded.effect.with("recovers from explicit continuation rejection", { tags: ["continuation", "recovery"] }, () =>
+    Effect.gen(function* () {
+      const channel = yield* makeChannel
+      const request = LLM.request({
+        id: "recorded_openai_responses_websocket_rejection",
+        model,
+        system: "Follow the user's exact reply instruction.",
+        prompt: "Reply exactly: Ready.",
+        generation: { maxTokens: 30 },
+        cache: "none",
+      })
+      const first = yield* LLMClient.generate(request, { webSocket: channel.executor })
+      const continuation = LLMRequest.update(request, {
+        messages: [...request.messages, first.message, Message.user("Reply exactly: Recovered.")],
+      })
+      yield* channel.reconnect(true)
+      const rejected = yield* LLMClient.generate(continuation, { webSocket: channel.executor }).pipe(Effect.flip)
+      const recovered = yield* LLMClient.generate(continuation, { webSocket: channel.executor })
+
+      expect(rejected).toMatchObject({
+        reason: { _tag: "Transport", delivery: "rejected", recovery: "retry-full" },
+      })
+      expect(recovered.text).toBe("Recovered.")
+      expect(channel.opens()).toBe(2)
+      expect(channel.sent[1]).toHaveProperty("previous_response_id", expect.any(String))
+      expect(channel.sent[2]).not.toHaveProperty("previous_response_id")
+      expect(channel.sent[2]).toMatchObject({
+        input: [
+          { role: "system", content: "Follow the user's exact reply instruction." },
+          { role: "user", content: [{ type: "input_text", text: "Reply exactly: Ready." }] },
+          { role: "assistant", content: [{ type: "output_text", text: "Ready." }] },
+          { role: "user", content: [{ type: "input_text", text: "Reply exactly: Recovered." }] },
+        ],
+      })
+    }),
+  )
+})

packages/ai/test/provider/openai-responses.test.ts

@@ -28,7 +28,7 @@ import * as Azure from "../../src/providers/azure.js"
 import * as OpenAI from "../../src/providers/openai.js"
 import * as XAI from "../../src/providers/xai.js"
 import * as OpenAIResponses from "../../src/protocols/openai-responses.js"
-import { OpenAIResponsesChannel } from "../../src/protocols/openai-responses-channel.js"
+import { OpenResponsesContinuation } from "../../src/protocols/open-responses-continuation.js"
 import * as ProviderShared from "../../src/protocols/shared.js"
 import { continuationRequest, nativeOpenAIResponsesContinuation } from "../continuation-scenarios.js"
 import { it } from "../lib/effect.js"
@@ -68,7 +68,7 @@ const baseChannelDriver = (message: string): WebSocketChannelDriver => ({
 
 const continuationDriver = (request: Readonly<Record<string, unknown>>) => {
   const message = ProviderShared.encodeJson(request)
-  return OpenAIResponsesChannel.driver({
+  return OpenResponsesContinuation.driver({
     id: "openai-responses",
     name: "OpenAI Responses",
     request,
@@ -263,7 +263,7 @@ describe("OpenAI Responses route", () => {
       expect(prepared.body.input).toEqual([
         { role: "user", content: [{ type: "input_text", text: "Before." }] },
         { role: "developer", content: "Operator update." },
-        { role: "assistant", content: [{ type: "output_text", text: "After." }] },
+        { type: "message", role: "assistant", content: [{ type: "output_text", text: "After." }] },
       ])
     }),
   )
@@ -485,7 +485,7 @@ describe("OpenAI Responses route", () => {
     }),
   )
 
-  it.effect("continues store-false reasoning without replaying the output-only item ID", () =>
+  it.effect("continues store-false reasoning while retaining the output item ID", () =>
     Effect.gen(function* () {
       const firstInput = [{ role: "user", content: [{ type: "input_text", text: "Think" }] }]
       const request = { type: "response.create", model: "gpt-5.2", store: false, input: firstInput }
@@ -515,6 +515,7 @@ describe("OpenAI Responses route", () => {
           ...firstInput,
           {
             type: "reasoning",
+            id: "rs_1",
             summary: [{ type: "summary_text", text: "Thought" }],
             encrypted_content: "encrypted",
           },
@@ -690,6 +691,134 @@ describe("OpenAI Responses route", () => {
     }),
   )
 
+  it.effect("builds xAI WebSocket requests without OpenAI handshake headers", () =>
+    Effect.gen(function* () {
+      const deps = Layer.succeed(
+        RequestExecutor.Service,
+        RequestExecutor.Service.of({ execute: () => Effect.die("unexpected HTTP request") }),
+      )
+      const response = yield* LLMClient.generate(LLM.request({ model: xaiModel, prompt: "Say hello." }), {
+        webSocket: {
+          execute: (exchange) =>
+            Effect.gen(function* () {
+              expect(exchange.connect.url).toBe("wss://api.x.ai/v1/responses")
+              expect(exchange.connect.rotateAfterMs).toBe(24 * 60 * 1000)
+              expect(exchange.connect.headers.authorization).toBe("Bearer test")
+              expect(exchange.connect.headers["openai-beta"]).toBeUndefined()
+              expect(JSON.parse((yield* exchange.driver.create(undefined)).message)).toMatchObject({
+                type: "response.create",
+                model: "grok-4.5",
+                store: false,
+              })
+              return {
+                frames: Stream.make(
+                  JSON.stringify({ type: "response.created", response: { id: "resp_xai" } }),
+                  JSON.stringify({ type: "response.completed", response: { id: "resp_xai" } }),
+                ),
+                complete: Effect.void,
+              }
+            }),
+        },
+      }).pipe(Effect.provide(LLMClient.layer.pipe(Layer.provide(deps))))
+
+      expect(response.finishReason.normalized).toBe("stop")
+    }),
+  )
+
+  it.effect("builds Azure WebSocket requests with v1 URLs and bearer auth", () =>
+    Effect.gen(function* () {
+      const deps = Layer.succeed(
+        RequestExecutor.Service,
+        RequestExecutor.Service.of({ execute: () => Effect.die("unexpected HTTP request") }),
+      )
+      const cases = [
+        {
+          model: Azure.configure({ resourceName: "opencode-test", apiKey: "azure-key" }).responses("deployment"),
+          authorization: "Bearer azure-key",
+        },
+        {
+          model: Azure.configure({ resourceName: "opencode-test", auth: Auth.bearer("entra-token") }).responses(
+            "deployment",
+          ),
+          authorization: "Bearer entra-token",
+        },
+      ]
+
+      yield* Effect.forEach(cases, (item) =>
+        LLMClient.generate(LLM.request({ model: item.model, prompt: "Say hello." }), {
+          webSocket: {
+            execute: (exchange) =>
+              Effect.gen(function* () {
+                expect(exchange.connect.url).toBe("wss://opencode-test.openai.azure.com/openai/v1/responses")
+                expect(exchange.connect.rotateAfterMs).toBe(55 * 60 * 1000)
+                expect(exchange.connect.headers.authorization).toBe(item.authorization)
+                expect(exchange.connect.headers["api-key"]).toBeUndefined()
+                expect(exchange.connect.headers["openai-beta"]).toBeUndefined()
+                expect(JSON.parse((yield* exchange.driver.create(undefined)).message)).toMatchObject({
+                  type: "response.create",
+                  model: "deployment",
+                  store: false,
+                })
+                return {
+                  frames: Stream.make(
+                    JSON.stringify({ type: "response.created", response: { id: "resp_azure" } }),
+                    JSON.stringify({ type: "response.completed", response: { id: "resp_azure" } }),
+                  ),
+                  complete: Effect.void,
+                }
+              }),
+          },
+        }).pipe(Effect.provide(LLMClient.layer.pipe(Layer.provide(deps)))),
+      )
+    }),
+  )
+
+  it.effect("keeps unsupported Azure endpoints and API versions on HTTP", () =>
+    Effect.gen(function* () {
+      const cases = [
+        {
+          model: Azure.configure({
+            resourceName: "opencode-test",
+            apiKey: "azure-key",
+            apiVersion: "2025-04-01-preview",
+          }).responses("deployment"),
+          url: "https://opencode-test.openai.azure.com/openai/v1/responses?api-version=2025-04-01-preview",
+        },
+        {
+          model: Azure.configure({
+            resourceName: "opencode-test",
+            apiKey: "azure-key",
+            useDeploymentBasedUrls: true,
+          }).responses("deployment"),
+          url: "https://opencode-test.openai.azure.com/openai/deployments/deployment/responses?api-version=v1",
+        },
+        {
+          model: Azure.configure({ baseURL: "https://gateway.example/azure", apiKey: "azure-key" }).responses(
+            "deployment",
+          ),
+          url: "https://gateway.example/azure/responses",
+        },
+      ]
+
+      yield* Effect.forEach(cases, (item) =>
+        LLMClient.generate(LLM.request({ model: item.model, prompt: "Say hello." }), {
+          webSocket: { execute: () => Effect.die("unexpected WebSocket request") },
+        }).pipe(
+          Effect.provide(
+            dynamicResponse((input) =>
+              Effect.gen(function* () {
+                expect(input.request.url).toBe(item.url)
+                return input.respond(sseEvents({ type: "response.completed", response: {} }), {
+                  headers: { "content-type": "text/event-stream" },
+                })
+              }),
+            ),
+          ),
+        ),
+      )
+    }),
+  )
+
   it.effect("uses exactly one HTTP request when no WebSocket executor is supplied", () =>
     Effect.gen(function* () {
       const attempts = yield* Ref.make(0)
@@ -1144,7 +1273,7 @@ describe("OpenAI Responses route", () => {
         {
           type: "input_file",
           filename: "report.pdf",
-          file_data: "data:application/pdf;base64,JVBERi0xLjQ=",
+          file_data: "JVBERi0xLjQ=",
         },
       ])
     }),
@@ -1171,12 +1300,12 @@ describe("OpenAI Responses route", () => {
       )
 
       expect(expectToolOutput(prepared.body).output).toEqual([
-        { type: "input_file", filename: "report.pdf", file_data: dataUrl },
+        { type: "input_file", filename: "report.pdf", file_data: base64 },
       ])
     }),
   )
 
-  it.effect("uses xAI inline file encoding for PDF tool results", () =>
+  it.effect("uses standard inline file encoding for xAI PDF tool results", () =>
     Effect.gen(function* () {
       const prepared = yield* compileRequest(
         LLM.request({
@@ -1205,7 +1334,6 @@ describe("OpenAI Responses route", () => {
           type: "input_file",
           filename: "report.pdf",
           file_data: "JVBERi0xLjQ=",
-          mime_type: "application/pdf",
         },
       ])
     }),
@@ -1230,7 +1358,61 @@ describe("OpenAI Responses route", () => {
       )
 
       expect(expectToolOutput(prepared.body).output).toEqual([
-        { type: "input_file", filename: "file", file_data: "data:audio/mpeg;base64,AAECAw==" },
+        { type: "input_file", filename: "file", file_data: "AAECAw==" },
+      ])
+    }),
+  )
+
+  it.effect("lowers remote tool-result media URLs without base64 wrapping", () =>
+    Effect.gen(function* () {
+      const prepared = yield* compileRequest(
+        LLM.request({
+          model,
+          messages: [
+            Message.assistant([ToolCallPart.make({ id: "call_1", name: "fetch", input: {} })]),
+            Message.tool({
+              id: "call_1",
+              name: "fetch",
+              resultType: "content",
+              result: [
+                { type: "file", uri: "https://example.com/image.png", mime: "image/png" },
+                { type: "file", uri: "https://example.com/report.pdf", mime: "application/pdf", name: "report.pdf" },
+              ],
+            }),
+          ],
+        }),
+      )
+
+      expect(expectToolOutput(prepared.body).output).toEqual([
+        { type: "input_image", image_url: "https://example.com/image.png" },
+        { type: "input_file", filename: "report.pdf", file_url: "https://example.com/report.pdf" },
+      ])
+    }),
+  )
+
+  it.effect("lowers tool-result videos as input_video", () =>
+    Effect.gen(function* () {
+      const prepared = yield* compileRequest(
+        LLM.request({
+          model,
+          messages: [
+            Message.assistant([ToolCallPart.make({ id: "call_1", name: "record", input: {} })]),
+            Message.tool({
+              id: "call_1",
+              name: "record",
+              resultType: "content",
+              result: [
+                { type: "file", uri: "data:video/mp4;base64,AAECAw==", mime: "video/mp4" },
+                { type: "file", uri: "https://example.com/demo.mp4", mime: "video/mp4" },
+              ],
+            }),
+          ],
+        }),
+      )
+
+      expect(expectToolOutput(prepared.body).output).toEqual([
+        { type: "input_video", video_url: "data:video/mp4;base64,AAECAw==" },
+        { type: "input_video", video_url: "https://example.com/demo.mp4" },
       ])
     }),
   )
@@ -1284,6 +1466,7 @@ describe("OpenAI Responses route", () => {
           model: OpenAI.configure({ baseURL: "https://api.openai.test/v1/", apiKey: "test" }).model("gpt-5.2"),
           prompt: "think",
           promptCacheKey: "session_123",
+          generation: { presencePenalty: 0.25, frequencyPenalty: -0.25 },
           tools: [
             ToolDefinition.make({ name: "read", description: "Read a file", inputSchema: { type: "object" } }),
             ToolDefinition.make({ name: "grep", description: "Search files", inputSchema: { type: "object" } }),
@@ -1293,6 +1476,10 @@ describe("OpenAI Responses route", () => {
             reasoningEffort: "high",
             reasoningSummary: "auto",
             include: ["reasoning.encrypted_content"],
+            metadata: { environment: "test", tenant: "acme" },
+            safetyIdentifier: "user_123",
+            streamOptions: { includeObfuscation: false },
+            topLogprobs: 5,
             truncation: "disabled",
             allowedTools: { toolNames: ["read", "grep"], mode: "required" },
             maxToolCalls: 4,
@@ -1306,6 +1493,12 @@ describe("OpenAI Responses route", () => {
       expect(prepared.body.include).toEqual(["reasoning.encrypted_content"])
       expect(prepared.body.reasoning).toEqual({ effort: "high", summary: "auto" })
       expect(prepared.body.text).toEqual({ verbosity: "low" })
+      expect(prepared.body.metadata).toEqual({ environment: "test", tenant: "acme" })
+      expect(prepared.body.safety_identifier).toBe("user_123")
+      expect(prepared.body.stream_options).toEqual({ include_obfuscation: false })
+      expect(prepared.body.top_logprobs).toBe(5)
+      expect(prepared.body.presence_penalty).toBe(0.25)
+      expect(prepared.body.frequency_penalty).toBe(-0.25)
       expect(prepared.body.truncation).toBe("disabled")
       expect(prepared.body.tool_choice).toEqual({
         type: "allowed_tools",
@@ -1473,7 +1666,7 @@ describe("OpenAI Responses route", () => {
       expect(response.text).toBe("Hello!")
       expect(response.events).toEqual([
         { type: "step-start", index: 0 },
-        { type: "text-start", id: "msg_1" },
+        { type: "text-start", id: "msg_1", providerMetadata: { openai: { itemId: "msg_1" } } },
         { type: "text-delta", id: "msg_1", text: "Hello" },
         { type: "text-delta", id: "msg_1", text: "!" },
         { type: "text-end", id: "msg_1" },
@@ -1494,6 +1687,108 @@ describe("OpenAI Responses route", () => {
     }),
   )
 
+  it.effect("preserves standard refusal content as ordinary assistant text", () =>
+    Effect.gen(function* () {
+      const response = yield* LLMClient.generate(request).pipe(
+        Effect.provide(
+          fixedResponse(
+            sseEvents(
+              {
+                type: "response.output_item.added",
+                output_index: 0,
+                item: { type: "message", id: "msg_refusal", content: [] },
+              },
+              {
+                type: "response.content_part.added",
+                item_id: "msg_refusal",
+                output_index: 0,
+                content_index: 0,
+                part: { type: "refusal", refusal: "" },
+              },
+              {
+                type: "response.refusal.delta",
+                item_id: "msg_refusal",
+                output_index: 0,
+                content_index: 0,
+                delta: "I can't",
+              },
+              {
+                type: "response.refusal.delta",
+                item_id: "msg_refusal",
+                output_index: 0,
+                content_index: 0,
+                delta: " help with that.",
+              },
+              {
+                type: "response.refusal.done",
+                item_id: "msg_refusal",
+                output_index: 0,
+                content_index: 0,
+                refusal: "I can't help with that.",
+              },
+              {
+                type: "response.content_part.done",
+                item_id: "msg_refusal",
+                output_index: 0,
+                content_index: 0,
+                part: { type: "refusal", refusal: "I can't help with that." },
+              },
+              {
+                type: "response.output_item.done",
+                output_index: 0,
+                item: {
+                  type: "message",
+                  id: "msg_refusal",
+                  phase: "final_answer",
+                  content: [{ type: "refusal", refusal: "I can't help with that." }],
+                },
+              },
+              { type: "response.completed", response: { id: "resp_1" } },
+            ),
+          ),
+        ),
+      )
+
+      expect(response.text).toBe("I can't help with that.")
+      expect(response.finishReason).toEqual({ normalized: "stop", raw: undefined })
+      expect(response.message.content).toEqual([
+        {
+          type: "text",
+          text: "I can't help with that.",
+          providerMetadata: { openai: { itemId: "msg_refusal", phase: "final_answer" } },
+        },
+      ])
+
+      const prepared = yield* compileRequest(LLM.request({ model, messages: [response.message] }))
+      expect(prepared.body.input).toEqual([
+        {
+          type: "message",
+          id: "msg_refusal",
+          role: "assistant",
+          content: [{ type: "output_text", text: "I can't help with that." }],
+          phase: "final_answer",
+        },
+      ])
+    }),
+  )
+
+  it.effect("rejects malformed refusal events", () =>
+    Effect.gen(function* () {
+      const events = [
+        { type: "response.refusal.delta", output_index: 0, content_index: 0, delta: "missing item" },
+        { type: "response.refusal.delta", item_id: "msg_1", output_index: 0, content_index: 0 },
+        { type: "response.refusal.done", item_id: "msg_1", output_index: 0, content_index: 0 },
+      ]
+      for (const event of events) {
+        const error = yield* LLMClient.generate(request).pipe(
+          Effect.provide(fixedResponse(sseEvents(event))),
+          Effect.flip,
+        )
+        expect(error.reason._tag).toBe("InvalidProviderOutput")
+      }
+    }),
+  )
+
   it.effect("preserves and replays assistant message phases", () =>
     Effect.gen(function* () {
       const response = yield* LLMClient.generate(request).pipe(
@@ -1532,33 +1827,39 @@ describe("OpenAI Responses route", () => {
         {
           type: "text",
           text: "Checking.",
-          providerMetadata: { openai: { phase: "commentary" } },
+          providerMetadata: { openai: { itemId: "msg_commentary", phase: "commentary" } },
         },
         {
           type: "text",
           text: "Finished.",
-          providerMetadata: { openai: { phase: "final_answer" } },
+          providerMetadata: { openai: { itemId: "msg_final", phase: "final_answer" } },
         },
         {
           type: "text",
           text: "Unclassified.",
-          providerMetadata: { openai: { phase: null } },
+          providerMetadata: { openai: { itemId: "msg_null", phase: null } },
         },
       ])
 
       const prepared = yield* compileRequest(LLM.request({ model, messages: [response.message] }))
       expect(prepared.body.input).toEqual([
         {
+          type: "message",
+          id: "msg_commentary",
           role: "assistant",
           content: [{ type: "output_text", text: "Checking." }],
           phase: "commentary",
         },
         {
+          type: "message",
+          id: "msg_final",
           role: "assistant",
           content: [{ type: "output_text", text: "Finished." }],
           phase: "final_answer",
         },
         {
+          type: "message",
+          id: "msg_null",
           role: "assistant",
           content: [{ type: "output_text", text: "Unclassified." }],
           phase: null,
@@ -1652,12 +1953,12 @@ describe("OpenAI Responses route", () => {
       )
 
       expect(response.events.filter((event) => event.type.startsWith("text-"))).toEqual([
-        { type: "text-start", id: "msg_1" },
+        { type: "text-start", id: "msg_1", providerMetadata: { openai: { itemId: "msg_1" } } },
         { type: "text-delta", id: "msg_1", text: "First" },
-        { type: "text-end", id: "msg_1" },
-        { type: "text-start", id: "msg_2" },
+        { type: "text-end", id: "msg_1", providerMetadata: undefined },
+        { type: "text-start", id: "msg_2", providerMetadata: { openai: { itemId: "msg_2" } } },
         { type: "text-delta", id: "msg_2", text: "Second" },
-        { type: "text-end", id: "msg_2" },
+        { type: "text-end", id: "msg_2", providerMetadata: { openai: { itemId: "msg_2" } } },
       ])
     }),
   )
@@ -1689,7 +1990,7 @@ describe("OpenAI Responses route", () => {
       expect(response.events.filter((event) => event.type === "finish")).toHaveLength(1)
       expect(response.message.content).toEqual([
         { type: "reasoning", text: "thinking" },
-        { type: "text", text: "Hello" },
+        { type: "text", text: "Hello", providerMetadata: { openai: { itemId: "msg_1" } } },
       ])
     }),
   )
@@ -1850,6 +2151,7 @@ describe("OpenAI Responses route", () => {
                   { role: "user", content: [{ type: "input_text", text: "What changed?" }] },
                   {
                     type: "reasoning",
+                    id: "rs_1",
                     encrypted_content: "encrypted-state",
                     summary: [{ type: "summary_text", text: "Checked the previous diff." }],
                   },
@@ -1857,7 +2159,6 @@ describe("OpenAI Responses route", () => {
                   { role: "user", content: [{ type: "input_text", text: "Summarize it." }] },
                 ],
               })
-              expect(body.input[1]).not.toHaveProperty("id")
               return input.respond(
                 sseEvents(
                   { type: "response.output_text.delta", item_id: "msg_1", delta: "Parser now round-trips reasoning." },
@@ -1901,13 +2202,14 @@ describe("OpenAI Responses route", () => {
       )
 
       expect(prepared.body.input).toEqual([
-        { role: "assistant", content: [{ type: "output_text", text: "Before." }] },
+        { type: "message", role: "assistant", content: [{ type: "output_text", text: "Before." }] },
         {
           type: "reasoning",
+          id: "rs_1",
           encrypted_content: "encrypted-state",
           summary: [{ type: "summary_text", text: "Checked order." }],
         },
-        { role: "assistant", content: [{ type: "output_text", text: "After." }] },
+        { type: "message", role: "assistant", content: [{ type: "output_text", text: "After." }] },
       ])
     }),
   )
@@ -2039,6 +2341,7 @@ describe("OpenAI Responses route", () => {
       expect(prepared.body.input).toEqual([
         {
           type: "reasoning",
+          id: "rs_1",
           encrypted_content: "encrypted-state",
           summary: [
             { type: "summary_text", text: "First" },
@@ -2171,6 +2474,50 @@ describe("OpenAI Responses route", () => {
           usage,
         },
       ])
+
+      const prepared = yield* compileRequest(LLM.request({ model, messages: [response.message] }))
+      expect(prepared.body.input).toEqual([
+        {
+          type: "function_call",
+          id: "item_1",
+          call_id: "call_1",
+          name: "lookup",
+          arguments: '{"query":"weather"}',
+        },
+      ])
+    }),
+  )
+
+  it.effect("finalizes a pending function call at response completion", () =>
+    Effect.gen(function* () {
+      const body = sseEvents(
+        {
+          type: "response.output_item.added",
+          item: { type: "function_call", id: "item_1", call_id: "call_1", name: "lookup", arguments: "" },
+        },
+        { type: "response.completed", response: { usage: { input_tokens: 5, output_tokens: 1 } } },
+      )
+      const response = yield* LLMClient.generate(request).pipe(Effect.provide(fixedResponse(body)))
+
+      expect(response.events.filter((event) => LLMEvent.is.toolInputEnd(event) || LLMEvent.is.toolCall(event))).toEqual(
+        [
+          {
+            type: "tool-input-end",
+            id: "call_1",
+            name: "lookup",
+            providerMetadata: { openai: { itemId: "item_1" } },
+          },
+          {
+            type: "tool-call",
+            id: "call_1",
+            name: "lookup",
+            input: {},
+            providerExecuted: undefined,
+            providerMetadata: { openai: { itemId: "item_1" } },
+          },
+        ],
+      )
+      expect(response.finishReason.normalized).toBe("tool-calls")
     }),
   )
 
@@ -2237,6 +2584,35 @@ describe("OpenAI Responses route", () => {
     }),
   )
 
+  it.effect("retains function call item metadata when output_item.added is absent", () =>
+    Effect.gen(function* () {
+      const response = yield* LLMClient.generate(request).pipe(
+        Effect.provide(
+          fixedResponse(
+            sseEvents(
+              {
+                type: "response.output_item.done",
+                item: {
+                  type: "function_call",
+                  id: "item_1",
+                  call_id: "call_1",
+                  name: "lookup",
+                  arguments: '{"query":"weather"}',
+                },
+              },
+              { type: "response.completed", response: { id: "resp_1" } },
+            ),
+          ),
+        ),
+      )
+
+      expect(response.events.find(LLMEvent.is.toolCall)).toMatchObject({
+        id: "call_1",
+        providerMetadata: { openai: { itemId: "item_1" } },
+      })
+    }),
+  )
+
   it.effect("decodes web_search_call as provider-executed tool-call + tool-result", () =>
     Effect.gen(function* () {
       const item = {
@@ -2389,7 +2765,7 @@ describe("OpenAI Responses route", () => {
             {
               type: "input_file",
               filename: "report.pdf",
-              file_data: "data:application/pdf;base64,JVBERi0xLjQ=",
+              file_data: "JVBERi0xLjQ=",
             },
           ],
         },
@@ -2397,7 +2773,7 @@ describe("OpenAI Responses route", () => {
     }),
   )
 
-  it.effect("uses xAI inline file encoding for user PDFs", () =>
+  it.effect("uses standard inline file encoding for xAI user PDFs", () =>
     Effect.gen(function* () {
       const prepared = yield* compileRequest(
         LLM.request({
@@ -2421,7 +2797,6 @@ describe("OpenAI Responses route", () => {
               type: "input_file",
               filename: "report.pdf",
               file_data: "JVBERi0xLjQ=",
-              mime_type: "application/pdf",
             },
           ],
         },
@@ -2446,7 +2821,7 @@ describe("OpenAI Responses route", () => {
             {
               type: "input_file",
               filename: "file",
-              file_data: "data:application/x-tar;base64,AAECAw==",
+              file_data: "AAECAw==",
             },
           ],
         },
@@ -2454,6 +2829,37 @@ describe("OpenAI Responses route", () => {
     }),
   )
 
+  it.effect("lowers remote user media URLs without base64 wrapping", () =>
+    Effect.gen(function* () {
+      const prepared = yield* compileRequest(
+        LLM.request({
+          model,
+          messages: [
+            Message.user([
+              { type: "media", mediaType: "image/png", data: "https://example.com/image.png" },
+              {
+                type: "media",
+                mediaType: "application/pdf",
+                data: "https://example.com/report.pdf",
+                filename: "report.pdf",
+              },
+            ]),
+          ],
+        }),
+      )
+
+      expect(prepared.body.input).toEqual([
+        {
+          role: "user",
+          content: [
+            { type: "input_image", image_url: "https://example.com/image.png" },
+            { type: "input_file", filename: "report.pdf", file_url: "https://example.com/report.pdf" },
+          ],
+        },
+      ])
+    }),
+  )
+
   it.effect("fails with a typed rate limit for provider error frames", () =>
     Effect.gen(function* () {
       const error = yield* LLMClient.generate(request).pipe(
@@ -2610,36 +3016,42 @@ describe("OpenAI Responses route", () => {
     }),
   )
 
-  it.effect("falls back to a stable default when error is null", () =>
+  it.effect("falls back to the raw payload when error is null", () =>
     Effect.gen(function* () {
       const error = yield* LLMClient.generate(request).pipe(
         Effect.provide(fixedResponse(sseEvents({ type: "error", error: null }))),
         Effect.flip,
       )
 
-      expect(error.reason).toMatchObject({ _tag: "UnknownProvider", message: "OpenAI Responses stream error" })
+      expect(error.reason).toMatchObject({ _tag: "UnknownProvider" })
+      expect(error.reason.message).toContain('"error":null')
+      expect(error.body).toBe(error.reason.message)
     }),
   )
 
-  it.effect("falls back to a stable default when both error and response are absent", () =>
+  it.effect("falls back to the raw payload when both error and response are absent", () =>
     Effect.gen(function* () {
       const error = yield* LLMClient.generate(request).pipe(
         Effect.provide(fixedResponse(sseEvents({ type: "error" }))),
         Effect.flip,
       )
 
-      expect(error.reason).toMatchObject({ _tag: "UnknownProvider", message: "OpenAI Responses stream error" })
+      expect(error.reason).toMatchObject({ _tag: "UnknownProvider" })
+      expect(error.reason.message).toContain('"type":"error"')
+      expect(error.body).toBe(error.reason.message)
     }),
   )
 
-  it.effect("falls back to a stable default when response.failed has no error payload", () =>
+  it.effect("keeps the raw response payload when response.failed has no error payload", () =>
     Effect.gen(function* () {
       const error = yield* LLMClient.generate(request).pipe(
         Effect.provide(fixedResponse(sseEvents({ type: "response.failed", response: { id: "resp_failed_3" } }))),
         Effect.flip,
       )
 
-      expect(error.reason).toMatchObject({ _tag: "UnknownProvider", message: "OpenAI Responses response failed" })
+      expect(error.reason).toMatchObject({ _tag: "UnknownProvider" })
+      expect(error.reason.message).toContain('"resp_failed_3"')
+      expect(error.body).toBe(error.reason.message)
     }),
   )
 

packages/ai/test/provider/pdf.recorded.test.ts

@@ -64,7 +64,7 @@ const targets: ReadonlyArray<{
     id: "xai",
     name: "xAI Grok 4.5",
     provider: "xai",
-    protocol: "openai-responses",
+    protocol: "xai-responses",
     requires: "XAI_API_KEY",
     filename: "verification.pdf",
     maxTokens: 40,

packages/ai/test/provider/xai-responses.test.ts

@@ -0,0 +1,77 @@
+import { describe, expect } from "bun:test"
+import { Effect } from "effect"
+import { LLM, LLMEvent } from "../../src/index.js"
+import { XAI } from "../../src/providers.js"
+import { OpenResponses } from "../../src/protocols/open-responses.js"
+import { OpenAIResponses } from "../../src/protocols/openai-responses.js"
+import { XAIResponses } from "../../src/protocols/xai-responses.js"
+import { LLMClient } from "../../src/route.js"
+import { compileRequest } from "../../src/route/client.js"
+import { it } from "../lib/effect.js"
+import { fixedResponse } from "../lib/http.js"
+import { sseEvents } from "../lib/sse.js"
+
+const model = XAI.configure({ apiKey: "test", baseURL: "https://api.x.ai/v1" }).responses("grok-4.6")
+
+describe("xAI Responses route", () => {
+  it.effect("extends the Open Responses baseline directly", () =>
+    Effect.gen(function* () {
+      expect(XAIResponses.protocol.body).toBe(OpenResponses.protocol.body)
+      expect(XAIResponses.protocol.body).not.toBe(OpenAIResponses.protocol.body)
+
+      const prepared = yield* compileRequest(LLM.request({ model, prompt: "Hello" }))
+      expect(prepared.protocol).toBe("xai-responses")
+    }),
+  )
+
+  it.effect("parses xAI reasoning text events", () =>
+    Effect.gen(function* () {
+      const response = yield* LLMClient.generate(LLM.request({ model, prompt: "Think" })).pipe(
+        Effect.provide(
+          fixedResponse(
+            sseEvents(
+              { type: "response.reasoning_text.delta", item_id: "reasoning_1", delta: "Considering." },
+              { type: "response.reasoning_text.done", item_id: "reasoning_1" },
+              {
+                type: "response.output_item.done",
+                item: { type: "reasoning", id: "reasoning_1", encrypted_content: "opaque" },
+              },
+              { type: "response.completed", response: { id: "response_1" } },
+            ),
+          ),
+        ),
+      )
+
+      expect(response.message.content.find((part) => part.type === "reasoning")).toMatchObject({
+        type: "reasoning",
+        text: "Considering.",
+        providerMetadata: { xai: { itemId: "reasoning_1", reasoningEncryptedContent: "opaque" } },
+      })
+    }),
+  )
+
+  it.effect("parses xAI hosted tool items", () =>
+    Effect.gen(function* () {
+      const response = yield* LLMClient.generate(LLM.request({ model, prompt: "Search X" })).pipe(
+        Effect.provide(
+          fixedResponse(
+            sseEvents(
+              {
+                type: "response.output_item.done",
+                item: { type: "x_search_call", id: "x_search_1", status: "completed", action: { query: "news" } },
+              },
+              { type: "response.completed", response: { id: "response_1" } },
+            ),
+          ),
+        ),
+      )
+
+      expect(response.events.find(LLMEvent.is.toolCall)).toMatchObject({
+        id: "x_search_1",
+        name: "x_search",
+        input: { query: "news" },
+        providerExecuted: true,
+      })
+    }),
+  )
+})

packages/ai/test/recorded-test.ts

@@ -1,5 +1,7 @@
 import { HttpRecorder } from "@opencode-ai/http-recorder"
+import { NodeSocket } from "@effect/platform-node"
 import { Layer } from "effect"
+import { Socket } from "effect/unstable/socket"
 import * as path from "node:path"
 import { fileURLToPath } from "node:url"
 import { LLMClient, RequestExecutor } from "../src/route.js"
@@ -16,7 +18,7 @@ import {
 const __dirname = path.dirname(fileURLToPath(import.meta.url))
 const FIXTURES_DIR = path.resolve(__dirname, "fixtures", "recordings")
 
-type RecordedEnv = RequestExecutorService | LLMClientService | ImageClientService
+type RecordedEnv = RequestExecutorService | LLMClientService | ImageClientService | Socket.WebSocketConstructor
 
 type RecordedTestsOptions = RecordedGroupOptions & {
   readonly options?: HttpRecorder.RecorderOptions
@@ -69,7 +71,7 @@ export const recordedTests = (options: RecordedTestsOptions) =>
         ...metadata,
       }
       if (recording) {
-        if (process.env.CI !== undefined) throw new Error("Unset CI before recording HTTP cassettes")
+        if (process.env.CI !== undefined) throw new Error("Unset CI before recording cassettes")
         HttpRecorder.removeCassetteSync(cassette, { directory: FIXTURES_DIR })
       }
       const requestExecutor = RequestExecutor.layer.pipe(
@@ -81,10 +83,16 @@ export const recordedTests = (options: RecordedTestsOptions) =>
           }),
         ),
       )
+      const webSocket = HttpRecorder.layerWebSocketConstructor(cassette, {
+        ...recorderOptions,
+        directory: FIXTURES_DIR,
+        metadata: recorderMetadata,
+      }).pipe(Layer.provide(NodeSocket.layerWebSocketConstructorWS))
       return Layer.mergeAll(
         requestExecutor,
         LLMClient.layer.pipe(Layer.provide(requestExecutor)),
         ImageClient.layer.pipe(Layer.provide(requestExecutor)),
+        webSocket,
       )
     },
   })

packages/app/AGENTS.md

@@ -19,6 +19,12 @@
 
 - Always prefer `createStore` over multiple `createSignal` calls
 
+## Typography
+
+- Use `--line-height-compact` (`16px`) for `13px` compact UI text and `--line-height-base` (`20px`) for body text.
+- Do not use `leading-none`, `line-height: 1`, or a `13px` line height for normal text. Inter descenders clip inside truncation and overflow containers.
+- Keep control and row heights explicit. Fix font metrics directly rather than using transforms, negative margins, or clip-padding compensation.
+
 ## Localization
 
 - NEVER hardcode user-visible English strings in production code. ALWAYS use an i18n key for visible copy, placeholders, accessible labels, tooltips, menus, dialogs, toasts, empty states, and displayed errors.

packages/app/e2e/performance/README.md

@@ -2,12 +2,34 @@
 
 The app's high-volume performance diagnostics live under `packages/app/e2e/performance` and are excluded from normal local and CI Playwright discovery. The benchmark config builds the app and serves the production bundle before running scenarios serially.
 
+The `devex` category is the explicit exception to the production-build rule. It measures development commands from submission through a user-visible ready state and has its own Playwright configuration.
+
 Run the suite explicitly from `packages/app`:
 
 ```sh
 bun run test:bench
 ```
 
+Run the desktop development startup benchmark from the repository root:
+
+```sh
+bun run bench:devex
+```
+
+It runs five serial samples of the exact `bun dev:desktop` command. Each sample uses a fresh desktop profile, database, service configuration, service registration, and service process; the desktop selects an isolated ephemeral loopback endpoint. It removes desktop build output and the desktop Vite cache before every run; dependencies, Bun's package cache, and Electron remain installed. The harness stops only that sample's service; it does not stop or change the elected global OpenCode service. The measured endpoint is a visible Home page whose empty-state controls pass Playwright actionability checks. The command's Electron installation check remains inside the measured interval.
+
+Set `DESKTOP_STARTUP_RUNS` only for focused diagnostics:
+
+```sh
+DESKTOP_STARTUP_RUNS=1 bun run bench:devex
+```
+
+Set `OPENCODE_PERFORMANCE_TRACE_DIR` to capture the renderer's CDP trace from attachment through actionable Home:
+
+```sh
+DESKTOP_STARTUP_RUNS=1 OPENCODE_PERFORMANCE_TRACE_DIR=/tmp/opencode-desktop-traces bun run bench:devex
+```
+
 PowerShell:
 
 ```powershell

packages/app/e2e/performance/chrome-trace.ts

@@ -18,9 +18,9 @@ const categories = [
   "disabled-by-default-v8.cpu_profiler",
 ]
 
-export async function startChromeTrace(page: Page, name: string) {
+export async function startChromeTrace(page: Page, name: string): Promise<undefined | (() => Promise<string>)> {
   const directory = process.env.OPENCODE_PERFORMANCE_TRACE_DIR
-  if (!directory) return
+  if (!directory) return undefined
 
   const selectors = process.env.OPENCODE_PERFORMANCE_SELECTOR_TRACE === "1"
   const file = await prepareChromeTrace(directory, name, selectors)

packages/app/e2e/performance/devex/desktop-startup-benchmark.spec.ts

@@ -0,0 +1,26 @@
+import { benchmark } from "../benchmark"
+import {
+  desktopBenchmarkContext,
+  runDesktopStartup,
+  summarizeDesktopStartup,
+  type DesktopStartupSample,
+} from "./desktop-startup"
+
+benchmark.describe("devex: desktop startup", () => {
+  benchmark("opens a cold desktop on Home", async ({ report }, testInfo) => {
+    benchmark.setTimeout(15 * 60_000)
+    const runs = Number(process.env.DESKTOP_STARTUP_RUNS ?? 5)
+    if (!Number.isSafeInteger(runs) || runs < 1) throw new Error("DESKTOP_STARTUP_RUNS must be a positive integer")
+
+    const samples: DesktopStartupSample[] = []
+    const context = await desktopBenchmarkContext(runs)
+    for (let run = 1; run <= runs; run++) {
+      const sample = await runDesktopStartup(run, testInfo).catch((error) => {
+        report(samples.length ? { samples, summary: summarizeDesktopStartup(samples) } : { samples }, context)
+        throw error
+      })
+      samples.push(sample)
+    }
+    report({ samples, summary: summarizeDesktopStartup(samples) }, context)
+  })
+})

packages/app/e2e/performance/devex/desktop-startup.ts

@@ -0,0 +1,526 @@
+import { Service } from "@opencode-ai/client/service"
+import { chromium, expect, type Browser, type Page, type TestInfo } from "@playwright/test"
+import { spawn, spawnSync, type ChildProcess } from "node:child_process"
+import { mkdir, mkdtemp, readFile, rm, writeFile } from "node:fs/promises"
+import { tmpdir } from "node:os"
+import { join, resolve } from "node:path"
+import { startChromeTrace } from "../chrome-trace"
+
+const repository = resolve(import.meta.dirname, "../../../../..")
+const milestones = [
+  "bunRootScript",
+  "bunDesktopScript",
+  "desktopPrepared",
+  "mainBundleReady",
+  "preloadBundleReady",
+  "rendererDevServerReady",
+  "electronSpawnStarted",
+  "debugEndpointReady",
+  "electronStarted",
+  "serviceEnsureStarted",
+  "serviceSpawnRequested",
+  "serviceReady",
+  "backgroundLoadingReady",
+  "rendererViteConnected",
+  "rendererInitializationStarted",
+  "rendererInitializationReady",
+  "windowVisible",
+  "homeReady",
+] as const
+const phases = [
+  "desktopPreparation",
+  "viteMainBundle",
+  "vitePreloadBundle",
+  "rendererServerStartup",
+  "electronStartup",
+  "serviceSpawnWait",
+  "serviceProcessStartup",
+  "rendererStartup",
+  "visibleWindowToHome",
+] as const
+
+type Milestone = (typeof milestones)[number]
+type Phase = (typeof phases)[number]
+type ServiceInfo = { id: string; version: string; url: string; pid: number }
+
+export type DesktopStartupSample = {
+  run: number
+  commandToHomeReadyMs: number
+  milestonesMs: Record<Milestone, number>
+  phasesMs: Record<Phase, number>
+  service: Omit<ServiceInfo, "id">
+}
+
+export async function runDesktopStartup(run: number, testInfo: TestInfo) {
+  const profile = await createColdProfile()
+  const desktop = await Promise.resolve()
+    .then(() => startDesktop(profile))
+    .catch(async (error) => {
+      await rm(profile.root, { recursive: true, force: true, maxRetries: 5, retryDelay: 100 })
+      throw error
+    })
+  try {
+    const page = await desktop.open()
+    const stopTrace = await startChromeTrace(page, `desktop-startup-${run}`)
+    try {
+      await startThemeObservation(page)
+      await waitForHome(page, desktop.mark)
+      await requireStableTheme(page)
+      return await desktop.result(run)
+    } finally {
+      await stopTrace?.()
+    }
+  } finally {
+    await desktop.close(testInfo, run)
+  }
+}
+
+export async function desktopBenchmarkContext(runs: number) {
+  const pkg = JSON.parse(await readFile(join(repository, "packages/desktop/package.json"), "utf8"))
+  const revision = spawnSync("git", ["rev-parse", "HEAD"], { cwd: repository })
+  if (revision.status !== 0) throw new Error("Failed to read the benchmark Git revision")
+  const status = spawnSync("git", ["status", "--porcelain"], { cwd: repository })
+  if (status.status !== 0) throw new Error("Failed to read the benchmark Git status")
+  const bun = spawnSync("bun", ["--version"], { cwd: repository })
+  if (bun.status !== 0) throw new Error("Failed to read the benchmark Bun version")
+  return {
+    arch: process.arch,
+    command: "bun dev:desktop",
+    runs,
+    profile: "fresh",
+    service: "isolated-cold",
+    install: "complete",
+    viteCache: "cold",
+    electronInstall: "present",
+    bunVersion: bun.stdout.toString().trim(),
+    electronVersion: pkg.devDependencies.electron,
+    electronViteVersionRange: pkg.devDependencies["electron-vite"],
+    gitCommit: revision.stdout.toString().trim(),
+    gitDirty: status.stdout.length > 0,
+    trace: Boolean(process.env.OPENCODE_PERFORMANCE_TRACE_DIR),
+  }
+}
+
+export function summarizeDesktopStartup(samples: DesktopStartupSample[]) {
+  return {
+    commandToHomeReadyMs: statistics(samples.map((sample) => sample.commandToHomeReadyMs)),
+    milestonesMs: Object.fromEntries(
+      milestones.map((name) => [name, statistics(samples.map((sample) => sample.milestonesMs[name]))]),
+    ),
+    phasesMs: Object.fromEntries(
+      phases.map((name) => [name, statistics(samples.map((sample) => sample.phasesMs[name]))]),
+    ),
+  }
+}
+
+export function milestoneForLine(line: string): Milestone | undefined {
+  const text = stripAnsi(line)
+  return milestonePatterns.find((item) => text.includes(item.text))?.name
+}
+
+const milestonePatterns: ReadonlyArray<{ name: Milestone; text: string }> = [
+  { name: "bunRootScript", text: "$ bun --cwd packages/desktop dev" },
+  { name: "bunDesktopScript", text: "$ bun ./scripts/dev.ts" },
+  { name: "desktopPrepared", text: "Copied dev icons from" },
+  { name: "mainBundleReady", text: "electron main process built successfully" },
+  { name: "preloadBundleReady", text: "electron preload scripts built successfully" },
+  { name: "rendererDevServerReady", text: "dev server running for the electron renderer process at:" },
+  { name: "electronSpawnStarted", text: "starting electron app..." },
+  { name: "debugEndpointReady", text: "DevTools listening on ws://" },
+  { name: "electronStarted", text: "app starting" },
+  { name: "serviceEnsureStarted", text: "starting v2 background service" },
+  { name: "serviceSpawnRequested", text: "v2 CLI background service starting" },
+  { name: "serviceReady", text: "v2 CLI background service ready" },
+  { name: "backgroundLoadingReady", text: "loading task finished" },
+  { name: "rendererViteConnected", text: "[vite] connected." },
+  { name: "rendererInitializationStarted", text: "awaiting server ready" },
+  { name: "rendererInitializationReady", text: "server ready" },
+  { name: "windowVisible", text: "main window visible" },
+]
+
+async function createColdProfile() {
+  await Promise.all(
+    ["packages/desktop/node_modules/.vite", "packages/desktop/out"].map((path) =>
+      rm(join(repository, path), { recursive: true, force: true }),
+    ),
+  )
+  const root = await mkdtemp(join(tmpdir(), "opencode-desktop-startup-"))
+  return initializeColdProfile(root).catch(async (error) => {
+    await rm(root, { recursive: true, force: true, maxRetries: 5, retryDelay: 100 })
+    throw error
+  })
+}
+
+async function initializeColdProfile(root: string) {
+  await Promise.all(
+    ["data", "config", "cache", "state", "desktop", "session", "home"].map((dir) =>
+      mkdir(join(root, dir), { recursive: true }),
+    ),
+  )
+  await Promise.all([
+    writeFile(
+      join(root, "desktop", "opencode.settings"),
+      JSON.stringify({ firstLaunchOnboardingComplete: true }),
+    ),
+    writeFile(join(root, "desktop", "opencode.global.dat"), JSON.stringify({ language: '{"locale":"en"}' })),
+  ])
+  const registration = join(root, "desktop", "opencode", "service-local.json")
+  await Service.stop({ file: registration })
+  return { root, registration }
+}
+
+function startDesktop(profile: Awaited<ReturnType<typeof createColdProfile>>) {
+  const started = performance.now()
+  const child = spawn("bun", ["dev:desktop"], {
+    cwd: repository,
+    detached: process.platform !== "win32",
+    env: {
+      ...process.env,
+      OPENCODE_CONFIG_DIR: join(profile.root, "config"),
+      OPENCODE_DB: join(profile.root, "data", "opencode.db"),
+      OPENCODE_TEST_HOME: join(profile.root, "home"),
+      OPENCODE_TEST_ONBOARDING: "0",
+      OPENCODE_DESKTOP_TEST_ROOT: profile.root,
+      OPENCODE_DESKTOP_REMOTE_DEBUGGING_PORT: "0",
+      OPENCODE_DESKTOP_DISABLE_PROTOCOL_REGISTRATION: "1",
+    },
+    stdio: ["ignore", "pipe", "pipe"],
+  })
+  if (!child.pid || !child.stdout || !child.stderr) throw new Error("Failed to start the desktop command")
+  const exited = childExit(child)
+  const observed: Partial<Record<Milestone, number>> = {}
+  const endpoint = Promise.withResolvers<string>()
+  const pageErrors: string[] = []
+  let browser: Browser | undefined
+  let service: ServiceInfo | undefined
+  const mark = (name: Milestone) => {
+    observed[name] ??= elapsed(started)
+  }
+  const record = (line: string) => {
+    const milestone = milestoneForLine(line)
+    if (milestone) mark(milestone)
+    const match = stripAnsi(line).match(/DevTools listening on (ws:\/\/\S+)/)
+    if (match?.[1]) endpoint.resolve(match[1])
+  }
+  const stdout = observeOutput(child.stdout, record)
+  const stderr = observeOutput(child.stderr, record)
+
+  return {
+    mark,
+    async open() {
+      const url = await Promise.race([
+        endpoint.promise,
+        exited.then((code) => {
+          throw new Error(`Desktop command exited with code ${code} before opening its debug endpoint`)
+        }),
+        sleep(120_000).then(() => {
+          throw new Error("Timed out waiting for the desktop debug endpoint")
+        }),
+      ])
+      browser = await chromium.connectOverCDP(url, { timeout: 120_000 })
+      const context = browser.contexts()[0]
+      if (!context) throw new Error("Electron did not expose a browser context")
+      await expect.poll(() => context.pages().length, { timeout: 120_000 }).toBeGreaterThan(0)
+      const page = context.pages()[0]
+      if (!page) throw new Error("Electron did not expose a renderer page")
+      page.on("pageerror", (error) => pageErrors.push(error.stack ?? error.message))
+      return page
+    },
+    async result(run: number): Promise<DesktopStartupSample> {
+      if (pageErrors.length) throw new Error(`Desktop renderer reported errors:\n\n${pageErrors.join("\n\n")}`)
+      service = await readService(profile)
+      const milestonesMs = requireMilestones(observed)
+      return {
+        run,
+        commandToHomeReadyMs: milestonesMs.homeReady,
+        milestonesMs,
+        phasesMs: calculatePhases(milestonesMs),
+        service: {
+          version: service.version,
+          url: service.url,
+          pid: service.pid,
+        },
+      }
+    },
+    async close(testInfo: TestInfo, run: number) {
+      const errors: unknown[] = []
+      await browser?.close().catch(() => undefined)
+      await stopProcessTree(child, exited).catch((error) => {
+        errors.push(error)
+        child.stdout?.destroy()
+        child.stderr?.destroy()
+      })
+      const [stdoutText, stderrText] = await Promise.all([stdout, stderr]).catch((error) => {
+        errors.push(error)
+        return ["", ""]
+      })
+      await Promise.all([
+        testInfo.attach(`desktop-startup-${run}-stdout`, { body: stdoutText, contentType: "text/plain" }),
+        testInfo.attach(`desktop-startup-${run}-stderr`, { body: stderrText, contentType: "text/plain" }),
+        pageErrors.length
+          ? testInfo.attach(`desktop-startup-${run}-page-errors`, {
+              body: pageErrors.join("\n\n"),
+              contentType: "text/plain",
+            })
+          : Promise.resolve(),
+      ]).catch((error) => errors.push(error))
+      await Service.stop({ file: profile.registration }).catch((error) => errors.push(error))
+      if (service && processAlive(service.pid))
+        errors.push(new Error(`Desktop service process ${service.pid} did not stop`))
+      await rm(profile.root, { recursive: true, force: true, maxRetries: 5, retryDelay: 100 }).catch((error) =>
+        errors.push(error),
+      )
+      if (errors.length) throw new AggregateError(errors, "Desktop benchmark cleanup failed")
+    },
+  }
+}
+
+async function waitForHome(page: Page, mark: (name: Milestone) => void) {
+  await expect.poll(() => page.evaluate(() => document.visibilityState), { timeout: 120_000 }).toBe("visible")
+
+  const projects = page.getByRole("complementary", { name: "Projects", exact: true })
+  const sessions = page.getByRole("region", { name: "Recent sessions", exact: true })
+  const search = page.getByRole("textbox", { name: "Search sessions", exact: true })
+  const addProject = projects.locator('button[data-action="home-add-project-row"]')
+  await expect(projects).toBeVisible({ timeout: 120_000 })
+  await expect(sessions).toBeVisible()
+  await expect(search).toBeEditable()
+  await expect(sessions.getByText("Nothing here yet", { exact: true })).toBeVisible()
+  await expect(addProject).toHaveCount(1)
+  await addProject.click({ trial: true })
+  mark("homeReady")
+}
+
+type ThemeWindow = Window & {
+  __OPENCODE_THEME_STATES__?: string[]
+  __OPENCODE_THEME_OBSERVER__?: MutationObserver
+}
+
+async function startThemeObservation(page: Page) {
+  await page.addInitScript(installThemeObservation)
+  await page.evaluate(installThemeObservation)
+}
+
+async function requireStableTheme(page: Page) {
+  const states = await page.evaluate(() => {
+    const target = window as ThemeWindow
+    target.__OPENCODE_THEME_OBSERVER__?.disconnect()
+    return target.__OPENCODE_THEME_STATES__ ?? []
+  })
+  if (states.length !== 1) throw new Error(`Desktop theme changed during startup: ${states.join(" -> ")}`)
+}
+
+function installThemeObservation() {
+  const target = window as ThemeWindow
+  const observeRoot = () => {
+    const root = document.documentElement
+    if (!root) return false
+    const state = () => {
+      const theme = root.dataset.theme
+      const scheme = root.dataset.colorScheme
+      return theme && scheme ? `${theme}:${scheme}` : undefined
+    }
+    const initial = state()
+    target.__OPENCODE_THEME_STATES__ = initial ? [initial] : []
+    target.__OPENCODE_THEME_OBSERVER__ = new MutationObserver(() => {
+      const next = state()
+      if (!next) return
+      if (target.__OPENCODE_THEME_STATES__?.at(-1) !== next) target.__OPENCODE_THEME_STATES__?.push(next)
+    })
+    target.__OPENCODE_THEME_OBSERVER__.observe(root, {
+      attributes: true,
+      attributeFilter: ["data-theme", "data-color-scheme"],
+    })
+    return true
+  }
+  if (observeRoot()) return
+  const documentObserver = new MutationObserver(() => {
+    if (!observeRoot()) return
+    documentObserver.disconnect()
+  })
+  target.__OPENCODE_THEME_OBSERVER__ = documentObserver
+  documentObserver.observe(document, { childList: true })
+}
+
+async function observeOutput(stream: NodeJS.ReadableStream, record: (line: string) => void) {
+  const decoder = new TextDecoder()
+  const output: string[] = []
+  let pending = ""
+  for await (const chunk of stream) {
+    const text = typeof chunk === "string" ? chunk : decoder.decode(chunk, { stream: true })
+    output.push(text)
+    pending += text
+    const lines = pending.split(/\r?\n/)
+    pending = lines.pop() ?? ""
+    lines.forEach(record)
+  }
+  const final = decoder.decode()
+  output.push(final)
+  pending += final
+  if (pending) record(pending)
+  return output.join("")
+}
+
+async function readService(profile: Awaited<ReturnType<typeof createColdProfile>>) {
+  const value: unknown = JSON.parse(await readFile(profile.registration, "utf8"))
+  if (!isServiceInfo(value)) throw new Error("Desktop service registration is invalid")
+  const url = new URL(value.url)
+  const port = Number(url.port)
+  if (url.hostname !== "127.0.0.1" || !Number.isInteger(port) || port <= 0)
+    throw new Error(`Desktop service used unexpected endpoint ${value.url}`)
+  if (!value.version.startsWith("2.0.0-local-"))
+    throw new Error(`Desktop service used unexpected version ${value.version}`)
+  return value
+}
+
+function isServiceInfo(value: unknown): value is ServiceInfo {
+  return (
+    typeof value === "object" &&
+    value !== null &&
+    "id" in value &&
+    typeof value.id === "string" &&
+    "version" in value &&
+    typeof value.version === "string" &&
+    "url" in value &&
+    typeof value.url === "string" &&
+    "pid" in value &&
+    typeof value.pid === "number"
+  )
+}
+
+function requireMilestones(observed: Partial<Record<Milestone, number>>) {
+  const get = (name: Milestone) => {
+    const value = observed[name]
+    if (value === undefined) throw new Error(`Desktop startup did not report milestone: ${name}`)
+    return round(value)
+  }
+  return {
+    bunRootScript: get("bunRootScript"),
+    bunDesktopScript: get("bunDesktopScript"),
+    desktopPrepared: get("desktopPrepared"),
+    mainBundleReady: get("mainBundleReady"),
+    preloadBundleReady: get("preloadBundleReady"),
+    rendererDevServerReady: get("rendererDevServerReady"),
+    electronSpawnStarted: get("electronSpawnStarted"),
+    debugEndpointReady: get("debugEndpointReady"),
+    electronStarted: get("electronStarted"),
+    serviceEnsureStarted: get("serviceEnsureStarted"),
+    serviceSpawnRequested: get("serviceSpawnRequested"),
+    serviceReady: get("serviceReady"),
+    backgroundLoadingReady: get("backgroundLoadingReady"),
+    rendererViteConnected: get("rendererViteConnected"),
+    rendererInitializationStarted: get("rendererInitializationStarted"),
+    rendererInitializationReady: get("rendererInitializationReady"),
+    windowVisible: get("windowVisible"),
+    homeReady: get("homeReady"),
+  }
+}
+
+function calculatePhases(value: Record<Milestone, number>): Record<Phase, number> {
+  return {
+    desktopPreparation: value.desktopPrepared,
+    viteMainBundle: round(value.mainBundleReady - value.desktopPrepared),
+    vitePreloadBundle: round(value.preloadBundleReady - value.mainBundleReady),
+    rendererServerStartup: round(value.rendererDevServerReady - value.preloadBundleReady),
+    electronStartup: round(value.electronStarted - value.electronSpawnStarted),
+    serviceSpawnWait: round(value.serviceSpawnRequested - value.serviceEnsureStarted),
+    serviceProcessStartup: round(value.serviceReady - value.serviceSpawnRequested),
+    rendererStartup: round(value.homeReady - value.rendererViteConnected),
+    visibleWindowToHome: round(value.homeReady - value.windowVisible),
+  }
+}
+
+function statistics(values: number[]) {
+  if (!values.length) throw new Error("Cannot summarize an empty benchmark")
+  const sorted = values.toSorted((left, right) => left - right)
+  const median = medianOf(sorted)
+  return {
+    min: round(sorted[0]),
+    median: round(median),
+    max: round(sorted.at(-1)!),
+    medianAbsoluteDeviation: round(medianOf(sorted.map((value) => Math.abs(value - median)).toSorted((a, b) => a - b))),
+  }
+}
+
+function medianOf(sorted: number[]) {
+  const middle = Math.floor(sorted.length / 2)
+  if (sorted.length % 2) return sorted[middle]
+  return (sorted[middle - 1] + sorted[middle]) / 2
+}
+
+async function stopProcessTree(child: ChildProcess, exited: Promise<number | null>) {
+  if (!child.pid) throw new Error("Desktop command has no process ID")
+  if (process.platform !== "win32") return stopProcessGroup(child.pid, exited)
+  if (child.exitCode !== null || (await exitsWithin(child, exited, 2_000))) return
+  const kill = spawn("taskkill.exe", ["/PID", String(child.pid), "/T", "/F"], {
+    stdio: "ignore",
+  })
+  await childExit(kill)
+  if (await exitsWithin(child, exited, 10_000)) return
+  if (!(await exitsWithin(child, exited, 5_000))) throw new Error(`Desktop command process ${child.pid} did not stop`)
+}
+
+async function stopProcessGroup(pid: number, exited: Promise<number | null>) {
+  await Promise.race([exited, sleep(2_000)])
+  if (!processGroupAlive(pid)) return
+  process.kill(-pid, "SIGTERM")
+  if (await processGroupStopsWithin(pid, 10_000)) return
+  process.kill(-pid, "SIGKILL")
+  if (!(await processGroupStopsWithin(pid, 5_000))) throw new Error(`Desktop command process group ${pid} did not stop`)
+}
+
+async function processGroupStopsWithin(pid: number, timeout: number) {
+  const deadline = Date.now() + timeout
+  while (Date.now() < deadline) {
+    if (!processGroupAlive(pid)) return true
+    await sleep(50)
+  }
+  return !processGroupAlive(pid)
+}
+
+function processGroupAlive(pid: number) {
+  try {
+    process.kill(-pid, 0)
+    return true
+  } catch {
+    return false
+  }
+}
+
+async function exitsWithin(child: ChildProcess, exited: Promise<number | null>, timeout: number) {
+  if (child.exitCode !== null) return true
+  const result = await Promise.race([exited.then(() => true), sleep(timeout).then(() => false)])
+  return result
+}
+
+function childExit(child: ChildProcess) {
+  return new Promise<number | null>((resolve, reject) => {
+    child.once("error", reject)
+    child.once("exit", (code) => resolve(code))
+  })
+}
+
+function sleep(milliseconds: number) {
+  return new Promise<void>((resolve) => setTimeout(resolve, milliseconds))
+}
+
+function processAlive(pid: number) {
+  try {
+    process.kill(pid, 0)
+    return true
+  } catch {
+    return false
+  }
+}
+
+function stripAnsi(value: string) {
+  return value.replace(/\x1b\[[0-?]*[ -/]*[@-~]/g, "")
+}
+
+function elapsed(started: number) {
+  return round(performance.now() - started)
+}
+
+function round(value: number) {
+  return Math.round(value * 100) / 100
+}

packages/app/e2e/performance/devex/playwright.config.ts

@@ -0,0 +1,17 @@
+import { defineConfig } from "@playwright/test"
+
+process.env.OPENCODE_PERFORMANCE_RUN_ID ??= `${new Date().toISOString().replace(/[:.]/g, "-")}-${process.pid}`
+
+export default defineConfig({
+  testDir: ".",
+  testMatch: "desktop-startup-benchmark.spec.ts",
+  outputDir: "../../test-results/performance-devex",
+  timeout: 15 * 60_000,
+  expect: {
+    timeout: 120_000,
+  },
+  fullyParallel: false,
+  workers: 1,
+  reporter: [["html", { outputFolder: "../../playwright-report/performance-devex", open: "never" }], ["line"]],
+  projects: [{ name: "desktop" }],
+})

packages/app/e2e/performance/playwright.config.ts

@@ -7,7 +7,7 @@ process.env.OPENCODE_PERFORMANCE_RUN_ID ??= `${new Date().toISOString().replace(
 export default {
   ...config,
   testDir: ".",
-  testIgnore: "unit/**",
+  testIgnore: ["unit/**", "devex/**"],
   outputDir: "../test-results/performance",
   fullyParallel: false,
   workers: 1,

packages/app/e2e/performance/timeline-stability/adverse.spec.ts

@@ -179,7 +179,7 @@ test.describe("timeline adverse visual stability", () => {
         userMessage(),
         assistantMessage([
           shell(shellID, "completed", wideLines(15)),
-          toolPart(contextIDs[0]!, "read", "completed", { filePath: "src/a.ts" }),
+          toolPart(contextIDs[0]!, "read", "completed", { path: "src/a.ts" }),
           toolPart(contextIDs[1]!, "glob", "completed", { path: ".", pattern: "**/*.ts" }),
           textPart(followingID, "Following responsive timeline content that wraps on narrow screens."),
         ]),

packages/app/e2e/performance/timeline-stability/file-matrix.spec.ts

@@ -1,4 +1,5 @@
 import { test } from "@playwright/test"
+import { createTwoFilesPatch } from "diff"
 import {
   defineVisualRegions,
   reportVisualStability,
@@ -18,16 +19,20 @@ import {
 } from "./fixture"
 
 const profiles = [
-  { name: "edit", tool: "edit", input: { filePath: "src/edit.ts" } },
+  {
+    name: "edit",
+    tool: "edit",
+    input: { path: "src/edit.ts", oldString: "export const value = 1", newString: "export const value = 2" },
+  },
   {
     name: "multi patch",
-    tool: "apply_patch",
-    input: { files: ["src/a.ts", "src/b.ts", "src/old.ts", "src/moved.ts"] },
+    tool: "patch",
+    input: { patchText: "Update generated files" },
   },
 ] as const
 
 for (const profile of profiles) {
-  test(`stabilizes ${profile.name} pending to completed`, async ({ page }, testInfo) => {
+  test(`stabilizes ${profile.name} streaming to completed`, async ({ page }, testInfo) => {
     const partID = `prt_file_matrix_${profiles.indexOf(profile)}`
     const followingID = `prt_file_matrix_following_${profiles.indexOf(profile)}`
     const timeline = await setupTimeline(page, {
@@ -35,7 +40,7 @@ for (const profile of profiles) {
         userMessage(),
         assistantMessage(
           [
-            toolPart(partID, profile.tool, "pending", profile.input),
+            toolPart(partID, profile.tool, "streaming", profile.input),
             textPart(followingID, `Following ${profile.name}`),
           ],
           { completed: false },
@@ -89,34 +94,27 @@ function completedPart(partID: string, profile: (typeof profiles)[number]) {
   if (profile.tool === "edit") {
     return toolPart(partID, profile.tool, "completed", profile.input, {
       metadata: {
-        filediff: {
-          file: "src/edit.ts",
-          additions: 50,
-          deletions: 50,
-          before: source(50, false),
-          after: source(50, true),
-        },
+        files: [patchFile("src/edit.ts", "modified", 50)],
       },
     })
   }
   const files = [
-    patchFile("src/a.ts", "update"),
-    patchFile("src/b.ts", "add"),
-    patchFile("src/old.ts", "delete"),
-    { ...patchFile("src/moved.ts", "move"), move: "src/new-place.ts" },
+    patchFile("src/a.ts", "modified", 20),
+    patchFile("src/b.ts", "added", 20),
+    patchFile("src/old.ts", "deleted", 20),
   ]
   return toolPart(partID, profile.tool, "completed", profile.input, { metadata: { files } })
 }
 
-function patchFile(filePath: string, type: "add" | "update" | "delete" | "move") {
+function patchFile(file: string, status: "added" | "modified" | "deleted", lines: number) {
+  const before = status === "added" ? "" : source(lines, false)
+  const after = status === "deleted" ? "" : source(lines, true)
   return {
-    filePath,
-    relativePath: filePath,
-    type,
-    additions: type === "delete" ? 0 : 20,
-    deletions: type === "add" ? 0 : 20,
-    before: type === "add" ? undefined : source(20, false),
-    after: type === "delete" ? undefined : source(20, true),
+    file,
+    status,
+    patch: createTwoFilesPatch(`a/${file}`, `b/${file}`, before, after),
+    additions: status === "deleted" ? 0 : lines,
+    deletions: status === "added" ? 0 : lines,
   }
 }
 

packages/app/e2e/performance/timeline-stability/file-mutation.spec.ts

@@ -1,4 +1,5 @@
 import { expect, test } from "@playwright/test"
+import { createTwoFilesPatch } from "diff"
 import {
   defineVisualRegions,
   reportVisualStability,
@@ -20,13 +21,13 @@ import {
 test("adds patch files incrementally without resetting outer expansion", async ({ page }, testInfo) => {
   const patchID = "prt_incremental_01_patch"
   const followingID = "prt_incremental_02_following"
-  const first = patchFile("src/a.ts", "update")
+  const first = patchFile("src/a.ts", "modified")
   const timeline = await setupTimeline(page, {
     messages: [
       userMessage(),
       assistantMessage(
         [
-          toolPart(patchID, "apply_patch", "running", { files: [first.filePath] }, { metadata: { files: [first] } }),
+          toolPart(patchID, "patch", "running", { patchText: "Update files" }, { metadata: { files: [first] } }),
           textPart(followingID, "Following incremental patch"),
         ],
         { completed: false },
@@ -55,15 +56,15 @@ test("adds patch files incrementally without resetting outer expansion", async (
     },
   })
   await startVisualProbe(page, regions)
-  const second = patchFile("src/b.ts", "add")
-  const third = patchFile("src/old.ts", "delete")
+  const second = patchFile("src/b.ts", "added")
+  const third = patchFile("src/old.ts", "deleted")
   await timeline.send(
     partUpdated(
       toolPart(
         patchID,
-        "apply_patch",
+        "patch",
         "running",
-        { files: [first.filePath, second.filePath] },
+        { patchText: "Update files" },
         { metadata: { files: [first, second] } },
       ),
     ),
@@ -73,9 +74,9 @@ test("adds patch files incrementally without resetting outer expansion", async (
     partUpdated(
       toolPart(
         patchID,
-        "apply_patch",
+        "patch",
         "completed",
-        { files: [first.filePath, second.filePath, third.filePath] },
+        { patchText: "Update files" },
         { metadata: { files: [first, second, third] } },
       ),
     ),
@@ -106,15 +107,15 @@ test("adds patch files incrementally without resetting outer expansion", async (
   await expect(page.locator('[data-scope="apply-patch"] [data-type="delete"]')).toBeVisible()
 })
 
-function patchFile(filePath: string, type: "add" | "update" | "delete") {
+function patchFile(file: string, status: "added" | "modified" | "deleted") {
+  const before = status === "added" ? "" : source(false)
+  const after = status === "deleted" ? "" : source(true)
   return {
-    filePath,
-    relativePath: filePath,
-    type,
-    additions: type === "delete" ? 0 : 4,
-    deletions: type === "add" ? 0 : 3,
-    before: type === "add" ? undefined : source(false),
-    after: type === "delete" ? undefined : source(true),
+    file,
+    status,
+    patch: createTwoFilesPatch(`a/${file}`, `b/${file}`, before, after),
+    additions: status === "deleted" ? 0 : 4,
+    deletions: status === "added" ? 0 : 3,
   }
 }
 

packages/app/e2e/performance/timeline-stability/fixture.test.ts

@@ -35,7 +35,7 @@ describe("timeline fixture validation", () => {
         userMessage(),
         {
           ...assistantMessage(),
-          content: [{ type: "tool", id: "call_invalid", name: "bash", state: { status: "completed" } }],
+          content: [{ type: "tool", id: "call_invalid", name: "shell", state: { status: "completed" } }],
         } as never,
       ]),
     ).toThrow()
@@ -60,12 +60,11 @@ if (false) {
   const userSeed = { id: "prt_type_user", type: "text", text: "typed" } satisfies PartSeed<"user">
   userMessage([userSeed])
 
-  // @ts-expect-error Tool completion fields are not valid while pending.
-  toolPart("prt_invalid_pending", "bash", "pending", {}, { output: "impossible" })
-  // @ts-expect-error Tool completion fields are not valid while running.
-  toolPart("prt_invalid_running", "bash", "running", {}, { output: "impossible" })
+  // @ts-expect-error Tool completion fields are not valid while streaming.
+  toolPart("prt_invalid_streaming", "shell", "streaming", {}, { output: "impossible" })
+  toolPart("prt_valid_running", "shell", "running", {}, { output: "progressive output" })
   // @ts-expect-error Tool error fields are not valid after completion.
-  toolPart("prt_invalid_completed", "bash", "completed", {}, { error: "impossible" })
+  toolPart("prt_invalid_completed", "shell", "completed", {}, { error: "impossible" })
 
   assistantMessage([
     // @ts-expect-error Agent references belong to user messages, not assistant messages.

packages/app/e2e/performance/timeline-stability/fixture.ts

@@ -60,17 +60,17 @@ type ReasoningSeed = {
 type ToolSeed = {
   id: string
   type: "tool"
-  callID: string
-  tool: string
+  name: string
   messageID?: string
   executed?: boolean
   providerState?: Record<string, unknown>
   providerResultState?: Record<string, unknown>
   state:
-    | { status: "pending"; input: Record<string, unknown>; raw: string }
+    | { status: "streaming"; input: Record<string, unknown>; raw: string }
     | {
         status: "running"
         input: Record<string, unknown>
+        output?: string
         title?: string
         metadata: Record<string, unknown>
         time: { start: number }
@@ -100,10 +100,10 @@ export type PartSeed<Owner extends "user" | "assistant"> = Owner extends "user"
   ? TextSeed | FileSeed | AgentSeed
   : TextSeed | ReasoningSeed | ToolSeed
 
-type ToolOptions<State extends ToolStatus> = State extends "pending"
+type ToolOptions<State extends ToolStatus> = State extends "streaming"
   ? { output?: never; title?: never; metadata?: never; error?: never }
   : State extends "running"
-    ? { title?: string; metadata?: Record<string, unknown>; output?: never; error?: never }
+    ? { title?: string; metadata?: Record<string, unknown>; output?: string; error?: never }
     : State extends "error"
       ? { error?: string; metadata?: Record<string, unknown>; output?: never; title?: never }
       : { output?: string; title?: string; metadata?: Record<string, unknown>; error?: never }
@@ -259,6 +259,34 @@ export function event(
   return makeEvent(type, data)
 }
 
+export function compactionStarted(data: Extract<OpenCodeEvent, { type: "session.compaction.started" }>["data"]) {
+  return makeEvent("session.compaction.started", data)
+}
+
+export function compactionDelta(data: Extract<OpenCodeEvent, { type: "session.compaction.delta" }>["data"]) {
+  return makeEvent("session.compaction.delta", data)
+}
+
+export function compactionEnded(data: Extract<OpenCodeEvent, { type: "session.compaction.ended" }>["data"]) {
+  return makeEvent("session.compaction.ended", data)
+}
+
+export function compactionFailed(data: Extract<OpenCodeEvent, { type: "session.compaction.failed" }>["data"]) {
+  return makeEvent("session.compaction.failed", data)
+}
+
+export function toolInputStarted(data: Extract<OpenCodeEvent, { type: "session.tool.input.started" }>["data"]) {
+  return makeEvent("session.tool.input.started", data)
+}
+
+export function toolInputEnded(data: Extract<OpenCodeEvent, { type: "session.tool.input.ended" }>["data"]) {
+  return makeEvent("session.tool.input.ended", data)
+}
+
+export function toolCalled(data: Extract<OpenCodeEvent, { type: "session.tool.called" }>["data"]) {
+  return makeEvent("session.tool.called", data)
+}
+
 export function validateTimelineEvent(input: unknown): OpenCodeEvent {
   if (!input || typeof input !== "object") throw new Error("Timeline event must be an object")
   if (!("type" in input) || typeof input.type !== "string") throw new Error("Timeline event requires a type")
@@ -371,6 +399,15 @@ export function partUpdated(part: PartSeed<"assistant">): readonly OpenCodeEvent
   }
   if (part.type === "reasoning") {
     startedParts.add(part.id)
+    if (!started && !part.text)
+      return [
+        makeEvent("session.reasoning.started", {
+          sessionID,
+          assistantMessageID: messageID,
+          ordinal: ref.ordinal!,
+          state: jsonRecord(part.metadata),
+        }),
+      ]
     return [
       ...(started
         ? []
@@ -542,9 +579,9 @@ export function reasoningPart(id: string, text: string): ReasoningSeed {
 export function toolPart(
   id: string,
   tool: string,
-  state: "pending",
+  state: "streaming",
   input: Record<string, unknown>,
-  options?: ToolOptions<"pending">,
+  options?: ToolOptions<"streaming">,
 ): ToolSeed
 export function toolPart(
   id: string,
@@ -574,14 +611,15 @@ export function toolPart(
   input: Record<string, unknown>,
   options: ToolOptions<ToolStatus> = {},
 ): ToolSeed {
-  const base = { id, type: "tool" as const, callID: id, tool }
-  if (state === "pending") return { ...base, state: { status: state, input, raw: "" } }
+  const base = { id, type: "tool" as const, name: tool }
+  if (state === "streaming") return { ...base, state: { status: state, input, raw: "" } }
   if (state === "running")
     return {
       ...base,
       state: {
         status: state,
         input,
+        ...(options.output === undefined ? {} : { output: options.output }),
         title: options.title,
         metadata: options.metadata ?? {},
         time: { start: 1700000001000 },
@@ -612,12 +650,10 @@ export function toolPart(
 }
 
 export function shell(id: string, state: ToolStatus, output = "", command = `echo ${id}`): ToolSeed {
-  if (state === "pending") return toolPart(id, "bash", state, { command })
-  if (state === "running")
-    return toolPart(id, "bash", state, { command }, { title: command, metadata: { command, output } })
-  if (state === "error")
-    return toolPart(id, "bash", state, { command }, { error: output || undefined, metadata: { command, output } })
-  return toolPart(id, "bash", state, { command }, { title: command, output, metadata: { command, output } })
+  if (state === "streaming") return toolPart(id, "shell", state, { command })
+  if (state === "running") return toolPart(id, "shell", state, { command }, { title: command, output })
+  if (state === "error") return toolPart(id, "shell", state, { command }, { error: output || undefined })
+  return toolPart(id, "shell", state, { command }, { title: command, output })
 }
 
 export function completedAssistantInfo(info: SessionMessageAssistant): SessionMessageAssistant {
@@ -655,7 +691,7 @@ function messageContent(
 ): SessionMessageAssistant["content"][number] {
   if (part.type === "tool") {
     partRefs.set(part.id, { messageID, type: part.type })
-    toolStates.set(part.callID, part.state.status)
+    toolStates.set(part.id, part.state.status)
   } else {
     partRefs.set(part.id, { messageID, type: part.type, ordinal: ordinals[part.type]++ })
     startedParts.add(part.id)
@@ -675,8 +711,8 @@ function messageContent(
   const completed = state.status === "completed" || state.status === "error" ? state.time.end : undefined
   const base = {
     type: "tool" as const,
-    id: part.callID,
-    name: part.tool,
+    id: part.id,
+    name: part.name,
     time: {
       created: time?.start ?? 1700000001000,
       ...(time?.start === undefined ? {} : { ran: time.start }),
@@ -686,11 +722,18 @@ function messageContent(
     ...(part.providerState ? { providerState: jsonRecord(part.providerState) } : {}),
     ...(part.providerResultState ? { providerResultState: jsonRecord(part.providerResultState) } : {}),
   }
-  if (state.status === "pending") return { ...base, state: { status: "streaming", input: state.raw } }
+  if (state.status === "streaming") return { ...base, state: { status: "streaming", input: state.raw } }
   if (state.status === "running")
     return {
       ...base,
-      state: { status: "running", input: jsonRecord(state.input), metadata: jsonRecord(state.metadata) },
+      state: {
+        status: "running",
+        input: jsonRecord(state.input),
+        metadata: jsonRecord({
+          ...state.metadata,
+          ...(state.output === undefined ? {} : { output: state.output }),
+        }),
+      },
     }
   if (state.status === "error")
     return {
@@ -714,7 +757,7 @@ function messageContent(
 }
 
 function toolEvents(part: ToolSeed, messageID: string): readonly OpenCodeEvent[] {
-  const previous = toolStates.get(part.callID)
+  const previous = toolStates.get(part.id)
   if (previous === "completed" || previous === "error") return []
 
   const events: OpenCodeEvent[] = []
@@ -723,27 +766,27 @@ function toolEvents(part: ToolSeed, messageID: string): readonly OpenCodeEvent[]
       makeEvent("session.tool.input.started", {
         sessionID,
         assistantMessageID: messageID,
-        id: part.callID,
-        name: part.tool,
+        id: part.id,
+        name: part.name,
       }),
     )
   }
-  if (part.state.status === "pending") {
-    toolStates.set(part.callID, part.state.status)
+  if (part.state.status === "streaming") {
+    toolStates.set(part.id, part.state.status)
     return events
   }
-  if (!previous || previous === "pending") {
+  if (!previous || previous === "streaming") {
     events.push(
       makeEvent("session.tool.input.ended", {
         sessionID,
         assistantMessageID: messageID,
-        id: part.callID,
+        id: part.id,
         text: JSON.stringify(part.state.input),
       }),
       makeEvent("session.tool.called", {
         sessionID,
         assistantMessageID: messageID,
-        id: part.callID,
+        id: part.id,
         input: part.state.input,
         executed: part.executed ?? true,
         state: jsonRecord(part.providerState),
@@ -751,16 +794,20 @@ function toolEvents(part: ToolSeed, messageID: string): readonly OpenCodeEvent[]
     )
   }
   if (part.state.status === "running") {
-    if (previous === "running" || Object.keys(part.state.metadata).length)
+    const metadata = {
+      ...part.state.metadata,
+      ...(part.state.output === undefined ? {} : { output: part.state.output }),
+    }
+    if (previous === "running" || Object.keys(metadata).length)
       events.push(
         makeEvent("session.tool.progress", {
           sessionID,
           assistantMessageID: messageID,
-          id: part.callID,
-          metadata: jsonRecord(part.state.metadata),
+          id: part.id,
+          metadata: jsonRecord(metadata),
         }),
       )
-    toolStates.set(part.callID, part.state.status)
+    toolStates.set(part.id, part.state.status)
     return events
   }
   if (part.state.status === "error") {
@@ -768,28 +815,28 @@ function toolEvents(part: ToolSeed, messageID: string): readonly OpenCodeEvent[]
       makeEvent("session.tool.failed", {
         sessionID,
         assistantMessageID: messageID,
-        id: part.callID,
+        id: part.id,
         error: { type: "ToolError", message: part.state.error },
         metadata: jsonRecord(part.state.metadata),
         executed: part.executed ?? true,
         resultState: jsonRecord(part.providerResultState),
       }),
     )
-    toolStates.set(part.callID, part.state.status)
+    toolStates.set(part.id, part.state.status)
     return events
   }
   events.push(
     makeEvent("session.tool.success", {
       sessionID,
       assistantMessageID: messageID,
-      id: part.callID,
+      id: part.id,
       content: [{ type: "text", text: part.state.output }],
       metadata: jsonRecord(part.state.metadata),
       executed: part.executed ?? true,
       resultState: jsonRecord(part.providerResultState),
     }),
   )
-  toolStates.set(part.callID, part.state.status)
+  toolStates.set(part.id, part.state.status)
   return events
 }
 

packages/app/e2e/performance/timeline-stability/interaction.spec.ts

@@ -1,4 +1,5 @@
 import { expect, test } from "@playwright/test"
+import { createTwoFilesPatch } from "diff"
 import {
   defineVisualRegions,
   reportVisualStability,
@@ -58,14 +59,14 @@ test("expands and collapses a long completed shell without overlap", async ({ pa
   await startVisualProbe(page, regions)
   await trigger.click()
   await expect(trigger).toHaveAttribute("aria-expanded", "true")
-  await page.waitForTimeout(500)
+  await waitForVisualSettle(page, [regions.shell.selector, regions.following.selector])
   const expanded = await stopVisualProbe<keyof typeof regions>(page)
   await reportVisualStability(testInfo, "shell-expand", expanded, plan)
 
   await startVisualProbe(page, regions)
   await trigger.click()
   await expect(trigger).toHaveAttribute("aria-expanded", "false")
-  await page.waitForTimeout(500)
+  await waitForVisualSettle(page, [regions.shell.selector, regions.following.selector])
   const collapsed = await stopVisualProbe<keyof typeof regions>(page)
   await reportVisualStability(testInfo, "shell-collapse", collapsed, plan)
 })
@@ -83,7 +84,7 @@ test("expands and collapses a completed context group without overlap", async ({
     messages: [
       userMessage(),
       assistantMessage([
-        toolPart(ids[0]!, "read", "completed", { filePath: "src/a.ts" }),
+        toolPart(ids[0]!, "read", "completed", { path: "src/a.ts" }),
         toolPart(ids[1]!, "glob", "completed", { path: ".", pattern: "**/*.ts" }),
         toolPart(ids[2]!, "grep", "completed", { path: ".", pattern: "stable" }),
         toolPart(ids[3]!, "list", "completed", { path: "src" }),
@@ -110,7 +111,7 @@ test("expands and collapses a completed context group without overlap", async ({
     await startVisualProbe(page, regions)
     await trigger.click()
     await expect(trigger).toHaveAttribute("aria-expanded", String(expanded))
-    await page.waitForTimeout(500)
+    await waitForVisualSettle(page, [regions.context.selector, regions.following.selector])
     const trace = await stopVisualProbe<keyof typeof regions>(page)
     await reportVisualStability(
       testInfo,
@@ -142,16 +143,23 @@ test("expands and collapses an edit diff without moving twice", async ({ page },
           editID,
           "edit",
           "completed",
-          { filePath: "src/edit.ts" },
+          { path: "src/edit.ts", oldString: "export const value = 1", newString: "export const value = 2" },
           {
             metadata: {
-              filediff: {
-                file: "src/edit.ts",
-                additions: 40,
-                deletions: 40,
-                before: source(40, false),
-                after: source(40, true),
-              },
+              files: [
+                {
+                  file: "src/edit.ts",
+                  patch: createTwoFilesPatch(
+                    "a/src/edit.ts",
+                    "b/src/edit.ts",
+                    source(40, false),
+                    source(40, true),
+                  ),
+                  additions: 40,
+                  deletions: 40,
+                  status: "modified",
+                },
+              ],
             },
           },
         ),
@@ -182,7 +190,7 @@ test("expands and collapses an edit diff without moving twice", async ({ page },
   await startVisualProbe(page, regions)
   await trigger.click()
   await expect(trigger).toHaveAttribute("aria-expanded", "true")
-  await page.waitForTimeout(900)
+  await waitForVisualSettle(page, [regions.edit.selector, regions.following.selector])
   const trace = await stopVisualProbe<keyof typeof regions>(page)
   await reportVisualStability(
     testInfo,

packages/app/e2e/performance/timeline-stability/scroll-interaction.spec.ts

@@ -1,4 +1,5 @@
 import { expect, test } from "@playwright/test"
+import { createTwoFilesPatch } from "diff"
 import {
   defineVisualRegions,
   reportVisualStability,
@@ -13,10 +14,63 @@ import {
   setupTimeline,
   shell,
   textPart,
+  toolPart,
   userMessage,
   type TimelineMessage,
 } from "./fixture"
 
+test("follows an expanded patch that arrives as the user reaches the bottom", async ({ page }) => {
+  const toolID = "prt_bottom_follow_patch"
+  const input = { patchText: "Update src/edit.ts" }
+  const timeline = await setupTimeline(page, {
+    messages: [
+      ...history(20),
+      userMessage(),
+      assistantMessage([textPart("prt_bottom_follow_text", "Working")], { completed: false }),
+    ],
+    settings: { editToolPartsExpanded: true },
+    reducedMotion: true,
+  })
+  const scroller = page.locator(".scroll-view__viewport", { has: page.locator("[data-timeline-row]") })
+  await scroller.evaluate((element) => {
+    element.scrollTop = Math.max(0, element.scrollHeight - element.clientHeight - 300)
+    element.dispatchEvent(new WheelEvent("wheel", { bubbles: true, cancelable: true, deltaY: 300 }))
+    element.scrollTop = element.scrollHeight
+  })
+  await expect
+    .poll(() => scroller.evaluate((element) => element.scrollHeight - element.clientHeight - element.scrollTop))
+    .toBeLessThanOrEqual(1)
+
+  await timeline.send(partUpdated(toolPart(toolID, "patch", "running", input)))
+  await timeline.send(
+    partUpdated(
+      toolPart(toolID, "patch", "completed", input, {
+        metadata: {
+          files: [
+            {
+              file: "src/edit.ts",
+              status: "modified",
+              patch: createTwoFilesPatch(
+                "a/src/edit.ts",
+                "b/src/edit.ts",
+                Array.from({ length: 40 }, (_, index) => `export const value${index} = ${index}\n`).join(""),
+                Array.from({ length: 40 }, (_, index) => `export const value${index} = ${index + 1}\n`).join(""),
+              ),
+              additions: 40,
+              deletions: 40,
+            },
+          ],
+        },
+      }),
+    ),
+  )
+
+  await timeline.waitForPart(toolID)
+  await expect
+    .poll(() => scroller.evaluate((element) => element.scrollHeight - element.clientHeight - element.scrollTop))
+    .toBeLessThanOrEqual(1)
+})
+
 test("does not reverse visible rows when the user wheels during shell remeasurement", async ({ page }, testInfo) => {
   const shellID = "prt_wheel_01_shell"
   const followingID = "prt_wheel_02_following"

packages/app/e2e/performance/timeline-stability/tool-mutation.spec.ts

@@ -17,32 +17,32 @@ import {
   userMessage,
 } from "./fixture"
 
-test("adds a task child-session link without replacing the task row", async ({ page }, testInfo) => {
-  const taskID = "prt_task_link"
-  const childID = "ses_task_child"
-  const input = { description: "Inspect child", subagent_type: "explore" }
+test("adds a subagent child-session link without replacing the row", async ({ page }, testInfo) => {
+  const taskID = "prt_subagent_link"
+  const childID = "ses_subagent_child"
+  const input = { description: "Inspect child", agent: "explore", prompt: "Inspect the child Session." }
   const timeline = await setupTimeline(page, {
-    messages: [userMessage(), assistantMessage([toolPart(taskID, "task", "running", input)], { completed: false })],
+    messages: [userMessage(), assistantMessage([toolPart(taskID, "subagent", "running", input)], { completed: false })],
     sessions: [session(), session({ id: childID, parentID: sessionID, title: "Inspect child" })],
     cpuRate: 4,
   })
   const regions = defineVisualRegions({
-    task: { selector: `[data-timeline-part-id="${renderedPartID(taskID)}"] [data-slot="collapsible-trigger"]` },
+    subagent: { selector: `[data-timeline-part-id="${renderedPartID(taskID)}"] [data-slot="collapsible-trigger"]` },
   })
   await startVisualProbe(page, regions)
   await timeline.send(
-    partUpdated(toolPart(taskID, "task", "completed", input, { metadata: { sessionId: childID } })),
+    partUpdated(toolPart(taskID, "subagent", "completed", input, { metadata: { sessionID: childID } })),
     500,
   )
   const trace = await stopVisualProbe<keyof typeof regions>(page)
   await reportVisualStability(
     testInfo,
-    "task-link",
+    "subagent-link",
     trace,
     visualPlan(regions, [
-      { type: "required", regions: ["task"] },
-      { type: "unique", regions: ["task"] },
-      { type: "stable", regions: ["task"] },
+      { type: "required", regions: ["subagent"] },
+      { type: "unique", regions: ["subagent"] },
+      { type: "stable", regions: ["subagent"] },
       { type: "opacity", regions: "all" },
       { type: "continuity", regions: "all" },
       { type: "motion", regions: "all", maxPositionReversals: 0 },

packages/app/e2e/performance/timeline-stability/tools.spec.ts

@@ -21,24 +21,30 @@ import {
 } from "./fixture"
 
 test.describe("timeline tool state stability", () => {
-  test("moves lightweight tools through pending, running, and completed without replacing rows", async ({
+  test("moves lightweight tools through streaming, running, and completed without replacing rows", async ({
     page,
   }, testInfo) => {
-    const ids = ["webfetch", "websearch", "task", "skill", "custom"] as const
+    const ids = ["webfetch", "websearch", "subagent", "skill", "custom"] as const
     const inputs = {
       webfetch: { url: "https://example.com/docs" },
       websearch: { query: "timeline stability" },
-      task: { description: "Inspect timeline", subagent_type: "explore" },
+      subagent: { description: "Inspect timeline", agent: "explore", prompt: "Inspect the timeline." },
       skill: { name: "stability" },
       custom: { target: "timeline", depth: 2 },
     }
-    const names = { webfetch: "webfetch", websearch: "websearch", task: "task", skill: "skill", custom: "mcp_probe" }
+    const names = {
+      webfetch: "webfetch",
+      websearch: "websearch",
+      subagent: "subagent",
+      skill: "skill",
+      custom: "mcp_probe",
+    }
     const questionID = "prt_state_question"
     const todoID = "prt_state_todo"
     const initial = [
-      ...ids.map((id) => toolPart(`prt_state_${id}`, names[id], "pending", inputs[id])),
-      toolPart(questionID, "question", "pending", questionInput()),
-      toolPart(todoID, "todowrite", "pending", { todos: [{ content: "Hidden", status: "pending" }] }),
+      ...ids.map((id) => toolPart(`prt_state_${id}`, names[id], "streaming", inputs[id])),
+      toolPart(questionID, "question", "streaming", questionInput()),
+      toolPart(todoID, "todowrite", "streaming", { todos: [{ content: "Hidden", status: "pending" }] }),
       textPart("prt_state_following", "Following lightweight tools"),
     ]
     const childID = "ses_timeline_child"
@@ -55,14 +61,14 @@ test.describe("timeline tool state stability", () => {
     const regionIDs = [
       "prt_state_webfetch",
       "prt_state_websearch",
-      "prt_state_task",
+      "prt_state_subagent",
       "prt_state_skill",
       "prt_state_custom",
     ] as const
     const regions = defineVisualRegions({
       prt_state_webfetch: toolRegion(regionIDs[0]),
       prt_state_websearch: toolRegion(regionIDs[1]),
-      prt_state_task: toolRegion(regionIDs[2]),
+      prt_state_subagent: toolRegion(regionIDs[2]),
       prt_state_skill: toolRegion(regionIDs[3]),
       prt_state_custom: toolRegion(regionIDs[4]),
     })
@@ -73,9 +79,9 @@ test.describe("timeline tool state stability", () => {
         [80, 240, 100, 360, 140][index],
       )
     }
-    for (const [index, id] of ["skill", "webfetch", "custom", "task", "websearch"].entries()) {
+    for (const [index, id] of ["skill", "webfetch", "custom", "subagent", "websearch"].entries()) {
       const key = id as (typeof ids)[number]
-      const metadata = key === "task" ? { sessionId: childID } : key === "websearch" ? { provider: "exa" } : {}
+      const metadata = key === "subagent" ? { sessionID: childID } : key === "websearch" ? { provider: "exa" } : {}
       const output = key === "websearch" ? "Result https://example.com/result" : "Completed"
       await timeline.send(
         partUpdated(toolPart(`prt_state_${key}`, names[key], "completed", inputs[key], { metadata, output })),
@@ -121,12 +127,12 @@ test.describe("timeline tool state stability", () => {
     const ids = ["prt_ctx_01_read", "prt_ctx_02_glob", "prt_ctx_03_grep", "prt_ctx_04_list"]
     const tools = ["read", "glob", "grep", "list"]
     const inputs = [
-      { filePath: "src/a.ts", offset: 0, limit: 120 },
+      { path: "src/a.ts", offset: 0, limit: 120 },
       { path: directory, pattern: "**/*.ts" },
       { path: directory, pattern: "stability", include: "*.ts" },
       { path: "src" },
     ]
-    const context = ids.map((id, index) => toolPart(id, tools[index]!, "pending", inputs[index]!))
+    const context = ids.map((id, index) => toolPart(id, tools[index]!, "streaming", inputs[index]!))
     const timeline = await setupTimeline(page, {
       messages: [
         userMessage(),

packages/app/e2e/performance/timeline/first-navigation-benchmark.spec.ts

@@ -11,7 +11,7 @@ import {
 } from "./timeline-test-helpers"
 import { waitForStableTimeline } from "./session-tab-switch-probe"
 
-const contentSelector = '[data-message-id], [data-component="prompt-input"]'
+const contentSelector = '[data-message-id], [data-component="composer-editor"]'
 const draftID = "draft_first_navigation"
 
 benchmark.describe("performance: first navigation paint", () => {
@@ -41,11 +41,42 @@ benchmark.describe("performance: first navigation paint", () => {
       href,
       destinationPath: href,
       sourceSelector: messageSelector(fixture.expected.sourceMessageIDs.at(-1)!),
-      destinationSelector: '[data-component="prompt-input"]',
+      destinationSelector: '[data-component="composer-editor"]',
       contentSelector,
       navigate: async () => {
         await page.locator(`[data-slot="titlebar-tabs"] a[href="${href}"]`).first().click()
-        await expect(page.locator('[data-component="prompt-input"]')).toBeVisible()
+        await expect(page.locator('[data-component="composer-editor"]')).toBeVisible()
+      },
+    })
+    report(result)
+    expect(result.summary.blankSamples).toBe(0)
+    expect(result.summary.unknownSamples).toBe(0)
+  })
+
+  benchmark("opens a session from the new session page without a blank frame", async ({ page, report }) => {
+    await mockStressTimeline(page)
+    await installTimelineSettings(page)
+    await installStressSessionTabs(page, { draftID })
+    await page.goto("/")
+
+    const draftHref = stressDraftHref(draftID)
+    const draftTab = page.locator(`[data-slot="titlebar-tabs"] a[href="${draftHref}"]`)
+    await expect(draftTab).toHaveCount(1)
+    await draftTab.click()
+    await expect(page.locator('[data-component="new-session"]')).toBeVisible()
+
+    const href = stressSessionHref(fixture.targetID)
+    const sessionTab = page.locator(`[data-slot="titlebar-tabs"] a[href="${href}"]`)
+    await expect(sessionTab).toHaveCount(1)
+    const result = await measureFirstNavigation(page, {
+      href,
+      destinationPath: href,
+      sourceSelector: '[data-component="new-session"]',
+      destinationSelector: messageSelector(fixture.expected.targetMessageIDs.at(-1)!),
+      contentSelector,
+      navigate: async () => {
+        await sessionTab.click()
+        await expectSessionTitle(page, fixture.expected.targetTitle)
       },
     })
     report(result)

packages/app/e2e/performance/timeline/session-timeline-benchmark.fixture.ts

@@ -4,6 +4,7 @@ import type { Page } from "@playwright/test"
 import { mockOpenCodeServer } from "../../utils/mock-server"
 import { expectAppVisible, expectSessionTitle } from "../../utils/waits"
 import { expect } from "../benchmark"
+import { createTwoFilesPatch } from "diff"
 
 const directory = "C:/OpenCode/TimelineStateRegression"
 const projectID = "proj_timeline_state_regression"
@@ -26,28 +27,29 @@ const userMessage = {
 
 const editPart: ToolSeed = {
   id: editPartID,
-  sessionID,
-  messageID: assistantMessageID,
   type: "tool",
-  callID: "call_edit_regression",
-  tool: "edit",
+  name: "edit",
   state: {
     status: "completed",
-    input: { filePath: "src/regression.ts" },
-    output: "Edited src/regression.ts",
-    title: "src/regression.ts",
+    input: {
+      path: "src/regression.ts",
+      oldString: "export const value = 'before'",
+      newString: "export const value = 'after'",
+    },
+    content: [{ type: "text", text: "Edited src/regression.ts" }],
     metadata: {
-      filediff: {
-        file: "src/regression.ts",
-        additions: 1,
-        deletions: 1,
-        before: "export const value = 'before'\n",
-        after: "export const value = 'after'\n",
-      },
-      diff: "diff --git a/src/regression.ts b/src/regression.ts\n-export const value = 'before'\n+export const value = 'after'\n",
+      files: [
+        currentFile(
+          "src/regression.ts",
+          "export const value = 'before'\n",
+          "export const value = 'after'\n",
+          1,
+          1,
+        ),
+      ],
     },
-    time: { start: 1700000001000, end: 1700000002000 },
   },
+  time: { created: 1700000001000, ran: 1700000001000, completed: 1700000002000 },
 }
 
 const assistantMessage = {
@@ -204,20 +206,20 @@ function performanceTurn(index: number) {
       ? [
           {
             id: `prt_0000_${suffix}_edit`,
-            sessionID,
-            messageID: assistantID,
             type: "tool",
-            callID: `call_0000_${suffix}_edit`,
-            tool: "edit",
+            name: "edit",
             state: {
               status: "completed",
-              input: { filePath: `src/history-${index}.ts` },
-              output: `Edited src/history-${index}.ts`,
-              title: `src/history-${index}.ts`,
+              input: { path: `src/history-${index}.ts`, oldString: before, newString: after },
+              content: [{ type: "text", text: `Edited src/history-${index}.ts` }],
               metadata: {
-                filediff: { file: `src/history-${index}.ts`, additions: 48, deletions: 48, before, after },
+                files: [currentFile(`src/history-${index}.ts`, before, after, 48, 48)],
               },
-              time: { start: 1690000001200 + index * 2_000, end: 1690000001400 + index * 2_000 },
+            },
+            time: {
+              created: 1690000001200 + index * 2_000,
+              ran: 1690000001200 + index * 2_000,
+              completed: 1690000001400 + index * 2_000,
             },
           },
         ]
@@ -226,20 +228,18 @@ function performanceTurn(index: number) {
       ? [
           {
             id: `prt_0000_${suffix}_write`,
-            sessionID,
-            messageID: assistantID,
             type: "tool",
-            callID: `call_0000_${suffix}_write`,
-            tool: "write",
+            name: "write",
             state: {
               status: "completed",
-              input: { filePath: `src/generated-${index}.tsx`, content: after },
-              output: `Wrote src/generated-${index}.tsx`,
-              title: `src/generated-${index}.tsx`,
-              metadata: {
-                filediff: { file: `src/generated-${index}.tsx`, additions: 32, deletions: 0, before: "", after },
-              },
-              time: { start: 1690000001400 + index * 2_000, end: 1690000001500 + index * 2_000 },
+              input: { path: `src/generated-${index}.tsx`, content: after },
+              content: [{ type: "text", text: `Wrote src/generated-${index}.tsx` }],
+              metadata: { files: [currentFile(`src/generated-${index}.tsx`, "", after, 32, 0)] },
+            },
+            time: {
+              created: 1690000001400 + index * 2_000,
+              ran: 1690000001400 + index * 2_000,
+              completed: 1690000001500 + index * 2_000,
             },
           },
         ]
@@ -248,31 +248,24 @@ function performanceTurn(index: number) {
       ? [
           {
             id: `prt_0000_${suffix}_patch`,
-            sessionID,
-            messageID: assistantID,
             type: "tool",
-            callID: `call_0000_${suffix}_patch`,
-            tool: "apply_patch",
+            name: "patch",
             state: {
               status: "completed",
               input: { patchText: realisticPatch(index) },
-              output: "Success. Updated src/components/SessionCard.tsx",
-              title: "src/components/SessionCard.tsx",
+              content: [{ type: "text", text: "Success. Updated src/components/SessionCard.tsx" }],
               metadata: {
                 files: [
                   {
-                    filePath: "src/components/SessionCard.tsx",
-                    relativePath: "src/components/SessionCard.tsx",
-                    type: "update",
-                    additions: 8,
-                    deletions: 3,
-                    patch: realisticPatch(index),
-                    before,
-                    after,
+                    ...currentFile("src/components/SessionCard.tsx", before, after, 8, 3),
                   },
                 ],
               },
-              time: { start: 1690000001500 + index * 2_000, end: 1690000001700 + index * 2_000 },
+            },
+            time: {
+              created: 1690000001500 + index * 2_000,
+              ran: 1690000001500 + index * 2_000,
+              completed: 1690000001700 + index * 2_000,
             },
           },
         ]
@@ -309,20 +302,16 @@ function performanceTurn(index: number) {
 }
 
 type ToolSeed = {
-  id?: string
-  sessionID?: string
-  messageID?: string
+  id: string
   type: "tool"
-  callID: string
-  tool: string
+  name: string
   state: {
-    status: string
+    status: "completed"
     input: Record<string, unknown>
-    output: string
-    title?: string
+    content: [{ type: "text"; text: string }]
     metadata: Record<string, unknown>
-    time: { start: number; end: number }
   }
+  time: { created: number; ran: number; completed: number }
 }
 
 type ContentSeedBase = { id?: string; sessionID?: string; messageID?: string }
@@ -335,13 +324,13 @@ type ContentSeed =
 function toolContent(part: ToolSeed): SessionMessageAssistant["content"][number] {
   return {
     type: "tool",
-    id: part.callID,
-    name: part.tool,
-    time: { created: part.state.time.start, ran: part.state.time.start, completed: part.state.time.end },
+    id: part.id,
+    name: part.name,
+    time: part.time,
     state: {
       status: "completed",
       input: part.state.input as Record<string, JsonValue>,
-      content: [{ type: "text", text: part.state.output }],
+      content: part.state.content,
       metadata: part.state.metadata as Record<string, JsonValue>,
     },
   }
@@ -422,6 +411,16 @@ export function MessageSummary(props: { messages: Message[]; locale: string }) {
 `
 }
 
+function currentFile(file: string, before: string, after: string, additions: number, deletions: number) {
+  return {
+    file,
+    patch: createTwoFilesPatch(`a/${file}`, `b/${file}`, before, after),
+    additions,
+    deletions,
+    status: before ? (after ? "modified" : "deleted") : "added",
+  }
+}
+
 function realisticPatch(index: number) {
   return `*** Begin Patch
 *** Update File: src/components/SessionCard.tsx

packages/app/e2e/performance/timeline/session-timeline-profile.ts

@@ -1,4 +1,6 @@
 import type { CDPSession, Page } from "@playwright/test"
+import path from "node:path"
+import { mkdir, writeFile } from "node:fs/promises"
 
 export async function startTimelineProfile(page: Page, options: { cpuThrottle: number; profileCPU: boolean }) {
   const cdp = await page.context().newCDPSession(page)
@@ -12,6 +14,13 @@ export async function startTimelineProfile(page: Page, options: { cpuThrottle: n
     async stop() {
       if (!options.profileCPU) return
       const result = await cdp.send("Profiler.stop")
+      const directory = process.env.TIMELINE_CPU_PROFILE_DIR
+      if (directory) {
+        await mkdir(directory, { recursive: true })
+        const file = path.join(directory, `${process.env.OPENCODE_PERFORMANCE_RUN_ID ?? "manual"}-timeline.cpuprofile`)
+        await writeFile(file, JSON.stringify(result.profile))
+        console.log("timeline cpu profile file", file)
+      }
       const self = new Map<number, number>()
       result.profile.samples?.forEach((id, index) => {
         const duration = (result.profile.timeDeltas?.[index] ?? 0) / 1_000

packages/app/e2e/performance/timeline/session-timeline-stress.fixture.ts

@@ -1,3 +1,5 @@
+import { createTwoFilesPatch } from "diff"
+
 const words = [
   "alpha",
   "bravo",
@@ -28,22 +30,21 @@ const directory = "C:/OpenCode/SmokeProject"
 const projectID = "proj_smoke_timeline"
 const model = { providerID: "opencode", modelID: "claude-opus-4-6", variant: "max" }
 
-type MessagePart = {
-  id: string
-  type: "text" | "reasoning" | "tool"
-  text?: string
-  time?: { start: number; end?: number }
-  callID?: string
-  tool?: string
-  state?: {
-    status: "completed"
-    input: Record<string, unknown>
-    output: string
-    title: unknown
-    metadata: Record<string, unknown>
-    time: { start: number; end: number }
-  }
-}
+type MessagePart =
+  | { id: string; type: "text"; text: string }
+  | { id: string; type: "reasoning"; text: string; time?: { start: number; end?: number } }
+  | {
+      id: string
+      type: "tool"
+      name: string
+      state: {
+        status: "completed"
+        input: Record<string, unknown>
+        content: [{ type: "text"; text: string }]
+        metadata: Record<string, unknown>
+      }
+      time: { created: number; ran: number; completed: number }
+    }
 
 function lorem(seed: number, length: number) {
   let out = ""
@@ -102,17 +103,16 @@ function messageContent(part: MessagePart): SessionMessageAssistant["content"][n
         ? { created: part.time.start, ...(part.time.end === undefined ? {} : { completed: part.time.end }) }
         : undefined,
     }
-  const state = part.state!
   return {
     type: "tool",
-    id: part.callID ?? part.id,
-    name: part.tool!,
-    time: { created: state.time.start, ran: state.time.start, completed: state.time.end },
+    id: part.id,
+    name: part.name,
+    time: part.time,
     state: {
       status: "completed",
-      input: state.input as Record<string, JsonValue>,
-      content: [{ type: "text", text: state.output }],
-      metadata: state.metadata as Record<string, JsonValue>,
+      input: part.state.input as Record<string, JsonValue>,
+      content: part.state.content,
+      metadata: part.state.metadata as Record<string, JsonValue>,
     },
   }
 }
@@ -149,43 +149,46 @@ function toolPart(
 ): MessagePart {
   const metadata =
     metadataOverride ??
-    (tool === "apply_patch"
-      ? { files: [patchFile(index, "update"), patchFile(index + 1, index % 2 === 0 ? "add" : "delete")] }
+    (tool === "patch"
+      ? {
+          files: [
+            patchFile(index, "modified"),
+            patchFile(index + 1, index % 2 === 0 ? "added" : "deleted"),
+          ],
+        }
       : tool === "edit" || tool === "write"
-        ? {
-            filediff: fileDiff(String(input.filePath ?? `src/generated/file-${index}.ts`), index),
-            diff: patch(index, outputLength),
-            preview: patch(index + 1, 420),
-          }
+        ? { files: [fileDiff(String(input.path ?? `src/generated/file-${index}.ts`), index)] }
         : tool === "question"
           ? { answers: [["Proceed"], ["Keep sample output"]] }
           : {})
   return {
-    id: id(`prt_tool_${tool}_${partIndex}`, index),
+    id: id(`call_${tool}_${partIndex}`, index),
     type: "tool",
-    callID: id("call", index * 10 + partIndex),
-    tool,
+    name: tool,
     state: {
       status: "completed",
       input,
-      output: lorem(index * 23 + partIndex, outputLength),
-      title: tool === "bash" ? "Verify generated output" : input.filePath || input.path || input.pattern || "completed",
+      content: [{ type: "text", text: lorem(index * 23 + partIndex, outputLength) }],
       metadata,
-      time: { start: 1700000000000 + index * 10_000, end: 1700000000000 + index * 10_000 + 400 },
+    },
+    time: {
+      created: 1700000000000 + index * 10_000,
+      ran: 1700000000000 + index * 10_000,
+      completed: 1700000000000 + index * 10_000 + 400,
     },
   }
 }
 
-function patchFile(seed: number, type: "add" | "update" | "delete") {
+function patchFile(seed: number, status: "added" | "modified" | "deleted") {
+  const file = `src/generated/patch-${seed}.ts`
+  const before = status === "added" ? "" : code(seed, 18)
+  const after = status === "deleted" ? "" : code(seed + 1, 24)
   return {
-    filePath: `src/generated/patch-${seed}.ts`,
-    relativePath: `src/generated/patch-${seed}.ts`,
-    type,
-    additions: (seed % 7) + 1,
-    deletions: type === "add" ? 0 : seed % 4,
-    patch: patch(seed, 520),
-    before: type === "add" ? undefined : code(seed, 18),
-    after: type === "delete" ? undefined : code(seed + 1, 24),
+    file,
+    status,
+    additions: status === "deleted" ? 0 : (seed % 7) + 1,
+    deletions: status === "added" ? 0 : seed % 4,
+    patch: createTwoFilesPatch(`a/${file}`, `b/${file}`, before, after),
   }
 }
 
@@ -200,17 +203,13 @@ function fileDiff(file: string, seed: number) {
         : before.replace("value4", "updatedValue4").replace("value20", "updatedValue20")
   return {
     file,
+    status: "modified" as const,
     additions: lines === 300 ? 300 : lines === 2 ? 1 : 2,
     deletions: lines === 300 ? 300 : lines === 2 ? 1 : 2,
-    before,
-    after,
+    patch: createTwoFilesPatch(`a/${file}`, `b/${file}`, before, after),
   }
 }
 
-function patch(seed: number, length: number) {
-  return `diff --git a/src/generated/file-${seed}.ts b/src/generated/file-${seed}.ts\n+${lorem(seed, length).replace(/\n/g, "\n+")}`
-}
-
 function code(seed: number, lines: number, width = 32) {
   return Array.from(
     { length: lines },
@@ -225,22 +224,24 @@ function turn(index: number): SessionMessageInfo[] {
     ...(index % 5 === 0 ? [reasoningPart(index, 0, 420)] : []),
     ...(index % 3 === 0
       ? [
-          toolPart(index, 0, "read", { filePath: `src/generated/file-${index}.ts`, offset: 0, limit: 80 }, 220),
+          toolPart(index, 0, "read", { path: `src/generated/file-${index}.ts`, offset: 0, limit: 80 }, 220),
           toolPart(index, 5, "glob", { path: directory, pattern: `**/*sample-${index}*.ts` }, 140),
           toolPart(index, 1, "grep", { path: directory, pattern: `sample-${index}`, include: "*.ts" }, 180),
           toolPart(index, 6, "list", { path: `src/generated/${index}` }, 120),
         ]
       : []),
     textPart(index, 2, 160 + (index % 6) * 90),
-    ...(index % 4 === 0 ? [toolPart(index, 3, "edit", { filePath: `src/generated/file-${index}.ts` }, 700)] : []),
+    ...(index % 4 === 0
+      ? [toolPart(index, 3, "edit", { path: `src/generated/file-${index}.ts`, oldString: "before", newString: "after" }, 700)]
+      : []),
     ...(index % 6 === 0
-      ? [toolPart(index, 7, "write", { filePath: `src/generated/write-${index}.ts`, content: code(index, 28) }, 560)]
+      ? [toolPart(index, 7, "write", { path: `src/generated/write-${index}.ts`, content: code(index, 28) }, 560)]
       : []),
     ...(index % 8 === 0
-      ? [toolPart(index, 8, "apply_patch", { files: [`src/generated/patch-${index}.ts`] }, 620)]
+      ? [toolPart(index, 8, "patch", { patchText: `Update generated patch ${index}` }, 620)]
       : []),
     ...(index % 7 === 0
-      ? [toolPart(index, 4, "bash", { command: "bun typecheck", description: "Verify generated output" }, 620)]
+      ? [toolPart(index, 4, "shell", { command: "bun typecheck", description: "Verify generated output" }, 620)]
       : []),
     ...(index % 10 === 0 ? [toolPart(index, 9, "webfetch", { url: "https://example.com/docs/sample" }, 120)] : []),
     ...(index % 11 === 0 ? [toolPart(index, 10, "websearch", { query: "sample movement notes" }, 240)] : []),
@@ -256,7 +257,15 @@ function turn(index: number): SessionMessageInfo[] {
         ]
       : []),
     ...(index % 17 === 0
-      ? [toolPart(index, 12, "task", { description: "Inspect generated fixture", subagent_type: "explore" }, 160)]
+      ? [
+          toolPart(
+            index,
+            12,
+            "subagent",
+            { description: "Inspect generated fixture", agent: "explore", prompt: "Inspect the fixture." },
+            160,
+          ),
+        ]
       : []),
   ]
   return [user, assistantMessage(targetID, index, user.id, parts)]
@@ -272,10 +281,10 @@ const sourceMessages = Array.from({ length: 12 }, (_, index) => [
           toolPart(
             index + 1000,
             1,
-            "task",
-            { description: "Inspect child navigation", subagent_type: "explore" },
+            "subagent",
+            { description: "Inspect child navigation", agent: "explore", prompt: "Inspect child navigation." },
             160,
-            { sessionId: childID },
+            { sessionID: childID },
           ),
         ]
       : []),

packages/app/e2e/performance/unit/desktop-startup.test.ts

@@ -0,0 +1,88 @@
+import { describe, expect, test } from "bun:test"
+import { inlineThemePreload } from "../../../vite.js"
+import { milestoneForLine, summarizeDesktopStartup, type DesktopStartupSample } from "../devex/desktop-startup"
+
+describe("desktop startup benchmark", () => {
+  test.each(["/oc-theme-preload.js", "./oc-theme-preload.js"])("inlines %s before the renderer runs", (path) => {
+    const html = inlineThemePreload(`<script id="oc-theme-preload-script" src="${path}"></script>`)
+    expect(html).not.toContain(" src=")
+    expect(html).toContain("opencode-color-scheme")
+  })
+
+  test("recognizes startup milestones in colored output", () => {
+    const cases = [
+      ["bunRootScript", "$ bun --cwd packages/desktop dev"],
+      ["bunDesktopScript", "$ bun ./scripts/dev.ts"],
+      ["desktopPrepared", "Copied dev icons from"],
+      ["mainBundleReady", "electron main process built successfully"],
+      ["preloadBundleReady", "electron preload scripts built successfully"],
+      ["rendererDevServerReady", "dev server running for the electron renderer process at:"],
+      ["electronSpawnStarted", "starting electron app..."],
+      ["debugEndpointReady", "DevTools listening on ws://"],
+      ["electronStarted", "app starting"],
+      ["serviceEnsureStarted", "starting v2 background service"],
+      ["serviceSpawnRequested", "v2 CLI background service starting"],
+      ["serviceReady", "v2 CLI background service ready"],
+      ["backgroundLoadingReady", "loading task finished"],
+      ["rendererViteConnected", "[vite] connected."],
+      ["rendererInitializationStarted", "awaiting server ready"],
+      ["rendererInitializationReady", "server ready"],
+      ["windowVisible", "main window visible"],
+    ] as const
+    cases.forEach(([milestone, line]) => {
+      expect(milestoneForLine(`\u001b[32m${line}\u001b[39m`)).toBe(milestone)
+    })
+    expect(milestoneForLine("12:30:00.000 › v2 CLI background service ready {")).toBe("serviceReady")
+    expect(milestoneForLine("unrelated output")).toBeUndefined()
+  })
+
+  test("keeps raw samples and reports median absolute deviation", () => {
+    const samples = [24, 20, 22, 28, 26].map((commandToHomeReadyMs, index) => sample(index + 1, commandToHomeReadyMs))
+    expect(summarizeDesktopStartup(samples).commandToHomeReadyMs).toEqual({
+      min: 20,
+      median: 24,
+      max: 28,
+      medianAbsoluteDeviation: 2,
+    })
+  })
+})
+
+function sample(run: number, commandToHomeReadyMs: number): DesktopStartupSample {
+  const milestonesMs = {
+    bunRootScript: 1,
+    bunDesktopScript: 2,
+    desktopPrepared: 3,
+    mainBundleReady: 4,
+    preloadBundleReady: 5,
+    rendererDevServerReady: 6,
+    electronSpawnStarted: 7,
+    debugEndpointReady: 8,
+    electronStarted: 9,
+    serviceEnsureStarted: 10,
+    serviceSpawnRequested: 11,
+    serviceReady: 12,
+    backgroundLoadingReady: 13,
+    rendererViteConnected: 14,
+    rendererInitializationStarted: 15,
+    rendererInitializationReady: 16,
+    windowVisible: 17,
+    homeReady: commandToHomeReadyMs,
+  }
+  return {
+    run,
+    commandToHomeReadyMs,
+    milestonesMs,
+    phasesMs: {
+      desktopPreparation: 3,
+      viteMainBundle: 1,
+      vitePreloadBundle: 1,
+      rendererServerStartup: 1,
+      electronStartup: 2,
+      serviceSpawnWait: 1,
+      serviceProcessStartup: 1,
+      rendererStartup: commandToHomeReadyMs - 14,
+      visibleWindowToHome: commandToHomeReadyMs - 17,
+    },
+    service: { version: "2.0.0-local-test", url: "http://127.0.0.1:3000", pid: run },
+  }
+}