refactor(app): establish v2 session controller

.opencode/plugins/tui/discovery-smoke.ts

@@ -2,12 +2,15 @@ import type { Context } from "../../../packages/plugin/src/tui/context"
 
 export default {
   id: "test.tui-discovery-smoke",
-  setup(_context: Context) {
+  setup(context: Context) {
-    // context.ui.toast.show({
+    const timer = setTimeout(() => {
-    //   title: "TUI plugin discovery works",
+      context.ui.toast.show({
-    //   message: "Loaded .opencode/plugins/tui/discovery-smoke.ts",
+        title: "TUI plugin discovery works",
-    //   variant: "success",
+        message: "Loaded .opencode/plugins/tui/discovery-smoke.ts",
-    //   duration: 30_000,
+        variant: "success",
-    // })
+        duration: 30_000,
+      })
+    }, 1_000)
+    return () => clearTimeout(timer)
   },
 }

bun.lock

@@ -67,7 +67,6 @@
         "@opencode-ai/sdk": "file:vendor/opencode-ai-sdk-1.18.8-dev.tgz",
         "@opencode-ai/session-ui": "workspace:*",
         "@opencode-ai/ui": "workspace:*",
-        "@opencode-ai/util": "workspace:*",
         "@pierre/trees": "1.0.0-beta.4",
         "@sentry/solid": "catalog:",
         "@shikijs/transformers": "3.9.2",
@@ -88,6 +87,7 @@
         "@tanstack/solid-query": "5.91.4",
         "@tanstack/solid-virtual": "catalog:",
         "@thisbeyond/solid-dnd": "0.7.5",
+        "diff": "catalog:",
         "effect": "catalog:",
         "fuzzysort": "catalog:",
         "ghostty-web": "github:anomalyco/ghostty-web#83c0a07b8628b748aed073b232cb4b52a6ca11c1",
@@ -143,10 +143,7 @@
         "open": "10.1.2",
         "semver": "catalog:",
         "solid-js": "catalog:",
-        "tree-sitter-bash": "0.25.0",
-        "tree-sitter-powershell": "0.25.10",
         "uqr": "0.1.3",
-        "web-tree-sitter": "0.25.10",
         "ws": "8.21.0",
       },
       "devDependencies": {
@@ -383,6 +380,7 @@
         "@opencode-ai/ai": "workspace:*",
         "@opencode-ai/codemode": "workspace:*",
         "@opencode-ai/effect-drizzle-sqlite": "workspace:*",
+        "@opencode-ai/effect-sqlite-node": "workspace:*",
         "@opencode-ai/plugin": "workspace:*",
         "@opencode-ai/schema": "workspace:*",
         "@opencode-ai/util": "workspace:*",
@@ -403,11 +401,9 @@
         "ignore": "7.0.5",
         "immer": "11.1.4",
         "jsonc-parser": "3.3.1",
-        "tree-sitter-bash": "0.25.0",
+        "semver": "^7.6.3",
-        "tree-sitter-powershell": "0.25.10",
         "turndown": "7.2.0",
         "venice-ai-sdk-provider": "2.1.1",
-        "web-tree-sitter": "0.25.10",
         "which": "6.0.1",
         "zod": "catalog:",
       },
@@ -424,6 +420,7 @@
         "@tsconfig/bun": "catalog:",
         "@types/bun": "catalog:",
         "@types/node": "catalog:",
+        "@types/semver": "catalog:",
         "@types/turndown": "5.0.5",
         "@types/which": "3.0.4",
         "drizzle-kit": "catalog:",
@@ -517,7 +514,6 @@
         "@opencode-ai/core": "workspace:*",
         "@opencode-ai/session-ui": "workspace:*",
         "@opencode-ai/ui": "workspace:*",
-        "@opencode-ai/util": "workspace:*",
         "@pierre/diffs": "catalog:",
         "@solidjs/meta": "catalog:",
         "@solidjs/router": "catalog:",
@@ -604,7 +600,6 @@
         "zod": "catalog:",
       },
       "devDependencies": {
-        "@opencode-ai/theme": "workspace:*",
         "@opentui/core": "catalog:",
         "@opentui/keymap": "catalog:",
         "@opentui/solid": "catalog:",
@@ -616,14 +611,12 @@
         "typescript": "catalog:",
       },
       "peerDependencies": {
-        "@opencode-ai/theme": "workspace:*",
         "@opentui/core": ">=0.4.5",
         "@opentui/keymap": ">=0.4.5",
         "@opentui/solid": ">=0.4.5",
         "solid-js": ">=1.9.0",
       },
       "optionalPeers": [
-        "@opencode-ai/theme",
         "@opentui/core",
         "@opentui/keymap",
         "@opentui/solid",
@@ -1077,11 +1070,9 @@
   },
   "trustedDependencies": [
     "esbuild",
-    "tree-sitter-powershell",
     "protobufjs",
     "electron",
     "web-tree-sitter",
-    "tree-sitter-bash",
   ],
   "patchedDependencies": {
     "@pierre/trees@1.0.0-beta.4": "patches/@pierre%2Ftrees@1.0.0-beta.4.patch",
@@ -4078,10 +4069,10 @@
 
     "fetch-blob": ["fetch-blob@3.2.0", "", { "dependencies": { "node-domexception": "^1.0.0", "web-streams-polyfill": "^3.0.3" } }, "sha512-7yAQpD2UMJzLi1Dqv7qFYnPbaPx7ZfFK6PiIxQ4PfkGPyNyl2Ugx+a/umUonmKqjhM4DnfbMvdX6otXq83soQQ=="],
 
-    "ffi-rs": ["ffi-rs@1.3.2", "", { "optionalDependencies": { "@yuuang/ffi-rs-android-arm64": "1.3.2", "@yuuang/ffi-rs-darwin-arm64": "1.3.2", "@yuuang/ffi-rs-darwin-x64": "1.3.2", "@yuuang/ffi-rs-linux-arm-gnueabihf": "1.3.2", "@yuuang/ffi-rs-linux-arm64-gnu": "1.3.2", "@yuuang/ffi-rs-linux-arm64-musl": "1.3.2", "@yuuang/ffi-rs-linux-x64-gnu": "1.3.2", "@yuuang/ffi-rs-linux-x64-musl": "1.3.2", "@yuuang/ffi-rs-win32-arm64-msvc": "1.3.2", "@yuuang/ffi-rs-win32-ia32-msvc": "1.3.2", "@yuuang/ffi-rs-win32-x64-msvc": "1.3.2" } }, "sha512-4s8dX9VbBw/jd5NOuE3EJRqXaIVdjMyiumeeDzrOhtjQRwp6Bz2za7iksWXTnvTQKV/tTdm1s1w7mObe92zPjQ=="],
-
     "file-uri-to-path": ["file-uri-to-path@1.0.0", "", {}, "sha512-0Zt+s3L7Vf1biwWZ29aARiVYLx7iMGnEUl9x33fbB/j3jR81u/O2LbqK+Bm1CDSNDKVtJ/YjwY7TUd5SkeLQLw=="],
 
+    "ffi-rs": ["ffi-rs@1.3.2", "", { "optionalDependencies": { "@yuuang/ffi-rs-android-arm64": "1.3.2", "@yuuang/ffi-rs-darwin-arm64": "1.3.2", "@yuuang/ffi-rs-darwin-x64": "1.3.2", "@yuuang/ffi-rs-linux-arm-gnueabihf": "1.3.2", "@yuuang/ffi-rs-linux-arm64-gnu": "1.3.2", "@yuuang/ffi-rs-linux-arm64-musl": "1.3.2", "@yuuang/ffi-rs-linux-x64-gnu": "1.3.2", "@yuuang/ffi-rs-linux-x64-musl": "1.3.2", "@yuuang/ffi-rs-win32-arm64-msvc": "1.3.2", "@yuuang/ffi-rs-win32-ia32-msvc": "1.3.2", "@yuuang/ffi-rs-win32-x64-msvc": "1.3.2" } }, "sha512-4s8dX9VbBw/jd5NOuE3EJRqXaIVdjMyiumeeDzrOhtjQRwp6Bz2za7iksWXTnvTQKV/tTdm1s1w7mObe92zPjQ=="],
+
     "filelist": ["filelist@1.0.6", "", { "dependencies": { "minimatch": "^5.0.1" } }, "sha512-5giy2PkLYY1cP39p17Ech+2xlpTRL9HLspOfEgm0L6CwBXBTgsK5ou0JtzYuepxkaQ/tvhCFIJ5uXo0OrM2DxA=="],
 
     "fill-range": ["fill-range@7.1.1", "", { "dependencies": { "to-regex-range": "^5.0.1" } }, "sha512-YsGpe3WHLK8ZYi4tWDg2Jy3ebRz2rXowDxnld4bkQB00cc/1Zw9AWnC0i9ztDJitivtQvaI9KaLyKrc+hBW0yg=="],
@@ -5640,10 +5631,6 @@
 
     "traverse": ["traverse@0.3.9", "", {}, "sha512-iawgk0hLP3SxGKDfnDJf8wTz4p2qImnyihM5Hh/sGvQ3K37dPi/w8sRhdNIxYA1TwFwc5mDhIJq+O0RsvXBKdQ=="],
 
-    "tree-sitter-bash": ["tree-sitter-bash@0.25.0", "", { "dependencies": { "node-addon-api": "^8.2.1", "node-gyp-build": "^4.8.2" }, "peerDependencies": { "tree-sitter": "^0.25.0" }, "optionalPeers": ["tree-sitter"] }, "sha512-gZtlj9+qFS81qKxpLfD6H0UssQ3QBc/F0nKkPsiFDyfQF2YBqYvglFJUzchrPpVhZe9kLZTrJ9n2J6lmka69Vg=="],
-
-    "tree-sitter-powershell": ["tree-sitter-powershell@0.25.10", "", { "dependencies": { "node-addon-api": "^7.1.0", "node-gyp-build": "^4.8.0" }, "peerDependencies": { "tree-sitter": "^0.25.0" }, "optionalPeers": ["tree-sitter"] }, "sha512-bEt8QoySpGFnU3aa8WedQyNMaN6aTwy/WUbvIVt0JSKF+BbJoSHNHu+wCbhj7xLMsfB0AuffmiJm+B8gzva8Lg=="],
-
     "treeverse": ["treeverse@3.0.0", "", {}, "sha512-gcANaAnd2QDZFmHFEOF4k7uc1J/6a6z3DJMd/QwEyxLoKGiptJRwid582r7QIsFlFMIZ3SnxfS52S4hm2DHkuQ=="],
 
     "trim-lines": ["trim-lines@3.0.1", "", {}, "sha512-kRj8B+YHZCc9kQYdWfJB2/oUl9rA99qbowYYBtr4ui4mZyAQ2JpvVBd/6U2YloATfqBhBTSMhTpgBHtU0Mf3Rg=="],
@@ -6866,8 +6853,6 @@
 
     "topojson-client/commander": ["commander@2.20.3", "", {}, "sha512-GpVkmM8vF2vQUkj2LvZmD35JxeJOLCwJ9cUkugyk2nuhbv3+mJvpLYYt+0+USMxE+oj+ey/lJEnhZw75x/OMcQ=="],
 
-    "tree-sitter-bash/node-addon-api": ["node-addon-api@8.9.0", "", {}, "sha512-ekZMeaaIzSQTSpr7X2X3iJM7lTzgnx8ahAG9pJfT/7+14mlEM8ZYQ9cgCDvSSRbReFK0oHli3WrZdCiRsgAT9Q=="],
-
     "tw-to-css/postcss": ["postcss@8.4.31", "", { "dependencies": { "nanoid": "^3.3.6", "picocolors": "^1.0.0", "source-map-js": "^1.0.2" } }, "sha512-PS08Iboia9mts/2ygV3eLpY5ghnUcfLV/EXTOW1E2qYxJKGGBUtNjN76FYHnMs36RmARn41bC0AZmn+rR0OVpQ=="],
 
     "tw-to-css/tailwindcss": ["tailwindcss@3.3.2", "", { "dependencies": { "@alloc/quick-lru": "^5.2.0", "arg": "^5.0.2", "chokidar": "^3.5.3", "didyoumean": "^1.2.2", "dlv": "^1.1.3", "fast-glob": "^3.2.12", "glob-parent": "^6.0.2", "is-glob": "^4.0.3", "jiti": "^1.18.2", "lilconfig": "^2.1.0", "micromatch": "^4.0.5", "normalize-path": "^3.0.0", "object-hash": "^3.0.0", "picocolors": "^1.0.0", "postcss": "^8.4.23", "postcss-import": "^15.1.0", "postcss-js": "^4.0.1", "postcss-load-config": "^4.0.1", "postcss-nested": "^6.0.1", "postcss-selector-parser": "^6.0.11", "postcss-value-parser": "^4.2.0", "resolve": "^1.22.2", "sucrase": "^3.32.0" }, "bin": { "tailwind": "lib/cli.js", "tailwindcss": "lib/cli.js" } }, "sha512-9jPkMiIBXvPc2KywkraqsUfbfj+dHDb+JPWtSJa9MLFdrPyazI7q6WX2sUrm7R9eVR7qqv3Pas7EvQFzxKnI6w=="],

docs/design/v1-v2-database-migration.md

@@ -1,118 +0,0 @@
-# V1 to V2 Database Migration
-
-## Approach
-
-- Use the `dev` branch database schema and migration registry as the V1 baseline.
-- Remove migrations that exist only on the V2 branch.
-- Generate one canonical migration from the `dev` schema to the final V2 schema.
-- Add explicit data operations to that migration where generated DDL is insufficient.
-- Test the migration against a populated database at the exact `dev` schema.
-
-## Preserve
-
-The canonical V1 data remains in its existing tables. In particular, preserve `session`, `message`, and `part` rows.
-
-Preserve `workspace` rows and existing `session.workspace_id` values unchanged. The migration must not clear or rebuild
-workspace relationships.
-
-Keep the `todo` table and its data unchanged. V2 does not currently migrate todos into another representation, and the
-generated migration must not drop the table.
-
-## Truncate
-
-Truncate these pre-launch V2 tables before applying schema changes:
-
-- `event`
-- `event_sequence`
-- `session_message`
-
-These rows are not canonical V1 data. Truncating `event` before adding the required `event.created` column means the
-column needs neither a backfill nor a default. After truncation, rebuild `session_message` from canonical V1 `message`
-and `part` rows rather than retaining its pre-launch V2 contents.
-
-## Message Backfill
-
-Backfill canonical V1 history from `message` and `part` into `session_message`. This is the main data transformation in
-the migration. Preserving the V1 tables alone keeps the data safe but does not make existing history visible through the
-V2 session APIs, which read `session_message`.
-
-Reuse each V1 `message.id` as the corresponding `session_message.id`. Stable IDs keep the migration deterministic and
-avoid rewriting other persisted state that may refer to a message.
-
-Within each session, order V1 messages by `time_created` and then `id`, matching the existing V1 message index. Assign
-contiguous `session_message.seq` values starting at `0`.
-
-Map ordinary V1 messages one-to-one by role. Each ordinary V1 user message becomes one V2 `user` row, and each ordinary
-V1 assistant message becomes one V2 `assistant` row. Fold the source message's ordered V1 parts into that row's V2
-payload.
-
-Handle semantic marker parts before applying the ordinary mapping. In particular, a V1 user message containing a
-`compaction` part and its paired assistant summary represent one compaction operation, not two ordinary messages. Special
-part mappings must be decided explicitly before implementing the backfill.
-
-V1 synthetic content is represented by user text parts with `synthetic: true`, not by a separate message role. A V1 user
-message whose visible text parts are all synthetic should become a V2 `synthetic` message. If a V1 user message mixes
-ordinary and synthetic content, preserve the ordinary content in the V2 `user` row and emit the synthetic content as an
-adjacent V2 `synthetic` row. Ignore text parts marked `ignored`, matching V1 model-history behavior.
-
-Use the V1 compaction user message ID as the ID of the collapsed V2 compaction message. This matches V2's use of the
-admitted compaction input ID and preserves references to the initiating message.
-
-For a completed compaction, create one V2 `compaction` row with `status: "completed"`. Set `reason` from the V1
-compaction part's `auto` flag, join the paired summary assistant's nonempty text parts with blank lines for `summary`, and
-serialize the retained V1 tail beginning at `tail_start_id` for `recent`. Use an empty `recent` value when no tail was
-retained, and use the compaction user message creation time. Do not emit the paired summary assistant as a separate V2
-assistant row.
-
-After rebuilding `session_message`, seed `event_sequence` with one row per migrated session. Set its watermark to that
-session's maximum backfilled `session_message.seq`. This prevents new V2 events from reusing sequence numbers or sorting
-before migrated history. The `event` table remains empty.
-
-## Drop
-
-Drop these pre-launch V2 tables without preserving or transforming their rows:
-
-- `session_input`
-- `session_context_epoch`
-
-Do not transfer `session_input` rows into `session_pending`.
-
-## Create Empty
-
-Let the generated migration create these tables empty:
-
-- `instruction_blob`
-- `instruction_entry`
-- `instruction_state`
-- `session_pending`
-- `kv`
-
-V1 has no canonical data to backfill into these tables. V2 initializes their state as it runs.
-
-## Fork Storage
-
-V1 has no fork-boundary state to backfill. New V2 forks use a required message boundary and persist it in
-`session.fork_boundary`. The durable fork event contains no parent sequence. Its resolved boundary is one of:
-
-- `before`: copy messages before the identified message.
-- `through`: copy messages through the identified message.
-
-Forking an empty session is not supported. `session.fork_seq` and `session.fork_message_id` are not part of the final V2
-schema.
-
-New nullable session columns, including `fork_session_id`, `fork_boundary`, and `time_suspended`, require no explicit
-backfill. Existing rows naturally receive `NULL` when the generated migration adds the columns.
-
-## Verification
-
-The canonical migration test should seed representative V1 sessions, messages, parts, todos, projects, accounts,
-credentials, permissions, shares, and workspaces. After migration, it should verify:
-
-- Preserved rows and encoded values remain unchanged.
-- Todo rows remain available in the unchanged `todo` table.
-- `event` is empty, and stale pre-launch rows are absent from the rebuilt projections.
-- Backfilled `session_message` rows represent the canonical V1 `message` and `part` history.
-- Each migrated session's `event_sequence` watermark matches its maximum backfilled message sequence.
-- Dropped tables no longer exist.
-- New tables exist and are empty.
-- The final schema has no ungenerated changes.

packages/ai/AGENTS.md

@@ -12,8 +12,6 @@
 
 Per-type constructors live on the type, not as top-level re-exports. Use `Message.system(...)`, `Message.user(...)`, `Message.assistant(...)`, `Message.tool(...)`, `Model.make(...)`, `ToolDefinition.make(...)`, `ToolCallPart.make(...)`, `ToolResultPart.make(...)`, `ToolChoice.make(...)`, `ToolChoice.named(...)`, `SystemPart.make(...)`, and `GenerationOptions.make(...)` directly. The top-level `LLM` namespace is reserved for request-shaped call APIs: `LLM.request`, `LLM.generate`, `LLM.stream`, `LLM.updateRequest`, and `LLM.generateObject`. Two ways to construct the same thing is one too many.
 
-- Keep provider-defined string enums forward-compatible. Expose known values for autocomplete while accepting future values with `Known | (string & {})`; use `Schema.String` at runtime unless rejecting unknown values is required for correctness.
-
 ## Tests
 
 - Use `testEffect(...)` from `test/lib/effect.ts` for tests requiring Effect layers.
@@ -48,7 +46,7 @@ const response = yield * LLMClient.generate(request)
 
 `LLM.request(...)` builds an `LLMRequest`. `LLMClient.generate(...)` reads the executable route carried by `request.model.route`, builds the provider-native body, asks the route's transport for a real `HttpClientRequest.HttpClientRequest`, sends it through `RequestExecutor.Service`, parses the provider stream into common `LLMEvent`s, and finally returns an `LLMResponse`.
 
-Use `LLMClient.stream(request)` when callers want incremental `LLMEvent`s. Use `LLMClient.generate(request)` when callers want those same events collected into an `LLMResponse`.
+Use `LLMClient.stream(request)` when callers want incremental `LLMEvent`s. Use `LLMClient.generate(request)` when callers want those same events collected into an `LLMResponse`. Use `LLMClient.prepare<Body>(request)` to compile a request through the route pipeline without sending it — the optional `Body` type argument narrows `.body` to the route's native shape (e.g. `prepare<OpenAIChatBody>(...)` returns a `PreparedRequestOf<OpenAIChatBody>`). The runtime body is identical; the generic is a type-level assertion.
 
 Filter or narrow `LLMEvent` streams with `LLMEvent.is.*` (camelCase guards, e.g. `events.filter(LLMEvent.is.toolCall)`). The kebab-case `LLMEvent.guards["tool-call"]` form also works but prefer `is.*` in new code.
 
@@ -140,13 +138,13 @@ packages/ai/src/
     ids.ts                  branded IDs, literal types, ProviderMetadata
     options.ts              Generation/Provider/Http options, Limits, Model, cache policy
     messages.ts             content parts, Message, ToolDefinition, LLMRequest
-    events.ts               Usage, individual events, LLMEvent, LLMResponse
+    events.ts               Usage, individual events, LLMEvent, PreparedRequest, LLMResponse
     errors.ts               error reasons, LLMError, ToolFailure
     index.ts                barrel
   llm.ts                    request constructors and convenience helpers
   route/
     index.ts                @opencode-ai/ai/route advanced barrel
-    client.ts               Route.make + LLMClient.stream/generate
+    client.ts               Route.make + LLMClient.prepare/stream/generate
     executor.ts             RequestExecutor service + transport error mapping
     protocol.ts             Protocol type + Protocol.make
     endpoint.ts             Endpoint type + Endpoint.path

packages/ai/README.md

@@ -196,6 +196,7 @@ The hosted result is represented as a provider-executed tool call and tool resul
 - **`LLM.generate` / `LLM.stream`** — re-exported from `LLMClient` for one-import use.
 - **`Message.user(...)` / `Message.assistant(...)` / `Message.tool(...)`** — message constructors from the canonical schema model.
 - **`Model.make(...)` / `ToolCallPart.make(...)` / `ToolResultPart.make(...)` / `ToolDefinition.make(...)`** — model and tool-related constructors from the canonical schema model.
+- **`LLMClient.prepare(request)`** — compile a request through protocol body construction, validation, and HTTP preparation without sending. Useful for inspection and testing.
 - **`LLMEvent.is.*`** — typed guards (`is.textDelta`, `is.toolCall`, `is.finish`, …) for filtering streams.
 - **`Image.generate({...})`** — generate images through a provider-neutral image request and response model.
 - **`ImageClient`** — Effect service and layer for image execution, parallel to `LLMClient`.

packages/ai/example/call-sites.md

@@ -568,7 +568,7 @@ App boundary = explicit durable-config -> typed-provider call
       calling `.model(...)`.
 - [x] Remove request-shaping defaults from `Model`; selected models now carry only
       id, provider, and configured route while defaults live on routes or requests.
-- [x] Rework `LLMClient.stream` / `generate` to read
+- [x] Rework `LLMClient.prepare` / `stream` / `generate` to read
       `request.model.route` directly instead of calling `registeredRoute(...)`.
 - [x] Remove `Route.make(...)` global registration from the normal execution
       path; keep route ids only as diagnostics/provider API labels.

packages/ai/example/tutorial.ts

@@ -50,6 +50,18 @@ const request = LLM.request({
   },
 })
 
+// `http` is intentionally not needed for normal calls. This shows the shape for
+// newly released provider fields before they deserve a typed provider option.
+const rawOverlayExample = LLM.request({
+  model,
+  prompt: "Show the final HTTP overlay shape.",
+  http: {
+    body: { metadata: { example: "tutorial" } },
+    headers: { "x-opencode-tutorial": "1" },
+    query: { debug: "1" },
+  },
+})
+
 // 3. `generate` sends the request and collects the event stream into one
 // response object. `response.text` is the collected text output.
 const generateOnce = Effect.gen(function* () {
@@ -210,15 +222,33 @@ const FakeEcho = {
   }),
 }
 
+// `LLMClient.prepare` is the lower-level inspection hook: it compiles through
+// body conversion, validation, endpoint, auth, and HTTP construction without
+// sending anything over the network.
+const inspectFakeProvider = Effect.gen(function* () {
+  const prepared = yield* LLMClient.prepare(
+    LLM.request({
+      model: FakeEcho.configure().model("tiny-echo"),
+      prompt: "Show me the provider pipeline.",
+    }),
+  )
+
+  console.log("\n== fake provider prepare ==")
+  console.log("route:", prepared.route)
+  console.log("body:", Formatter.formatJson(prepared.body, { space: 2 }))
+})
+
 // Provide the LLM runtime and the HTTP request executor once. Keep one path
-// enabled at a time so the tutorial can demonstrate generate, stream, or
+// enabled at a time so the tutorial can demonstrate generate, prepare, stream,
-// tool-loop behavior without spending tokens on every example.
+// or tool-loop behavior without spending tokens on every example.
 const requestExecutorLayer = RequestExecutor.fetchLayer
 const llmDeps = Layer.mergeAll(requestExecutorLayer, WebSocketExecutor.layer)
 const llmClientLayer = LLMClient.layer.pipe(Layer.provide(llmDeps))
 
 const program = Effect.gen(function* () {
   // yield* generateOnce
+  // yield* inspectFakeProvider
+  // yield* LLMClient.prepare(rawOverlayExample).pipe(Effect.andThen((prepared) => Effect.sync(() => console.log(prepared.body))))
   // yield* streamText
   // yield* generateStructuredObject
   // yield* generateDynamicObject.pipe(Effect.andThen((response) => Effect.sync(() => console.log(response.object))))

packages/ai/src/cache-policy.ts

@@ -38,7 +38,7 @@ const resolve = (policy: CachePolicy | undefined): CachePolicyObject => {
 // Protocols whose wire format ignores inline cache markers (OpenAI's implicit
 // prefix caching, Gemini's implicit + out-of-band CachedContent). Skip the
 // whole policy pass for these — emitting hints would be harmless but pointless.
-const RESPECTS_INLINE_HINTS = new Set(["anthropic-messages", "bedrock-converse", "openrouter"])
+const RESPECTS_INLINE_HINTS = new Set(["anthropic-messages", "bedrock-converse"])
 
 const makeHint = (ttlSeconds: number | undefined): CacheHint =>
   ttlSeconds !== undefined ? new CacheHint({ type: "ephemeral", ttlSeconds }) : new CacheHint({ type: "ephemeral" })
@@ -133,7 +133,6 @@ const countHints = (request: LLMRequest) =>
 
 export const applyCachePolicy = (request: LLMRequest): LLMRequest => {
   if (!RESPECTS_INLINE_HINTS.has(request.model.route.id)) return request
-  if (request.model.route.id === "openrouter" && (request.cache === undefined || request.cache === "auto")) return request
   const policy = resolve(request.cache)
   if (!policy.tools && !policy.system && !policy.messages) return request
 

packages/ai/src/llm.ts

@@ -9,28 +9,25 @@ import {
   LLMRequest,
   LLMResponse,
   Message,
-  Model,
   SystemPart,
   ToolChoice,
   ToolDefinition,
   type ContentPart,
-  type ModelProviderOptions,
 } from "./schema"
 import { make as makeTool, toDefinitions, type ToolSchema } from "./tool"
 
 /** Input accepted by `LLM.request`, normalized into the canonical `LLMRequest` class. */
-export type RequestInput<SelectedModel extends Model = Model> = Omit<
+export type RequestInput = Omit<
   ConstructorParameters<typeof LLMRequest>[0],
-  "model" | "system" | "messages" | "tools" | "toolChoice" | "generation" | "http" | "providerOptions"
+  "system" | "messages" | "tools" | "toolChoice" | "generation" | "http" | "providerOptions"
 > & {
-  readonly model: SelectedModel
   readonly system?: string | SystemPart | ReadonlyArray<SystemPart>
   readonly prompt?: string | ContentPart | ReadonlyArray<ContentPart>
   readonly messages?: ReadonlyArray<Message | Message.Input>
   readonly tools?: ReadonlyArray<ToolDefinition.Input>
   readonly toolChoice?: ToolChoice.Input
   readonly generation?: GenerationOptions.Input
-  readonly providerOptions?: NoInfer<ModelProviderOptions<SelectedModel>>
+  readonly providerOptions?: ConstructorParameters<typeof LLMRequest>[0]["providerOptions"]
   readonly http?: HttpOptions.Input
 }
 
@@ -38,7 +35,7 @@ export const generate = LLMClient.generate
 
 export const stream = LLMClient.stream
 
-export const request = <const SelectedModel extends Model>(input: RequestInput<SelectedModel>) => {
+export const request = (input: RequestInput) => {
   const {
     system: requestSystem,
     prompt,
@@ -66,7 +63,7 @@ const GENERATE_OBJECT_TOOL_NAME = "generate_object"
 
 const GENERATE_OBJECT_TOOL_DESCRIPTION = "Return the structured result by calling this tool."
 
-type GenerateObjectBase<SelectedModel extends Model = Model> = Omit<RequestInput<SelectedModel>, "tools" | "toolChoice">
+type GenerateObjectBase = Omit<RequestInput, "tools" | "toolChoice">
 
 export class GenerateObjectResponse<T> {
   constructor(
@@ -83,13 +80,11 @@ export class GenerateObjectResponse<T> {
   }
 }
 
-export interface GenerateObjectOptions<S extends ToolSchema<any>, SelectedModel extends Model = Model>
+export interface GenerateObjectOptions<S extends ToolSchema<any>> extends GenerateObjectBase {
-  extends GenerateObjectBase<SelectedModel> {
   readonly schema: S
 }
 
-export interface GenerateObjectDynamicOptions<SelectedModel extends Model = Model>
+export interface GenerateObjectDynamicOptions extends GenerateObjectBase {
-  extends GenerateObjectBase<SelectedModel> {
   /** Raw JSON Schema object describing the expected output shape. */
   readonly jsonSchema: JsonSchema.JsonSchema
 }
@@ -142,11 +137,11 @@ const runGenerateObject = Effect.fn("LLM.generateObject")(function* (
  * 2. `jsonSchema: JsonSchema.JsonSchema` — `.object` is `unknown`. Use when
  *    the schema is only available at runtime (MCP, plugin manifests). Caller validates.
  */
-export function generateObject<const SelectedModel extends Model, S extends ToolSchema<any>>(
+export function generateObject<S extends ToolSchema<any>>(
-  options: GenerateObjectOptions<S, SelectedModel>,
+  options: GenerateObjectOptions<S>,
 ): Effect.Effect<GenerateObjectResponse<Schema.Schema.Type<S>>, LLMError>
-export function generateObject<const SelectedModel extends Model>(
+export function generateObject(
-  options: GenerateObjectDynamicOptions<SelectedModel>,
+  options: GenerateObjectDynamicOptions,
 ): Effect.Effect<GenerateObjectResponse<unknown>, LLMError>
 export function generateObject(options: GenerateObjectOptions<ToolSchema<any>> | GenerateObjectDynamicOptions) {
   if ("schema" in options) {

packages/ai/src/protocols/gemini.ts

@@ -29,30 +29,9 @@ export const DEFAULT_BASE_URL = "https://generativelanguage.googleapis.com/v1bet
 
 export interface OptionsInput {
   readonly [key: string]: unknown
-  readonly cachedContent?: string
-  readonly safetySettings?: ReadonlyArray<{
-    readonly category:
-      | "HARM_CATEGORY_UNSPECIFIED"
-      | "HARM_CATEGORY_HATE_SPEECH"
-      | "HARM_CATEGORY_DANGEROUS_CONTENT"
-      | "HARM_CATEGORY_HARASSMENT"
-      | "HARM_CATEGORY_SEXUALLY_EXPLICIT"
-      | "HARM_CATEGORY_CIVIC_INTEGRITY"
-      | (string & {})
-    readonly threshold:
-      | "HARM_BLOCK_THRESHOLD_UNSPECIFIED"
-      | "BLOCK_LOW_AND_ABOVE"
-      | "BLOCK_MEDIUM_AND_ABOVE"
-      | "BLOCK_ONLY_HIGH"
-      | "BLOCK_NONE"
-      | "OFF"
-      | (string & {})
-  }>
-  readonly serviceTier?: "standard" | "flex" | "priority" | (string & {})
   readonly thinkingConfig?: {
     readonly thinkingBudget?: number
     readonly includeThoughts?: boolean
-    readonly thinkingLevel?: "minimal" | "low" | "medium" | "high" | (string & {})
   }
 }
 
@@ -132,12 +111,6 @@ const GeminiToolConfig = Schema.Struct({
 const GeminiThinkingConfig = Schema.Struct({
   thinkingBudget: Schema.optional(Schema.Number),
   includeThoughts: Schema.optional(Schema.Boolean),
-  thinkingLevel: Schema.optional(Schema.String),
-})
-
-const GeminiSafetySetting = Schema.Struct({
-  category: Schema.String,
-  threshold: Schema.String,
 })
 
 const GeminiGenerationConfig = Schema.Struct({
@@ -150,10 +123,7 @@ const GeminiGenerationConfig = Schema.Struct({
 })
 
 const GeminiBodyFields = {
-  cachedContent: Schema.optional(Schema.String),
   contents: Schema.Array(GeminiContent),
-  safetySettings: optionalArray(GeminiSafetySetting),
-  serviceTier: Schema.optional(Schema.String),
   systemInstruction: Schema.optional(GeminiSystemInstruction),
   tools: optionalArray(GeminiTool),
   toolConfig: Schema.optional(GeminiToolConfig),
@@ -346,38 +316,17 @@ const lowerMessages = Effect.fn("Gemini.lowerMessages")(function* (request: LLMR
 })
 
 const resolveOptions = (request: LLMRequest) => {
-  const input = request.providerOptions?.gemini
+  const value = request.providerOptions?.gemini?.thinkingConfig
-  const value = input?.thinkingConfig
+  if (!ProviderShared.isRecord(value)) return {}
   const thinkingConfig = {
-    thinkingBudget:
+    thinkingBudget: typeof value.thinkingBudget === "number" ? value.thinkingBudget : undefined,
-      ProviderShared.isRecord(value) && typeof value.thinkingBudget === "number" ? value.thinkingBudget : undefined,
+    includeThoughts: typeof value.includeThoughts === "boolean" ? value.includeThoughts : undefined,
-    includeThoughts:
-      ProviderShared.isRecord(value) && typeof value.includeThoughts === "boolean"
-        ? value.includeThoughts
-        : ProviderShared.isRecord(value)
-          ? true
-          : undefined,
-    thinkingLevel:
-      ProviderShared.isRecord(value) && typeof value.thinkingLevel === "string" ? value.thinkingLevel : undefined,
   }
   return {
-    cachedContent: typeof input?.cachedContent === "string" ? input.cachedContent : undefined,
-    safetySettings: mapSafetySettings(input?.safetySettings),
-    serviceTier: typeof input?.serviceTier === "string" ? input.serviceTier : undefined,
     thinkingConfig: Object.values(thinkingConfig).some((item) => item !== undefined) ? thinkingConfig : undefined,
   }
 }
 
-function mapSafetySettings(value: unknown) {
-  if (!Array.isArray(value)) return undefined
-  const settings = value.flatMap((item) =>
-    ProviderShared.isRecord(item) && typeof item.category === "string" && typeof item.threshold === "string"
-      ? [{ category: item.category, threshold: item.threshold }]
-      : [],
-  )
-  return settings
-}
-
 const fromRequest = Effect.fn("Gemini.fromRequest")(function* (request: LLMRequest) {
   const hasTools = request.tools.length > 0
   const generation = request.generation
@@ -393,10 +342,7 @@ const fromRequest = Effect.fn("Gemini.fromRequest")(function* (request: LLMReque
   }
 
   return {
-    cachedContent: options.cachedContent,
     contents: yield* lowerMessages(request),
-    safetySettings: options.safetySettings,
-    serviceTier: options.serviceTier,
     systemInstruction:
       request.system.length === 0 ? undefined : { parts: [{ text: ProviderShared.joinText(request.system) }] },
     tools: hasTools

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

@@ -11,7 +11,6 @@ import {
   Usage,
   type FinishReason,
   type FinishReasonDetails,
-  type CacheHint,
   type JsonSchema,
   type LLMRequest,
   type MediaPart,
@@ -39,11 +38,6 @@ export const PATH = "/chat/completions"
 // The body schema is the provider-native JSON body. `fromRequest` below builds
 // this shape from the common `LLMRequest`, then `Route.make` validates and
 // JSON-encodes it before transport.
-const OpenAIChatCacheControl = Schema.Struct({
-  type: Schema.Literal("ephemeral"),
-  ttl: Schema.optional(Schema.String),
-})
-
 const OpenAIChatFunction = Schema.Struct({
   name: Schema.String,
   description: Schema.String,
@@ -53,7 +47,6 @@ const OpenAIChatFunction = Schema.Struct({
 const OpenAIChatTool = Schema.Struct({
   type: Schema.tag("function"),
   function: OpenAIChatFunction,
-  cache_control: Schema.optional(OpenAIChatCacheControl),
 })
 type OpenAIChatTool = Schema.Schema.Type<typeof OpenAIChatTool>
 
@@ -68,11 +61,7 @@ const OpenAIChatAssistantToolCall = Schema.Struct({
 type OpenAIChatAssistantToolCall = Schema.Schema.Type<typeof OpenAIChatAssistantToolCall>
 
 const OpenAIChatUserContent = Schema.Union([
-  Schema.Struct({
+  Schema.Struct({ type: Schema.Literal("text"), text: Schema.String }),
-    type: Schema.Literal("text"),
-    text: Schema.String,
-    cache_control: Schema.optional(OpenAIChatCacheControl),
-  }),
   Schema.Struct({
     type: Schema.Literal("image_url"),
     image_url: Schema.Struct({ url: Schema.String }),
@@ -80,10 +69,7 @@ const OpenAIChatUserContent = Schema.Union([
 ])
 
 const OpenAIChatMessage = Schema.Union([
-  Schema.Struct({
+  Schema.Struct({ role: Schema.Literal("system"), content: Schema.String }),
-    role: Schema.Literal("system"),
-    content: Schema.Union([Schema.String, Schema.Array(OpenAIChatUserContent)]),
-  }),
   Schema.Struct({
     role: Schema.Literal("user"),
     content: Schema.Union([Schema.String, Schema.Array(OpenAIChatUserContent)]),
@@ -97,16 +83,10 @@ const OpenAIChatMessage = Schema.Union([
       reasoning: Schema.optional(Schema.String),
       reasoning_text: Schema.optional(Schema.String),
       reasoning_details: Schema.optional(Schema.Unknown),
-      cache_control: Schema.optional(OpenAIChatCacheControl),
     }),
     [Schema.Record(Schema.String, Schema.Unknown)],
   ),
-  Schema.Struct({
+  Schema.Struct({ role: Schema.Literal("tool"), tool_call_id: Schema.String, content: Schema.String }),
-    role: Schema.Literal("tool"),
-    tool_call_id: Schema.String,
-    content: Schema.String,
-    cache_control: Schema.optional(OpenAIChatCacheControl),
-  }),
 ]).pipe(Schema.toTaggedUnion("role"))
 type OpenAIChatMessage = Schema.Schema.Type<typeof OpenAIChatMessage>
 
@@ -127,7 +107,6 @@ export const bodyFields = {
   stream_options: Schema.optional(Schema.Struct({ include_usage: Schema.Boolean })),
   store: Schema.optional(Schema.Boolean),
   reasoning_effort: Schema.optional(OpenAIOptions.OpenAIReasoningEffort),
-  max_completion_tokens: Schema.optional(Schema.Number),
   max_tokens: Schema.optional(Schema.Number),
   temperature: Schema.optional(Schema.Number),
   top_p: Schema.optional(Schema.Number),
@@ -230,20 +209,13 @@ export interface ParserState {
 // Lowering is the only place that knows how common LLM messages map onto the
 // OpenAI Chat wire format. Keep provider quirks here instead of leaking native
 // fields into `LLMRequest`.
-interface LoweringOptions {
+const lowerTool = (tool: ToolDefinition, inputSchema: JsonSchema): OpenAIChatTool => ({
-  readonly cacheControl?: (
-    cache: CacheHint | undefined,
-  ) => Schema.Schema.Type<typeof OpenAIChatCacheControl> | undefined
-}
-
-const lowerTool = (tool: ToolDefinition, inputSchema: JsonSchema, options: LoweringOptions): OpenAIChatTool => ({
   type: "function",
   function: {
     name: tool.name,
     description: tool.description,
     parameters: ToolSchemaProjection.openAI(inputSchema),
   },
-  cache_control: options.cacheControl?.(tool.cache),
 })
 
 const lowerToolChoice = (toolChoice: NonNullable<LLMRequest["toolChoice"]>) =>
@@ -285,14 +257,11 @@ const reasoningDetails = (parts: ReadonlyArray<ReasoningPart>, native: unknown)
   if (isRecord(native) && Array.isArray(native.reasoning_details)) return native.reasoning_details
 }
 
-const lowerUserMessage = Effect.fn("OpenAIChat.lowerUserMessage")(function* (
+const lowerUserMessage = Effect.fn("OpenAIChat.lowerUserMessage")(function* (message: OpenAIChatRequestMessage) {
-  message: OpenAIChatRequestMessage,
-  options: LoweringOptions,
-) {
   const content: Array<Schema.Schema.Type<typeof OpenAIChatUserContent>> = []
   for (const part of message.content) {
     if (part.type === "text") {
-      content.push({ type: "text", text: part.text, cache_control: options.cacheControl?.(part.cache) })
+      content.push({ type: "text", text: part.text })
       continue
     }
     if (part.type === "media") {
@@ -301,18 +270,14 @@ const lowerUserMessage = Effect.fn("OpenAIChat.lowerUserMessage")(function* (
     }
     return yield* ProviderShared.unsupportedContent("OpenAI Chat", "user", ["text", "media"])
   }
-  if (content.every((part) => part.type === "text" && part.cache_control === undefined))
+  if (content.every((part) => part.type === "text"))
-    return {
+    return { role: "user" as const, content: content.map((part) => part.text).join("") }
-      role: "user" as const,
-      content: content.map((part) => (part.type === "text" ? part.text : "")).join(""),
-    }
   return { role: "user" as const, content }
 })
 
 const lowerAssistantMessage = Effect.fn("OpenAIChat.lowerAssistantMessage")(function* (
   message: OpenAIChatRequestMessage,
   configuredField?: string,
-  options: LoweringOptions = {},
 ) {
   const content: TextPart[] = []
   const reasoning: ReasoningPart[] = []
@@ -350,44 +315,29 @@ const lowerAssistantMessage = Effect.fn("OpenAIChat.lowerAssistantMessage")(func
     if (reasoning.length === 0) return nativeReasoning
     return text
   })()
-  const cached = message.content.findLast((part) => "cache" in part && part.cache !== undefined)
   const result = {
     role: "assistant" as const,
     content: content.length === 0 ? null : ProviderShared.joinText(content),
     tool_calls: toolCalls.length === 0 ? undefined : toolCalls,
     reasoning_details: details,
-    cache_control: options.cacheControl?.(cached && "cache" in cached ? cached.cache : undefined),
   }
   if (field === undefined || reasoningText === undefined) return result
   return { ...result, [field]: reasoningText }
 })
 
-const lowerToolMessages = Effect.fn("OpenAIChat.lowerToolMessages")(function* (
+const lowerToolMessages = Effect.fn("OpenAIChat.lowerToolMessages")(function* (message: OpenAIChatRequestMessage) {
-  message: OpenAIChatRequestMessage,
-  options: LoweringOptions,
-) {
   const messages: OpenAIChatMessage[] = []
   const images: Array<Schema.Schema.Type<typeof OpenAIChatUserContent>> = []
   for (const part of message.content) {
     if (!ProviderShared.supportsContent(part, ["tool-result"]))
       return yield* ProviderShared.unsupportedContent("OpenAI Chat", "tool", ["tool-result"])
     if (part.result.type !== "content") {
-      messages.push({
+      messages.push({ role: "tool", tool_call_id: part.id, content: ProviderShared.toolResultText(part) })
-        role: "tool",
-        tool_call_id: part.id,
-        content: ProviderShared.toolResultText(part),
-        cache_control: options.cacheControl?.(part.cache),
-      })
       continue
     }
     const content: ReadonlyArray<Tool.Content> = part.result.value
     const text = content.filter((item) => item.type === "text").map((item) => item.text)
-    messages.push({
+    messages.push({ role: "tool", tool_call_id: part.id, content: text.join("\n") })
-      role: "tool",
-      tool_call_id: part.id,
-      content: text.join("\n"),
-      cache_control: options.cacheControl?.(part.cache),
-    })
     const files = content.filter((item) => item.type === "file")
     images.push(
       ...(yield* Effect.forEach(files, (item) =>
@@ -401,29 +351,15 @@ const lowerToolMessages = Effect.fn("OpenAIChat.lowerToolMessages")(function* (
 const lowerMessage = Effect.fn("OpenAIChat.lowerMessage")(function* (
   message: OpenAIChatRequestMessage,
   reasoningField?: string,
-  options: LoweringOptions = {},
 ) {
-  if (message.role === "user") return [yield* lowerUserMessage(message, options)]
+  if (message.role === "user") return [yield* lowerUserMessage(message)]
-  if (message.role === "assistant") return [yield* lowerAssistantMessage(message, reasoningField, options)]
+  if (message.role === "assistant") return [yield* lowerAssistantMessage(message, reasoningField)]
-  return (yield* lowerToolMessages(message, options)).messages
+  return (yield* lowerToolMessages(message)).messages
 })
 
-const lowerMessages = Effect.fn("OpenAIChat.lowerMessages")(function* (request: LLMRequest, options: LoweringOptions) {
+const lowerMessages = Effect.fn("OpenAIChat.lowerMessages")(function* (request: LLMRequest) {
   const system: OpenAIChatMessage[] =
-    request.system.length === 0
+    request.system.length === 0 ? [] : [{ role: "system", content: ProviderShared.joinText(request.system) }]
-      ? []
-      : request.system.some((part) => part.cache !== undefined) && options.cacheControl !== undefined
-        ? [
-            {
-              role: "system",
-              content: request.system.map((part) => ({
-                type: "text",
-                text: part.text,
-                cache_control: options.cacheControl?.(part.cache),
-              })),
-            },
-          ]
-        : [{ role: "system", content: ProviderShared.joinText(request.system) }]
   const messages = [...system]
   const pendingImages: Array<Schema.Schema.Type<typeof OpenAIChatUserContent>> = []
   const flushImages = () => {
@@ -434,53 +370,28 @@ const lowerMessages = Effect.fn("OpenAIChat.lowerMessages")(function* (request:
     if (message.role === "system") {
       const part = yield* ProviderShared.wrappedSystemUpdate("OpenAI Chat", message)
       if (pendingImages.length > 0) {
-        messages.push({
+        messages.push({ role: "user", content: [...pendingImages.splice(0), { type: "text", text: part.text }] })
-          role: "user",
-          content: [
-            ...pendingImages.splice(0),
-            { type: "text", text: part.text, cache_control: options.cacheControl?.(part.cache) },
-          ],
-        })
         continue
       }
       const previous = messages.at(-1)
       if (previous?.role === "user" && typeof previous.content === "string")
-        messages[messages.length - 1] = options.cacheControl?.(part.cache)
+        messages[messages.length - 1] = { role: "user", content: `${previous.content}\n${part.text}` }
-          ? {
-              role: "user",
-              content: [
-                { type: "text", text: previous.content },
-                { type: "text", text: part.text, cache_control: options.cacheControl(part.cache) },
-              ],
-            }
-          : { role: "user", content: `${previous.content}\n${part.text}` }
       else if (previous?.role === "user" && Array.isArray(previous.content))
         messages[messages.length - 1] = {
           role: "user",
-          content: [
+          content: [...previous.content, { type: "text", text: part.text }],
-            ...previous.content,
-            { type: "text", text: part.text, cache_control: options.cacheControl?.(part.cache) },
-          ],
         }
-      else
+      else messages.push({ role: "user", content: part.text })
-        messages.push(
-          options.cacheControl?.(part.cache)
-            ? {
-                role: "user",
-                content: [{ type: "text", text: part.text, cache_control: options.cacheControl(part.cache) }],
-              }
-            : { role: "user", content: part.text },
-        )
       continue
     }
     if (message.role === "tool") {
-      const lowered = yield* lowerToolMessages(message, options)
+      const lowered = yield* lowerToolMessages(message)
       messages.push(...lowered.messages)
       pendingImages.push(...lowered.images)
       continue
     }
     flushImages()
-    messages.push(...(yield* lowerMessage(message, request.model.compatibility?.reasoningField, options)))
+    messages.push(...(yield* lowerMessage(message, request.model.compatibility?.reasoningField)))
   }
   flushImages()
   return messages
@@ -494,10 +405,7 @@ const lowerOptions = (request: LLMRequest) => {
   }
 }
 
-export const fromRequest = Effect.fn("OpenAIChat.fromRequest")(function* (
+const fromRequest = Effect.fn("OpenAIChat.fromRequest")(function* (request: LLMRequest) {
-  request: LLMRequest,
-  options: LoweringOptions = {},
-) {
   // `fromRequest` returns the provider body only. Endpoint, auth, framing,
   // validation, and HTTP execution are composed by `Route.make`.
   const reasoningField = request.model.compatibility?.reasoningField
@@ -507,26 +415,19 @@ export const fromRequest = Effect.fn("OpenAIChat.fromRequest")(function* (
     )
   const generation = request.generation
   const toolSchemaCompatibility = request.model.compatibility?.toolSchema
-  const maxTokensField = request.model.compatibility?.maxTokensField ?? "max_tokens"
   return {
     model: request.model.id,
-    messages: yield* lowerMessages(request, options),
+    messages: yield* lowerMessages(request),
     tools:
       request.tools.length === 0
         ? undefined
         : request.tools.map((tool) =>
-            lowerTool(
+            lowerTool(tool, ToolSchemaProjection.modelCompatibility(tool.inputSchema, toolSchemaCompatibility)),
-              tool,
-              ToolSchemaProjection.modelCompatibility(tool.inputSchema, toolSchemaCompatibility),
-              options,
-            ),
           ),
     tool_choice: request.toolChoice ? yield* lowerToolChoice(request.toolChoice) : undefined,
     stream: true as const,
     stream_options: { include_usage: true },
-    ...(maxTokensField === "max_completion_tokens"
+    max_tokens: generation?.maxTokens,
-      ? { max_completion_tokens: generation?.maxTokens }
-      : { max_tokens: generation?.maxTokens }),
     temperature: generation?.temperature,
     top_p: generation?.topP,
     frequency_penalty: generation?.frequencyPenalty,

packages/ai/src/provider-error.ts

@@ -16,6 +16,7 @@ import {
 
 const patterns = [
   /prompt is too long/i,
+  /request_too_large/i,
   /input is too long for requested model/i,
   /exceeds the context window/i,
   /exceeds (?:the )?(?:model'?s )?maximum context length(?: of [\d,]+ tokens?|\s*\([\d,]+\))/i,
@@ -32,6 +33,7 @@ const patterns = [
   /context window exceeds limit/i,
   /exceeded model token limit/i,
   /context[_ ]length[_ ]exceeded/i,
+  /request entity too large/i,
   /context length is only \d+ tokens/i,
   /input length.*exceeds.*context length/i,
   /prompt too long; exceeded (?:max )?context length/i,
@@ -42,15 +44,11 @@ const patterns = [
   /token limit exceeded/i,
 ]
 
-const payloadPatterns = [/request_too_large/i, /request entity too large/i, /payload too large/i, /request too large/i]
-
 const exclusions = [/^(throttling error|service unavailable):/i, /rate limit/i, /too many requests/i]
 
 export const isContextOverflow = (message: string) =>
   !exclusions.some((pattern) => pattern.test(message)) &&
-  (patterns.some((pattern) => pattern.test(message)) || /^400\s*(status code)?\s*\(no body\)/i.test(message))
+  (patterns.some((pattern) => pattern.test(message)) || /^4(00|13)\s*(status code)?\s*\(no body\)/i.test(message))
-
-export const isPayloadTooLarge = (message: string) => payloadPatterns.some((pattern) => pattern.test(message))
 
 export const isContextOverflowFailure = (failure: unknown) =>
   failure instanceof LLMError
@@ -102,8 +100,6 @@ export function classifyProviderFailure(input: ProviderFailure): LLMError["reaso
       isContextOverflow(text))
   )
     return new InvalidRequestReason({ ...common, classification: "context-overflow" })
-  if (input.status === 413 || isPayloadTooLarge(text))
-    return new InvalidRequestReason({ ...common, classification: "payload-too-large" })
   if (CONTENT_POLICY_TEXT.test(text)) return new ContentPolicyReason(common)
   if (codes.some((code) => QUOTA_CODES.has(code)) || (input.status === 429 && QUOTA_TEXT.test(text)))
     return new QuotaExceededReason(common)
@@ -146,7 +142,12 @@ export function classifyProviderFailure(input: ProviderFailure): LLMError["reaso
       retryAfterMs: input.retryAfterMs,
     })
   if (codes.some((code) => INVALID_REQUEST_CODES.has(code))) return new InvalidRequestReason(common)
-  if (input.status === 400 || input.status === 404 || input.status === 413 || input.status === 422)
+  if (
+    input.status === 400 ||
+    input.status === 404 ||
+    input.status === 413 ||
+    input.status === 422
+  )
     return new InvalidRequestReason(common)
   return new UnknownProviderReason({ ...common, status: input.status })
 }

packages/ai/src/provider-package.ts

@@ -1,21 +1,16 @@
-import type { Model, ProviderOptions } from "./schema"
+import type { Model } from "./schema"
 
 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<
+export interface Definition<ProviderSettings extends Settings = Settings> {
-  ProviderSettings extends Settings = Settings,
+  readonly model: (modelID: string, settings: ProviderSettings) => Model
-  Options extends ProviderOptions = ProviderOptions,
-> {
-  readonly model: (modelID: string, settings: ProviderSettings) => Model<Options>
 }
 
 export * as ProviderPackage from "./provider-package"

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

@@ -47,7 +47,7 @@ export const configure = (input: Config) => {
   })
   return {
     id: ProviderID.make(provider),
-    model: (modelID: string | ModelID) => route.model<AnthropicMessages.ProviderOptionsInput>({ id: modelID }),
+    model: (modelID: string | ModelID) => route.model({ id: modelID }),
     configure,
   }
 }
@@ -57,10 +57,7 @@ export const provider = {
   configure,
 }
 
-export const model: ProviderPackage.Definition<Settings, AnthropicMessages.ProviderOptionsInput>["model"] = (
+export const model: ProviderPackage.Definition<Settings>["model"] = (modelID, settings) => {
-  modelID,
-  settings,
-) => {
   if (settings.apiKey !== undefined && settings.authToken !== undefined)
     throw new Error("Anthropic-compatible apiKey cannot be combined with authToken")
   return configure({

packages/ai/src/providers/anthropic.ts

@@ -52,10 +52,7 @@ export const configure = (input: Config = {}) => {
 }
 
 export const provider = configure()
-export const model: ProviderPackage.Definition<Settings, AnthropicMessages.ProviderOptionsInput>["model"] = (
+export const model: ProviderPackage.Definition<Settings>["model"] = (modelID, settings) => {
-  modelID,
-  settings,
-) => {
   if (settings.apiKey !== undefined && settings.authToken !== undefined)
     throw new Error("Anthropic apiKey cannot be combined with authToken")
   return configure({

packages/ai/src/providers/azure.ts

@@ -99,14 +99,10 @@ export const configure = (input: Config) => {
   const modelDefaults = defaults(input)
 
   const responses = (modelID: string | ModelID) =>
-    configuredResponsesRoute
+    configuredResponsesRoute.with(withOpenAIOptions(modelID, modelDefaults)).model({ id: modelID })
-      .with(withOpenAIOptions(modelID, modelDefaults))
-      .model<OpenAIProviderOptionsInput>({ id: modelID })
 
   const chat = (modelID: string | ModelID) =>
-    configuredChatRoute
+    configuredChatRoute.with(withOpenAIOptions(modelID, modelDefaults)).model({ id: modelID })
-      .with(withOpenAIOptions(modelID, modelDefaults))
-      .model<OpenAIProviderOptionsInput>({ id: modelID })
 
   return {
     id,
@@ -137,12 +133,8 @@ const config = (settings: Settings): Config => {
   throw new Error("Azure requires resourceName or baseURL")
 }
 
-export const responsesModel: ProviderPackage.Definition<Settings, OpenAIProviderOptionsInput>["model"] = (
+export const responsesModel: ProviderPackage.Definition<Settings>["model"] = (modelID, settings) =>
-  modelID,
+  configure(config(settings)).responses(modelID)
-  settings,
+export const chatModel: ProviderPackage.Definition<Settings>["model"] = (modelID, settings) =>
-) => configure(config(settings)).responses(modelID)
+  configure(config(settings)).chat(modelID)
-export const chatModel: ProviderPackage.Definition<Settings, OpenAIProviderOptionsInput>["model"] = (
-  modelID,
-  settings,
-) => configure(config(settings)).chat(modelID)
 export const model = responsesModel

packages/ai/src/providers/cloudflare.ts

@@ -4,7 +4,6 @@ import { Auth } from "../route/auth"
 import { AuthOptions, type AtLeastOne, type ProviderAuthOption } from "../route/auth-options"
 import type { RouteDefaultsInput } from "../route/client"
 import { ProviderID, type ModelID } from "../schema"
-import type { OpenAIProviderOptionsInput } from "./openai-options"
 
 export const aiGatewayID = ProviderID.make("cloudflare-ai-gateway")
 export const workersAIID = ProviderID.make("cloudflare-workers-ai")
@@ -21,11 +20,10 @@ type GatewayURL = AtLeastOne<{
 }
 
 export type AIGatewayOptions = GatewayURL &
-  Omit<RouteDefaultsInput, "providerOptions"> &
+  RouteDefaultsInput &
   ProviderAuthOption<"optional"> & {
     /** Cloudflare AI Gateway authentication token. Sent as `cf-aig-authorization`. */
     readonly gatewayApiKey?: CloudflareSecret
-    readonly providerOptions?: OpenAIProviderOptionsInput
   }
 
 type WorkersAIURL = AtLeastOne<{
@@ -33,11 +31,7 @@ type WorkersAIURL = AtLeastOne<{
   readonly baseURL: string
 }>
 
-export type WorkersAIOptions = WorkersAIURL &
+export type WorkersAIOptions = WorkersAIURL & RouteDefaultsInput & ProviderAuthOption<"optional">
-  Omit<RouteDefaultsInput, "providerOptions"> &
-  ProviderAuthOption<"optional"> & {
-    readonly providerOptions?: OpenAIProviderOptionsInput
-  }
 
 export const aiGatewayBaseURL = (input: GatewayURL) => {
   if (input.baseURL) return input.baseURL
@@ -104,7 +98,7 @@ const configureAIGateway = (options: AIGatewayOptions) => {
   })
   return {
     id: aiGatewayID,
-    model: (modelID: string | ModelID) => route.model<OpenAIProviderOptionsInput>({ id: modelID }),
+    model: (modelID: string | ModelID) => route.model({ id: modelID }),
     configure: configureAIGateway,
   }
 }
@@ -117,7 +111,7 @@ const configureWorkersAI = (options: WorkersAIOptions) => {
   })
   return {
     id: workersAIID,
-    model: (modelID: string | ModelID) => route.model<OpenAIProviderOptionsInput>({ id: modelID }),
+    model: (modelID: string | ModelID) => route.model({ id: modelID }),
     configure: configureWorkersAI,
   }
 }

packages/ai/src/providers/github-copilot.ts

@@ -50,11 +50,9 @@ export const configure = (options: ModelOptions) => {
   const responsesRoute = configuredResponsesRoute(options)
   const chatRoute = configuredChatRoute(options)
   const responses = (modelID: string | ModelID) =>
-    responsesRoute
+    responsesRoute.with(withOpenAIOptions(modelID, defaults(options))).model({ id: modelID })
-      .with(withOpenAIOptions(modelID, defaults(options)))
-      .model<OpenAIProviderOptionsInput>({ id: modelID })
   const chat = (modelID: string | ModelID) =>
-    chatRoute.with(withOpenAIOptions(modelID, defaults(options))).model<OpenAIProviderOptionsInput>({ id: modelID })
+    chatRoute.with(withOpenAIOptions(modelID, defaults(options))).model({ id: modelID })
   return {
     id,
     model: (modelID: string | ModelID) =>

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

@@ -1,9 +1,8 @@
 import type { ProviderPackage } from "../provider-package"
 import { OpenAICompatibleChat } from "../protocols/openai-compatible-chat"
 import type { RouteDefaultsInput } from "../route/client"
-import { ProviderID, type ModelID } from "../schema"
+import { ProviderID, type ModelID, type ProviderOptions } from "../schema"
 import { GoogleVertexShared } from "./google-vertex-shared"
-import type { OpenAIProviderOptionsInput } from "./openai-options"
 
 export const id = ProviderID.make("google-vertex")
 
@@ -12,7 +11,6 @@ export type Config = RouteDefaultsInput &
     readonly baseURL?: string
     readonly location?: string
     readonly project?: string
-    readonly providerOptions?: OpenAIProviderOptionsInput
   }
 
 export interface Settings extends ProviderPackage.Settings {
@@ -21,7 +19,7 @@ export interface Settings extends ProviderPackage.Settings {
   readonly baseURL?: string
   readonly location?: string
   readonly project?: string
-  readonly providerOptions?: OpenAIProviderOptionsInput
+  readonly providerOptions?: ProviderOptions
 }
 
 const route = OpenAICompatibleChat.route.with({
@@ -58,7 +56,7 @@ export const configure = (input: Config = {}) => {
   const route = configuredRoute(input)
   return {
     id,
-    model: (modelID: string | ModelID) => route.model<OpenAIProviderOptionsInput>({ id: modelID }),
+    model: (modelID: string | ModelID) => route.model({ id: modelID }),
     configure,
   }
 }
@@ -68,7 +66,7 @@ export const provider = {
   configure,
 }
 
-export const model: ProviderPackage.Definition<Settings, OpenAIProviderOptionsInput>["model"] = (modelID, settings) => {
+export const model: ProviderPackage.Definition<Settings>["model"] = (modelID, settings) => {
   if (settings.apiKey !== undefined) throw new Error("Google Vertex Chat does not support API keys")
   return configure({
     accessToken: settings.accessToken,

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

@@ -91,7 +91,7 @@ export const configure = (input: Config = {}) => {
   const route = configuredRoute(input)
   return {
     id,
-    model: (modelID: string | ModelID) => route.model<AnthropicMessages.ProviderOptionsInput>({ id: modelID }),
+    model: (modelID: string | ModelID) => route.model({ id: modelID }),
     configure,
   }
 }
@@ -101,10 +101,7 @@ export const provider = {
   configure,
 }
 
-export const model: ProviderPackage.Definition<Settings, AnthropicMessages.ProviderOptionsInput>["model"] = (
+export const model: ProviderPackage.Definition<Settings>["model"] = (modelID, settings) => {
-  modelID,
-  settings,
-) => {
   if (settings.apiKey !== undefined) throw new Error("Google Vertex Messages does not support API keys")
   return configure({
     accessToken: settings.accessToken,

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

@@ -1,9 +1,8 @@
 import type { ProviderPackage } from "../provider-package"
 import { OpenAICompatibleResponses } from "../protocols/openai-compatible-responses"
 import type { RouteDefaultsInput } from "../route/client"
-import { ProviderID, type ModelID } from "../schema"
+import { ProviderID, type ModelID, type ProviderOptions } from "../schema"
 import { GoogleVertexShared } from "./google-vertex-shared"
-import type { OpenResponsesProviderOptionsInput } from "./open-responses-options"
 
 export const id = ProviderID.make("google-vertex")
 
@@ -12,7 +11,6 @@ export type Config = RouteDefaultsInput &
     readonly baseURL?: string
     readonly location?: string
     readonly project?: string
-    readonly providerOptions?: OpenResponsesProviderOptionsInput
   }
 
 export interface Settings extends ProviderPackage.Settings {
@@ -21,7 +19,7 @@ export interface Settings extends ProviderPackage.Settings {
   readonly baseURL?: string
   readonly location?: string
   readonly project?: string
-  readonly providerOptions?: OpenResponsesProviderOptionsInput
+  readonly providerOptions?: ProviderOptions
 }
 
 const route = OpenAICompatibleResponses.route.with({
@@ -60,7 +58,7 @@ export const configure = (input: Config = {}) => {
   const route = configuredRoute(input)
   return {
     id,
-    model: (modelID: string | ModelID) => route.model<OpenResponsesProviderOptionsInput>({ id: modelID }),
+    model: (modelID: string | ModelID) => route.model({ id: modelID }),
     configure,
   }
 }
@@ -70,10 +68,7 @@ export const provider = {
   configure,
 }
 
-export const model: ProviderPackage.Definition<Settings, OpenResponsesProviderOptionsInput>["model"] = (
+export const model: ProviderPackage.Definition<Settings>["model"] = (modelID, settings) => {
-  modelID,
-  settings,
-) => {
   if (settings.apiKey !== undefined) throw new Error("Google Vertex Responses does not support API keys")
   return configure({
     accessToken: settings.accessToken,

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

@@ -77,8 +77,7 @@ const configuredRoute = (input: Config, modelID: string | ModelID) => {
 export const configure = (input: Config = {}) => {
   return {
     id,
-    model: (modelID: string | ModelID) =>
+    model: (modelID: string | ModelID) => configuredRoute(input, modelID).model({ id: modelID }),
-      configuredRoute(input, modelID).model<Gemini.ProviderOptionsInput>({ id: modelID }),
     configure,
   }
 }
@@ -87,10 +86,7 @@ export const provider = {
   id,
   configure,
 }
-export const model: ProviderPackage.Definition<Settings, Gemini.ProviderOptionsInput>["model"] = (
+export const model: ProviderPackage.Definition<Settings>["model"] = (modelID, settings) => {
-  modelID,
-  settings,
-) => {
   if (settings.apiKey !== undefined && settings.accessToken !== undefined)
     throw new Error("Google Vertex apiKey cannot be combined with accessToken or auth")
   return configure({

packages/ai/src/providers/google.ts

@@ -50,14 +50,14 @@ export const configure = (input: Config = {}) => {
     })
   return {
     id,
-    model: (modelID: string | ModelID) => route.model<Gemini.ProviderOptionsInput>({ id: modelID }),
+    model: (modelID: string | ModelID) => route.model({ id: modelID }),
     image,
     configure,
   }
 }
 
 export const provider = configure()
-export const model: ProviderPackage.Definition<Settings, Gemini.ProviderOptionsInput>["model"] = (modelID, settings) =>
+export const model: ProviderPackage.Definition<Settings>["model"] = (modelID, settings) =>
   configure({
     apiKey: settings.apiKey,
     baseURL: settings.baseURL,

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

@@ -36,7 +36,7 @@ export const configure = (input: Config) => {
   })
   return {
     id: ProviderID.make(provider),
-    model: (modelID: string | ModelID) => route.model<OpenResponsesProviderOptionsInput>({ id: modelID }),
+    model: (modelID: string | ModelID) => route.model({ id: modelID }),
     configure,
   }
 }
@@ -46,10 +46,7 @@ export const provider = {
   configure,
 }
 
-export const model: ProviderPackage.Definition<Settings, OpenResponsesProviderOptionsInput>["model"] = (
+export const model: ProviderPackage.Definition<Settings>["model"] = (modelID, settings) =>
-  modelID,
-  settings,
-) =>
   configure({
     apiKey: settings.apiKey,
     baseURL: settings.baseURL,

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

@@ -4,15 +4,13 @@ import type { RouteDefaultsInput } from "../route/client"
 import { AuthOptions, type ProviderAuthOption } from "../route/auth-options"
 import type { ProviderPackage } from "../provider-package"
 import { profiles, type OpenAICompatibleProfile } from "./openai-compatible-profile"
-import type { OpenAIProviderOptionsInput } from "./openai-options"
 
 export const id = ProviderID.make("openai-compatible")
 
-type GenericModelOptions = Omit<RouteDefaultsInput, "providerOptions"> &
+type GenericModelOptions = RouteDefaultsInput &
   ProviderAuthOption<"optional"> & {
     readonly provider?: string
     readonly baseURL: string
-    readonly providerOptions?: OpenAIProviderOptionsInput
   }
 
 export interface Settings extends ProviderPackage.Settings {
@@ -21,10 +19,9 @@ export interface Settings extends ProviderPackage.Settings {
   readonly provider?: string
 }
 
-export type FamilyModelOptions = Omit<RouteDefaultsInput, "providerOptions"> &
+export type FamilyModelOptions = RouteDefaultsInput &
   ProviderAuthOption<"optional"> & {
     readonly baseURL?: string
-    readonly providerOptions?: OpenAIProviderOptionsInput
   }
 
 export const routes = [OpenAICompatibleChat.route]
@@ -40,8 +37,7 @@ export const configure = (input: GenericModelOptions) => {
   })
   return {
     id: ProviderID.make(provider),
-    model: (modelID: string | ModelID) =>
+    model: (modelID: string | ModelID) => route.model({ id: modelID, provider: ProviderID.make(provider) }),
-      route.model<OpenAIProviderOptionsInput>({ id: modelID, provider: ProviderID.make(provider) }),
     configure,
   }
 }
@@ -67,7 +63,7 @@ export const provider = {
   configure,
 }
 
-export const model: ProviderPackage.Definition<Settings, OpenAIProviderOptionsInput>["model"] = (modelID, settings) =>
+export const model: ProviderPackage.Definition<Settings>["model"] = (modelID, settings) =>
   configure({
     apiKey: settings.apiKey,
     baseURL: settings.baseURL,

packages/ai/src/providers/openai.ts

@@ -86,15 +86,10 @@ export const configure = (input: Config = {}) => {
   const chatRoute = configuredRoute(OpenAIChat.route, input)
   const modelDefaults = defaults(input)
   const responses = (id: string | ModelID) =>
-    responsesRoute
+    responsesRoute.with(withOpenAIOptions(id, modelDefaults, { textVerbosity: true })).model({ id })
-      .with(withOpenAIOptions(id, modelDefaults, { textVerbosity: true }))
-      .model<OpenAIProviderOptionsInput>({ id })
   const responsesWebSocket = (id: string | ModelID) =>
-    responsesWebSocketRoute
+    responsesWebSocketRoute.with(withOpenAIOptions(id, modelDefaults, { textVerbosity: true })).model({ id })
-      .with(withOpenAIOptions(id, modelDefaults, { textVerbosity: true }))
+  const chat = (id: string | ModelID) => chatRoute.with(withOpenAIOptions(id, modelDefaults)).model({ id })
-      .model<OpenAIProviderOptionsInput>({ id })
-  const chat = (id: string | ModelID) =>
-    chatRoute.with(withOpenAIOptions(id, modelDefaults)).model<OpenAIProviderOptionsInput>({ id })
   const image = (modelID: string | ModelID) =>
     OpenAIImages.model({
       id: modelID,
@@ -137,17 +132,15 @@ const config = (settings: Settings): Config => {
   }
 }
 
-export const model: ProviderPackage.Definition<Settings, OpenAIProviderOptionsInput>["model"] = (modelID, settings) => {
+export const model: ProviderPackage.Definition<Settings>["model"] = (modelID, settings) => {
   const configured = configure(config(settings))
   if (settings.transport === undefined || settings.transport === "http") return configured.responses(modelID)
   if (settings.transport === "websocket") return configured.responsesWebSocket(modelID)
   throw new Error(`Unsupported OpenAI Responses transport: ${String(settings.transport)}`)
 }
 
-export const chatModel: ProviderPackage.Definition<Settings, OpenAIProviderOptionsInput>["model"] = (
+export const chatModel: ProviderPackage.Definition<Settings>["model"] = (modelID, settings) =>
-  modelID,
+  configure(config(settings)).chat(modelID)
-  settings,
-) => configure(config(settings)).chat(modelID)
 export const responses = provider.responses
 export const responsesWebSocket = provider.responsesWebSocket
 export const chat = provider.chat

packages/ai/src/providers/openrouter.ts

@@ -4,72 +4,20 @@ import { Endpoint } from "../route/endpoint"
 import { Framing } from "../route/framing"
 import { Protocol } from "../route/protocol"
 import { AuthOptions, type ProviderAuthOption } from "../route/auth-options"
-import { ProviderID, type CacheHint, type ModelID, type ProviderOptions } from "../schema"
+import { ProviderID, type ModelID, type ProviderOptions } from "../schema"
-import type { ProviderPackage } from "../provider-package"
 import * as OpenAICompatibleProfiles from "./openai-compatible-profile"
 import * as OpenAIChat from "../protocols/openai-chat"
-import { newBreakpoints, ttlBucket } from "../protocols/utils/cache"
 import { isRecord } from "../protocols/shared"
 
 export const profile = OpenAICompatibleProfiles.profiles.openrouter
 export const id = ProviderID.make(profile.provider)
 const ADAPTER = "openrouter"
 
-type OpenRouterString<Known extends string> = Known | (string & {})
-
-export interface OpenRouterProviderRouting {
-  readonly [key: string]: unknown
-  readonly order?: ReadonlyArray<string>
-  readonly allow_fallbacks?: boolean
-  readonly require_parameters?: boolean
-  readonly data_collection?: OpenRouterString<"allow" | "deny">
-  readonly only?: ReadonlyArray<string>
-  readonly ignore?: ReadonlyArray<string>
-  readonly quantizations?: ReadonlyArray<string>
-  readonly sort?: OpenRouterString<"price" | "throughput" | "latency">
-  readonly max_price?: Readonly<{
-    prompt?: number | string
-    completion?: number | string
-    image?: number | string
-    audio?: number | string
-    request?: number | string
-  }>
-  readonly zdr?: boolean
-}
-
-export type OpenRouterPlugin =
-  | Readonly<{
-      id: "web"
-      max_results?: number
-      search_prompt?: string
-      engine?: OpenRouterString<"native" | "exa">
-    }>
-  | Readonly<{ id: "file-parser"; max_files?: number; pdf?: { engine?: string } }>
-  | Readonly<{ id: "moderation" }>
-  | Readonly<{ id: "response-healing" }>
-  | Readonly<{ id: "auto-router"; allowed_models?: ReadonlyArray<string> }>
-  | Readonly<{ id: string & {}; [key: string]: unknown }>
-
 export interface OpenRouterOptions {
   readonly [key: string]: unknown
-  readonly debug?: Readonly<{ echo_upstream_body?: boolean }>
+  readonly usage?: boolean | Record<string, unknown>
-  readonly models?: ReadonlyArray<string>
+  readonly reasoning?: Record<string, unknown>
-  readonly plugins?: ReadonlyArray<OpenRouterPlugin>
   readonly promptCacheKey?: string
-  readonly provider?: OpenRouterProviderRouting
-  readonly reasoning?: Readonly<{
-    enabled?: boolean
-    exclude?: boolean
-    effort?: OpenRouterString<"none" | "minimal" | "low" | "medium" | "high" | "xhigh" | "max">
-    max_tokens?: number
-  }>
-  readonly usage?: boolean | Readonly<{ include: boolean }>
-  readonly user?: string
-  readonly web_search_options?: Readonly<{
-    max_results?: number
-    search_prompt?: string
-    engine?: OpenRouterString<"native" | "exa">
-  }>
 }
 
 export type OpenRouterProviderOptionsInput = ProviderOptions & {
@@ -82,12 +30,6 @@ export type ModelOptions = Omit<RouteDefaultsInput, "providerOptions"> &
     readonly providerOptions?: OpenRouterProviderOptionsInput
   }
 
-export interface Settings extends ProviderPackage.Settings {
-  readonly apiKey?: string
-  readonly baseURL?: string
-  readonly providerOptions?: OpenRouterProviderOptionsInput
-}
-
 const OpenRouterBody = Schema.StructWithRest(Schema.Struct(OpenAIChat.bodyFields), [
   Schema.Record(Schema.String, Schema.Any),
 ])
@@ -98,7 +40,7 @@ export const protocol = Protocol.make({
   body: {
     schema: OpenRouterBody,
     from: (request) =>
-      OpenAIChat.fromRequest(request, { cacheControl: cacheControl() }).pipe(
+      OpenAIChat.protocol.body.from(request).pipe(
         Effect.map((body) => {
           const sourceAssistants = request.messages.filter((message) => message.role === "assistant")
           let assistantIndex = 0
@@ -129,39 +71,16 @@ export const protocol = Protocol.make({
   stream: OpenAIChat.protocol.stream,
 })
 
-const cacheControl = () => {
-  const breakpoints = newBreakpoints(4)
-  return (cache: CacheHint | undefined) => {
-    if (cache === undefined || breakpoints.remaining === 0) return undefined
-    breakpoints.remaining -= 1
-    return {
-      type: "ephemeral" as const,
-      ...(ttlBucket(cache.ttlSeconds) === "1h" ? { ttl: "1h" } : {}),
-    }
-  }
-}
-
 const bodyOptions = (input: unknown) => {
   const openrouter = isRecord(input) ? input : {}
-  const { usage, models, provider, plugins, web_search_options, debug, user, reasoning, promptCacheKey, ...options } =
-    openrouter
   return {
-    ...options,
+    ...(openrouter.usage === true
-    ...(usage === undefined || usage === true
       ? { usage: { include: true } }
-      : usage === false
+      : isRecord(openrouter.usage)
-        ? { usage: { include: false } }
+        ? { usage: openrouter.usage }
-        : isRecord(usage)
+        : {}),
-          ? { usage }
+    ...(isRecord(openrouter.reasoning) ? { reasoning: openrouter.reasoning } : {}),
-          : {}),
+    ...(typeof openrouter.promptCacheKey === "string" ? { prompt_cache_key: openrouter.promptCacheKey } : {}),
-    ...(Array.isArray(models) ? { models } : {}),
-    ...(isRecord(provider) ? { provider } : {}),
-    ...(Array.isArray(plugins) ? { plugins } : {}),
-    ...(isRecord(web_search_options) ? { web_search_options } : {}),
-    ...(isRecord(debug) ? { debug } : {}),
-    ...(typeof user === "string" ? { user } : {}),
-    ...(isRecord(reasoning) ? { reasoning } : {}),
-    ...(typeof promptCacheKey === "string" ? { prompt_cache_key: promptCacheKey } : {}),
   }
 }
 
@@ -188,21 +107,10 @@ export const configure = (input: ModelOptions = {}) => {
   const route = configuredRoute(input)
   return {
     id,
-    model: (modelID: string | ModelID) => route.model<OpenRouterProviderOptionsInput>({ id: modelID }),
+    model: (modelID: string | ModelID) => route.model({ id: modelID }),
     configure,
   }
 }
 
 export const provider = configure()
-export const model: ProviderPackage.Definition<Settings, OpenRouterProviderOptionsInput>["model"] = (
+export const model = provider.model
-  modelID,
-  settings,
-) =>
-  configure({
-    apiKey: settings.apiKey,
-    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

@@ -1,62 +1,29 @@
 import { AuthOptions, type ProviderAuthOption } from "../route/auth-options"
-import { Route, type RouteDefaultsInput } from "../route/client"
+import type { RouteDefaultsInput } from "../route/client"
-import { Endpoint } from "../route/endpoint"
+import { HttpOptions, ProviderID, type ModelID } from "../schema"
-import { HttpOptions, ProviderID, type ModelID, type ProviderOptions } from "../schema"
 import * as OpenAICompatibleProfiles from "./openai-compatible-profile"
 import * as OpenAICompatibleChat from "../protocols/openai-compatible-chat"
-import * as OpenAIChat from "../protocols/openai-chat"
 import * as OpenAIResponses from "../protocols/openai-responses"
 import { XAIImages } from "../protocols/xai-images"
-import type { OpenAIOptionsInput } from "./openai-options"
-import type { ProviderPackage } from "../provider-package"
 
 export const id = ProviderID.make("xai")
 
-export type XAIProviderOptionsInput = ProviderOptions & {
+export type ModelOptions = RouteDefaultsInput &
-  readonly xai?: OpenAIOptionsInput
-}
-
-export type ModelOptions = Omit<RouteDefaultsInput, "providerOptions"> &
   ProviderAuthOption<"optional"> & {
     readonly baseURL?: string
-    readonly providerOptions?: XAIProviderOptionsInput
   }
 
-export interface Settings extends ProviderPackage.Settings {
-  readonly apiKey?: string
-  readonly baseURL?: string
-  readonly providerOptions?: XAIProviderOptionsInput
-}
-
 export type { XAIImageOptions } from "../protocols/xai-images"
 
-const responsesRoute = Route.make({
+export const routes = [OpenAIResponses.route, OpenAICompatibleChat.route]
-  id: "openai-responses",
-  provider: id,
-  providerMetadataKey: "xai",
-  protocol: OpenAIResponses.protocol,
-  endpoint: Endpoint.path("/responses", { baseURL: OpenAICompatibleProfiles.profiles.xai.baseURL }),
-  transport: OpenAIResponses.httpTransport,
-  defaults: { providerOptions: { xai: { store: false } } },
-})
-
-const chatRoute = Route.make({
-  id: "openai-compatible-chat",
-  provider: id,
-  providerMetadataKey: "xai",
-  protocol: OpenAIChat.protocol,
-  endpoint: Endpoint.path("/chat/completions", { baseURL: OpenAICompatibleProfiles.profiles.xai.baseURL }),
-  transport: OpenAICompatibleChat.route.transport,
-})
-
-export const routes = [responsesRoute, chatRoute]
 
 const auth = (options: ProviderAuthOption<"optional">) => AuthOptions.bearer(options, "XAI_API_KEY")
 
 const configuredResponsesRoute = (input: ModelOptions) => {
   const { apiKey: _, auth: _auth, baseURL, ...rest } = input
-  return responsesRoute.with({
+  return OpenAIResponses.route.with({
     ...rest,
+    provider: id,
     endpoint: { baseURL: baseURL ?? OpenAICompatibleProfiles.profiles.xai.baseURL },
     auth: auth(input),
   })
@@ -64,8 +31,9 @@ const configuredResponsesRoute = (input: ModelOptions) => {
 
 const configuredChatRoute = (input: ModelOptions) => {
   const { apiKey: _, auth: _auth, baseURL, ...rest } = input
-  return chatRoute.with({
+  return OpenAICompatibleChat.route.with({
     ...rest,
+    provider: id,
     endpoint: { baseURL: baseURL ?? OpenAICompatibleProfiles.profiles.xai.baseURL },
     auth: auth(input),
   })
@@ -74,8 +42,8 @@ const configuredChatRoute = (input: ModelOptions) => {
 export const configure = (input: ModelOptions = {}) => {
   const responsesRoute = configuredResponsesRoute(input)
   const chatRoute = configuredChatRoute(input)
-  const responses = (modelID: string | ModelID) => responsesRoute.model<XAIProviderOptionsInput>({ id: modelID })
+  const responses = (modelID: string | ModelID) => responsesRoute.model({ id: modelID })
-  const chat = (modelID: string | ModelID) => chatRoute.model<XAIProviderOptionsInput>({ id: modelID })
+  const chat = (modelID: string | ModelID) => chatRoute.model({ id: modelID })
   const image = (modelID: string | ModelID) =>
     XAIImages.model({
       id: modelID,
@@ -95,15 +63,7 @@ export const configure = (input: ModelOptions = {}) => {
 }
 
 export const provider = configure()
-export const model: ProviderPackage.Definition<Settings, XAIProviderOptionsInput>["model"] = (modelID, settings) =>
+export const model = provider.model
-  configure({
-    apiKey: settings.apiKey,
-    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
 export const chat = provider.chat
 export const image = provider.image

packages/ai/src/route/client.ts

@@ -5,12 +5,12 @@ import { Endpoint, type EndpointPatch } from "./endpoint"
 import { RequestExecutor } from "./executor"
 import { Framing } from "./framing"
 import { HttpTransport } from "./transport"
-import type { HttpMiddleware, Transport, TransportRuntime } from "./transport"
+import type { Transport, TransportRuntime } from "./transport"
 import { WebSocketExecutor } from "./transport"
 import type { Protocol } from "./protocol"
 import { applyCachePolicy } from "../cache-policy"
 import * as ProviderShared from "../protocols/shared"
-import type { LLMError, ProtocolID, ProviderOptions } from "../schema"
+import type { LLMError, PreparedRequestOf, ProtocolID, ProviderOptions } from "../schema"
 import {
   GenerationOptions,
   HttpOptions,
@@ -20,6 +20,7 @@ import {
   ModelLimits,
   LLMError as LLMErrorClass,
   LLMEvent,
+  PreparedRequest,
   ProviderID,
   mergeGenerationOptions,
   mergeHttpOptions,
@@ -45,12 +46,8 @@ export interface Route<Body, Prepared = unknown> {
   readonly defaults: RouteDefaults
   readonly body: RouteBody<Body>
   readonly with: (patch: RoutePatch<Body, Prepared>) => Route<Body, Prepared>
-  readonly model: <Options extends ProviderOptions = ProviderOptions>(input: RouteMappedModelInput) => Model<Options>
+  readonly model: (input: RouteMappedModelInput) => Model
-  readonly prepareTransport: (
+  readonly prepareTransport: (body: Body, request: LLMRequest) => Effect.Effect<Prepared, LLMError>
-    body: Body,
-    request: LLMRequest,
-    options?: StreamOptions,
-  ) => Effect.Effect<Prepared, LLMError>
   readonly streamPrepared: (
     prepared: Prepared,
     request: LLMRequest,
@@ -96,15 +93,12 @@ export interface RoutePatch<Body, Prepared> extends RouteDefaultsInput {
 
 type RouteMappedModelInput = RouteModelInput | RouteRoutedModelInput
 
-const makeRouteModel = <Options extends ProviderOptions = ProviderOptions>(
+const makeRouteModel = (route: AnyRoute, mapped: RouteMappedModelInput) => {
-  route: AnyRoute,
-  mapped: RouteMappedModelInput,
-) => {
   const provider = route.provider ?? ("provider" in mapped ? mapped.provider : undefined)
   if (!provider) throw new Error(`Route.model(${route.id}) requires a provider`)
   if (!endpointBaseURL(route.endpoint))
     throw new Error(`Route.model(${route.id}) requires an endpoint baseURL — configure it on the route first`)
-  return Model.make<Options>({
+  return Model.make({
     ...mapped,
     provider,
     route,
@@ -148,20 +142,27 @@ export const httpOptions = (input: HttpOptionsInput | undefined) => {
 }
 
 export interface Interface {
+  /**
+   * Compile a request through protocol body construction, validation, and HTTP
+   * preparation without sending it. Returns the prepared request including the
+   * provider-native body.
+   *
+   * Pass a `Body` type argument to statically expose the route's body
+   * shape (e.g. `prepare<OpenAIChatBody>(...)`) — the runtime body is
+   * identical, so this is a type-level assertion the caller makes about which
+   * route the request will resolve to.
+   */
+  readonly prepare: <Body = unknown>(request: LLMRequest) => Effect.Effect<PreparedRequestOf<Body>, LLMError>
   readonly stream: StreamMethod
   readonly generate: GenerateMethod
 }
 
-export interface StreamOptions {
-  readonly http?: HttpMiddleware
-}
-
 export interface StreamMethod {
-  (request: LLMRequest, options?: StreamOptions): Stream.Stream<LLMEvent, LLMError>
+  (request: LLMRequest): Stream.Stream<LLMEvent, LLMError>
 }
 
 export interface GenerateMethod {
-  (request: LLMRequest, options?: StreamOptions): Effect.Effect<LLMResponse, LLMError>
+  (request: LLMRequest): Effect.Effect<LLMResponse, LLMError>
 }
 
 export class Service extends Context.Service<Service, Interface>()("@opencode/LLMClient") {}
@@ -295,9 +296,8 @@ function makeFromTransport<Body, Prepared, Frame, Event, State>(
           defaults: mergeRouteDefaults(route.defaults, defaults),
         })
       },
-      model: <Options extends ProviderOptions = ProviderOptions>(input: RouteMappedModelInput) =>
+      model: (input) => makeRouteModel(route, input),
-        makeRouteModel<Options>(route, input),
+      prepareTransport: (body, request) =>
-      prepareTransport: (body, request, options) =>
         routeInput.transport.prepare({
           body,
           request,
@@ -305,7 +305,6 @@ function makeFromTransport<Body, Prepared, Frame, Event, State>(
           auth: routeInput.auth ?? Auth.none,
           encodeBody,
           headers: routeInput.headers,
-          middleware: options?.http,
         }),
       streamPrepared: (prepared: Prepared, request: LLMRequest, runtime: TransportRuntime) => {
         const route = `${request.model.provider}/${request.model.route.id}`
@@ -371,14 +370,17 @@ export function make<Body, Prepared, Frame, Event, State>(
   })
 }
 
-const compile = Effect.fn("LLM.compile")(function* (request: LLMRequest, options?: StreamOptions) {
+// `compile` is the important boundary: it turns a common `LLMRequest` into a
+// validated provider body plus transport-private prepared data, but does not
+// execute transport.
+const compile = Effect.fn("LLM.compile")(function* (request: LLMRequest) {
   const resolved = applyCachePolicy(resolveRequestOptions(request))
   const route = resolved.model.route
 
   const body = yield* route.body
     .from(resolved)
     .pipe(Effect.flatMap(ProviderShared.validateWith(Schema.decodeUnknownEffect(route.body.schema))))
-  const prepared = yield* route.prepareTransport(body, resolved, options)
+  const prepared = yield* route.prepareTransport(body, resolved)
 
   return {
     request: resolved,
@@ -388,30 +390,30 @@ const compile = Effect.fn("LLM.compile")(function* (request: LLMRequest, options
   }
 })
 
-/** @internal Test-only projection of the execution compiler; not exported from package barrels. */
+const prepareWith = Effect.fn("LLMClient.prepare")(function* (request: LLMRequest) {
-export const compileRequest = Effect.fn("LLM.compileRequest")(function* (request: LLMRequest) {
   const compiled = yield* compile(request)
-  return {
+
+  return new PreparedRequest({
     id: compiled.request.id ?? "request",
     route: compiled.route.id,
     protocol: compiled.route.protocol,
     model: compiled.request.model,
     body: compiled.body,
     metadata: { transport: compiled.route.transport.id },
-  }
+  })
 })
 
-const streamRequestWith = (runtime: TransportRuntime) => (request: LLMRequest, options?: StreamOptions) =>
+const streamRequestWith = (runtime: TransportRuntime) => (request: LLMRequest) =>
   Stream.unwrap(
     Effect.gen(function* () {
-      const compiled = yield* compile(request, options)
+      const compiled = yield* compile(request)
       return compiled.route.streamPrepared(compiled.prepared, compiled.request, runtime)
     }),
   )
 
 const generateWith = (stream: Interface["stream"]) =>
-  Effect.fn("LLM.generate")(function* (request: LLMRequest, options?: StreamOptions) {
+  Effect.fn("LLM.generate")(function* (request: LLMRequest) {
-    const state = yield* stream(request, options).pipe(Stream.runFold(LLMResponse.empty, LLMResponse.reduce))
+    const state = yield* stream(request).pipe(Stream.runFold(LLMResponse.empty, LLMResponse.reduce))
     const response = LLMResponse.complete(state)
     if (response) return response
     return yield* ProviderShared.eventError(
@@ -420,24 +422,27 @@ const generateWith = (stream: Interface["stream"]) =>
     )
   })
 
-export function stream(request: LLMRequest, options?: StreamOptions): Stream.Stream<LLMEvent, LLMError> {
+export const prepare = <Body = unknown>(request: LLMRequest) =>
+  prepareWith(request) as Effect.Effect<PreparedRequestOf<Body>, LLMError>
+
+export function stream(request: LLMRequest): Stream.Stream<LLMEvent, LLMError> {
   return Stream.unwrap(
     Effect.gen(function* () {
-      return (yield* Service).stream(request, options)
+      return (yield* Service).stream(request)
     }),
   ) as Stream.Stream<LLMEvent, LLMError>
 }
 
-export function generate(request: LLMRequest, options?: StreamOptions): Effect.Effect<LLMResponse, LLMError> {
+export function generate(request: LLMRequest): Effect.Effect<LLMResponse, LLMError> {
   return Effect.gen(function* () {
-    return yield* (yield* Service).generate(request, options)
+    return yield* (yield* Service).generate(request)
   }) as Effect.Effect<LLMResponse, LLMError>
 }
 
-export const streamRequest = (request: LLMRequest, options?: StreamOptions) =>
+export const streamRequest = (request: LLMRequest) =>
   Stream.unwrap(
     Effect.gen(function* () {
-      return (yield* Service).stream(request, options)
+      return (yield* Service).stream(request)
     }),
   )
 
@@ -448,7 +453,7 @@ export const layer: Layer.Layer<Service, never, RequestExecutor.Service> = Layer
       http: yield* RequestExecutor.Service,
       webSocket: Option.getOrUndefined(yield* Effect.serviceOption(WebSocketExecutor.Service)),
     })
-    return Service.of({ stream, generate: generateWith(stream) })
+    return Service.of({ prepare: prepareWith as Interface["prepare"], stream, generate: generateWith(stream) })
   }),
 )
 
@@ -457,6 +462,7 @@ export const Route = { make } as const
 export const LLMClient = {
   Service,
   layer,
+  prepare,
   stream,
   generate,
 } as const

packages/ai/src/route/executor.ts

@@ -1,4 +1,4 @@
-import { Cause, Context, Effect, Layer, Stream } from "effect"
+import { Cause, Context, Effect, Layer } from "effect"
 import {
   FetchHttpClient,
   Headers,
@@ -20,13 +20,9 @@ import { classifyProviderFailure } from "../provider-error"
 export interface Interface {
   readonly execute: (
     request: HttpClientRequest.HttpClientRequest,
-    middleware?: HttpMiddleware,
   ) => Effect.Effect<HttpClientResponse.HttpClientResponse, LLMError>
 }
 
-export type HttpHandler = (request: Request) => Effect.Effect<Response, Error>
-export type HttpMiddleware = (request: Request, handler: HttpHandler) => Effect.Effect<Response, Error>
-
 export class Service extends Context.Service<Service, Interface>()("@opencode/LLM/RequestExecutor") {}
 
 const BODY_LIMIT = 16_384
@@ -286,41 +282,12 @@ export const layer: Layer.Layer<Service, never, HttpClient.HttpClient> = Layer.e
   Service,
   Effect.gen(function* () {
     const http = yield* HttpClient.HttpClient
-    const executeOnce = (request: HttpClientRequest.HttpClientRequest, middleware?: HttpMiddleware) =>
+    const executeOnce = (request: HttpClientRequest.HttpClientRequest) =>
       Effect.gen(function* () {
         const redactedNames = yield* Headers.CurrentRedactedNames
-        if (!middleware)
+        return yield* http
-          return yield* http
+          .execute(request)
-            .execute(request)
+          .pipe(Effect.mapError(toHttpError(redactedNames)), Effect.flatMap(statusError(request, redactedNames)))
-            .pipe(Effect.mapError(toHttpError(redactedNames)), Effect.flatMap(statusError(request, redactedNames)))
-
-        let sent = request
-        const origins = new WeakMap<Response, HttpClientRequest.HttpClientRequest>()
-        const response = yield* HttpClientRequest.toWeb(request).pipe(
-          Effect.flatMap((web) =>
-            middleware(web, (input) =>
-              Effect.gen(function* () {
-                sent = HttpClientRequest.fromWeb(input)
-                if (input.body)
-                  sent = HttpClientRequest.bodyUint8Array(
-                    sent,
-                    new Uint8Array(yield* Effect.promise(() => input.arrayBuffer())),
-                    input.headers.get("content-type") ?? undefined,
-                  )
-                const response = yield* http
-                  .execute(sent)
-                  .pipe(Effect.mapError((cause) => (cause instanceof Error ? cause : new Error(String(cause)))))
-                const body = yield* Stream.toReadableStreamEffect(response.stream)
-                const web = new Response(body, { status: response.status, headers: response.headers })
-                origins.set(web, sent)
-                return web
-              }),
-            ),
-          ),
-          Effect.mapError(toHttpError(redactedNames)),
-        )
-        const origin = origins.get(response) ?? sent
-        return yield* statusError(origin, redactedNames)(HttpClientResponse.fromWeb(origin, response))
       })
     return Service.of({
       execute: executeOnce,

packages/ai/src/route/index.ts

@@ -8,7 +8,6 @@ export type {
   AnyRoute,
   Interface as LLMClientShape,
   Service as LLMClientService,
-  StreamOptions,
 } from "./client"
 export * from "./executor"
 export { Auth } from "./auth"
@@ -23,4 +22,4 @@ export type { ApiKeyMode, AuthOverride, ProviderAuthOption } from "./auth-option
 export type { Definition as EndpointFn, EndpointInput } from "./endpoint"
 export type { Definition as FramingDef } from "./framing"
 export type { Protocol as ProtocolDef } from "./protocol"
-export type { HttpHandler, HttpMiddleware, Transport as TransportDef, TransportRuntime } from "./transport"
+export type { Transport as TransportDef, TransportRuntime } from "./transport"

packages/ai/src/route/transport/http.ts

@@ -3,7 +3,7 @@ import { Headers, HttpClientRequest } from "effect/unstable/http"
 import { Auth } from "../auth"
 import { render as renderEndpoint } from "../endpoint"
 import { Framing } from "../framing"
-import type { HttpMiddleware, Transport, TransportPrepareInput } from "./index"
+import type { Transport, TransportPrepareInput } from "./index"
 import * as ProviderShared from "../../protocols/shared"
 import { mergeJsonRecords, type LLMRequest } from "../../schema"
 
@@ -19,7 +19,6 @@ export interface JsonRequestParts<Body = unknown> {
 export interface HttpPrepared<Frame> {
   readonly request: HttpClientRequest.HttpClientRequest
   readonly framing: Framing.Definition<Frame>
-  readonly middleware?: HttpMiddleware
 }
 
 const applyQuery = (url: string, query: Record<string, string> | undefined) => {
@@ -29,9 +28,57 @@ const applyQuery = (url: string, query: Record<string, string> | undefined) => {
   return next.toString()
 }
 
+const PROTOCOL_BODY_OVERLAY_DENYLIST = new Set([
+  "anthropic_version",
+  "content",
+  "contents",
+  "frequencyPenalty",
+  "frequency_penalty",
+  "generationConfig",
+  "inferenceConfig",
+  "input",
+  "maxTokens",
+  "max_tokens",
+  "messages",
+  "model",
+  "presencePenalty",
+  "presence_penalty",
+  "responseFormat",
+  "response_format",
+  "seed",
+  "stop",
+  "stopSequences",
+  "stop_sequences",
+  "stream",
+  "streamOptions",
+  "stream_options",
+  "system",
+  "systemInstruction",
+  "system_instruction",
+  "temperature",
+  "thinking",
+  "toolChoice",
+  "toolConfig",
+  "tool_choice",
+  "tool_config",
+  "tools",
+  "topK",
+  "topP",
+  "top_k",
+  "top_p",
+])
+
+const forbiddenBodyOverlayKeys = (body: Record<string, unknown>) =>
+  Object.keys(body).filter((key) => PROTOCOL_BODY_OVERLAY_DENYLIST.has(key))
+
 const bodyWithOverlay = <Body>(body: Body, request: LLMRequest, encodeBody: (body: Body) => string) =>
   Effect.gen(function* () {
     if (request.http?.body === undefined) return { jsonBody: body, bodyText: encodeBody(body) }
+    const forbiddenKeys = forbiddenBodyOverlayKeys(request.http.body)
+    if (forbiddenKeys.length > 0)
+      return yield* ProviderShared.invalidRequest(
+        `http.body cannot overlay protocol-owned field(s): ${forbiddenKeys.join(", ")}`,
+      )
     if (ProviderShared.isRecord(body)) {
       const overlaid = mergeJsonRecords(body, request.http.body) ?? {}
       return { jsonBody: overlaid, bodyText: ProviderShared.encodeJson(overlaid) }
@@ -73,23 +120,18 @@ export const httpJson = <Body, Frame>(input: HttpJsonInput<Body, Frame>): HttpJs
   id: "http-json",
   with: (patch) => httpJson({ ...input, ...patch }),
   prepare: (prepareInput) =>
-    Effect.gen(function* () {
+    jsonRequestParts({
-      const parts = yield* jsonRequestParts({ ...prepareInput })
+      ...prepareInput,
-      const request = ProviderShared.jsonPost({
+    }).pipe(
-        url: parts.url,
+      Effect.map((parts) => ({
-        body: parts.bodyText,
+        request: ProviderShared.jsonPost({ url: parts.url, body: parts.bodyText, headers: parts.headers }),
-        headers: parts.headers,
-      })
-      return {
-        request,
         framing: input.framing,
-        middleware: prepareInput.middleware,
+      })),
-      }
+    ),
-    }),
   frames: (prepared, request, runtime) =>
     Stream.unwrap(
       runtime.http
-        .execute(prepared.request, prepared.middleware)
+        .execute(prepared.request)
         .pipe(
           Effect.map((response) =>
             prepared.framing.frame(

packages/ai/src/route/transport/index.ts

@@ -1,7 +1,7 @@
 import type { Effect, Stream } from "effect"
 import { Endpoint } from "../endpoint"
 import { Auth } from "../auth"
-import type { HttpMiddleware, Interface as RequestExecutorInterface } from "../executor"
+import type { Interface as RequestExecutorInterface } from "../executor"
 import type { Interface as WebSocketExecutorInterface } from "./websocket"
 import type { LLMError, LLMRequest } from "../../schema"
 
@@ -27,9 +27,7 @@ export interface TransportPrepareInput<Body> {
   readonly auth: Auth.Definition
   readonly encodeBody: (body: Body) => string
   readonly headers?: (input: { readonly request: LLMRequest }) => Record<string, string>
-  readonly middleware?: HttpMiddleware
 }
 
 export * as HttpTransport from "./http"
-export type { HttpHandler, HttpMiddleware } from "../executor"
 export { WebSocketExecutor, WebSocketTransport } from "./websocket"

packages/ai/src/schema/errors.ts

@@ -2,7 +2,7 @@ import { Schema } from "effect"
 import { Tool } from "@opencode-ai/schema/tool"
 import { ModelID, ProviderID, ProviderMetadata, RouteID } from "./ids"
 
-export const ProviderFailureClassification = Schema.Literals(["context-overflow", "payload-too-large"])
+export const ProviderFailureClassification = Schema.Literal("context-overflow")
 export type ProviderFailureClassification = typeof ProviderFailureClassification.Type
 
 export class HttpRequestDetails extends Schema.Class<HttpRequestDetails>("LLM.HttpRequestDetails")({

packages/ai/src/schema/events.ts

@@ -1,5 +1,6 @@
 import { Schema } from "effect"
-import { ContentBlockID, FinishReason, ProviderMetadata, ToolCallID } from "./ids"
+import { ContentBlockID, FinishReason, ProtocolID, ProviderMetadata, RouteID, ToolCallID } from "./ids"
+import { ModelSchema } from "./options"
 import { Message, ToolCallPart, ToolOutput, ToolResultPart, ToolResultValue, type ContentPart } from "./messages"
 import { ProviderFailureClassification } from "./errors"
 
@@ -313,6 +314,29 @@ export const LLMEvent = Object.assign(llmEventTagged, {
 })
 export type LLMEvent = Schema.Schema.Type<typeof llmEventTagged>
 
+export class PreparedRequest extends Schema.Class<PreparedRequest>("LLM.PreparedRequest")({
+  id: Schema.String,
+  route: RouteID,
+  protocol: ProtocolID,
+  model: ModelSchema,
+  body: Schema.Unknown,
+  metadata: Schema.optional(Schema.Record(Schema.String, Schema.Unknown)),
+}) {}
+
+/**
+ * A `PreparedRequest` whose `body` is typed as `Body`. Use with the generic
+ * on `LLMClient.prepare<Body>(...)` when the caller knows which route their
+ * request will resolve to and wants its native shape statically exposed
+ * (debug UIs, request previews, plan rendering).
+ *
+ * The runtime body is identical — the route still emits `body: unknown` — so
+ * this is a type-level assertion the caller makes about what they expect to
+ * find. The prepare runtime does not validate the assertion.
+ */
+export type PreparedRequestOf<Body> = Omit<PreparedRequest, "body"> & {
+  readonly body: Body
+}
+
 const responseText = (events: ReadonlyArray<LLMEvent>) =>
   events
     .filter(LLMEvent.is.textDelta)

packages/ai/src/schema/messages.ts

@@ -124,7 +124,6 @@ export const ToolCallPart = Object.assign(
     name: Schema.String,
     input: Schema.Unknown,
     providerExecuted: Schema.optional(Schema.Boolean),
-    cache: Schema.optional(CacheHint),
     metadata: Schema.optional(Schema.Record(Schema.String, Schema.Unknown)),
     providerMetadata: Schema.optional(ProviderMetadata),
   }).annotate({ identifier: "LLM.Content.ToolCall" }),
@@ -169,7 +168,6 @@ export const ReasoningPart = Schema.Struct({
   type: Schema.Literal("reasoning"),
   text: Schema.String,
   encrypted: Schema.optional(Schema.String),
-  cache: Schema.optional(CacheHint),
   metadata: Schema.optional(Schema.Record(Schema.String, Schema.Unknown)),
   providerMetadata: Schema.optional(ProviderMetadata),
 }).annotate({ identifier: "LLM.Content.Reasoning" })

packages/ai/src/schema/options.ts

@@ -123,7 +123,6 @@ export const mergeGenerationOptions = (...items: ReadonlyArray<GenerationOptions
 
 export class ModelLimits extends Schema.Class<ModelLimits>("LLM.ModelLimits")({
   context: Schema.optional(Schema.Number),
-  input: Schema.optional(Schema.Number),
   output: Schema.optional(Schema.Number),
 }) {}
 
@@ -167,13 +166,9 @@ export namespace ModelDefaults {
 export const ModelToolSchemaCompatibility = Schema.Literals(["gemini", "moonshot"])
 export type ModelToolSchemaCompatibility = Schema.Schema.Type<typeof ModelToolSchemaCompatibility>
 
-export const ModelMaxTokensFieldCompatibility = Schema.Literals(["max_completion_tokens", "max_tokens"])
-export type ModelMaxTokensFieldCompatibility = Schema.Schema.Type<typeof ModelMaxTokensFieldCompatibility>
-
 export class ModelCompatibility extends Schema.Class<ModelCompatibility>("LLM.ModelCompatibility")({
   toolSchema: Schema.optional(ModelToolSchemaCompatibility),
   reasoningField: Schema.optional(Schema.String),
-  maxTokensField: Schema.optional(ModelMaxTokensFieldCompatibility),
 }) {}
 
 export namespace ModelCompatibility {
@@ -183,8 +178,7 @@ export namespace ModelCompatibility {
   export const make = (input: Input) => (input instanceof ModelCompatibility ? input : new ModelCompatibility(input))
 }
 
-export class Model<Options extends ProviderOptions = ProviderOptions> {
+export class Model {
-  declare protected readonly _ProviderOptions: Options
   readonly id: ModelID
   readonly provider: ProviderID
   readonly route: AnyRoute
@@ -199,8 +193,8 @@ export class Model<Options extends ProviderOptions = ProviderOptions> {
     this.compatibility = input.compatibility
   }
 
-  static make<Options extends ProviderOptions = ProviderOptions>(input: Model.Input) {
+  static make(input: Model.Input) {
-    return new Model<Options>({
+    return new Model({
       id: ModelID.make(input.id),
       provider: ProviderID.make(input.provider),
       route: input.route,
@@ -209,7 +203,7 @@ export class Model<Options extends ProviderOptions = ProviderOptions> {
     })
   }
 
-  static input<Options extends ProviderOptions>(model: Model<Options>): Model.ConstructorInput {
+  static input(model: Model): Model.ConstructorInput {
     return {
       id: model.id,
       provider: model.provider,
@@ -219,9 +213,9 @@ export class Model<Options extends ProviderOptions = ProviderOptions> {
     }
   }
 
-  static update<Options extends ProviderOptions>(model: Model<Options>, patch: Partial<Model.Input>) {
+  static update(model: Model, patch: Partial<Model.Input>) {
     if (Object.keys(patch).length === 0) return model
-    return Model.make<Options>({
+    return Model.make({
       ...Model.input(model),
       ...patch,
     })
@@ -247,8 +241,6 @@ export namespace Model {
 
 export type ModelInput = Model.Input
 
-export type ModelProviderOptions<SelectedModel> = SelectedModel extends Model<infer Options> ? Options : never
-
 export const ModelSchema = Schema.declare((value): value is Model => value instanceof Model, { expected: "LLM.Model" })
 
 export class CacheHint extends Schema.Class<CacheHint>("LLM.CacheHint")({

packages/ai/src/testing.ts

@@ -99,6 +99,7 @@ export const layer = (options: LayerOptions = {}) =>
         )
       }) as LLMClientShape["stream"]
       const client = LLMClient.Service.of({
+        prepare: () => Effect.die("TestLLM does not prepare provider-native requests"),
         stream,
         generate: (request) =>
           stream(request).pipe(

packages/ai/test/adapter.test.ts

@@ -2,7 +2,6 @@ import { describe, expect } from "bun:test"
 import { Effect, Schema, Stream } from "effect"
 import { LLM, LLMRequest, LLMResponse } from "../src"
 import { Route, Endpoint, LLMClient, Protocol, type FramingDef } from "../src/route"
-import { compileRequest } from "../src/route/client"
 import { Model } from "../src/schema"
 import { testEffect } from "./lib/effect"
 import { dynamicResponse } from "./lib/http"
@@ -140,7 +139,8 @@ describe("llm route", () => {
 
   it.effect("selects routes by model route value", () =>
     Effect.gen(function* () {
-      const prepared = yield* compileRequest(
+      const llm = yield* LLMClient.Service
+      const prepared = yield* llm.prepare(
         LLMRequest.update(request, { model: updateModel(request.model, { route: configuredGemini }) }),
       )
 
@@ -173,7 +173,7 @@ describe("llm route", () => {
         framing: fakeFraming,
       })
 
-      const prepared = yield* compileRequest(
+      const prepared = yield* (yield* LLMClient.Service).prepare(
         LLMRequest.update(request, { model: updateModel(request.model, { route: duplicate }) }),
       )
 

packages/ai/test/cache-policy.test.ts

@@ -1,8 +1,7 @@
 import { describe, expect, test } from "bun:test"
 import { Effect } from "effect"
 import { CacheHint, LLM, Message } from "../src"
-import { Auth } from "../src/route"
+import { Auth, LLMClient } from "../src/route"
-import { compileRequest } from "../src/route/client"
 import { AmazonBedrock } from "../src/providers"
 import * as AnthropicMessages from "../src/protocols/anthropic-messages"
 import * as Gemini from "../src/protocols/gemini"
@@ -32,7 +31,7 @@ const geminiModel = Gemini.route
 describe("applyCachePolicy", () => {
   it.effect("undefined cache resolves to 'auto' (the recommended default)", () =>
     Effect.gen(function* () {
-      const prepared = yield* compileRequest(
+      const prepared = yield* LLMClient.prepare(
         LLM.request({
           model: anthropicModel,
           system: "You are concise.",
@@ -51,7 +50,7 @@ describe("applyCachePolicy", () => {
 
   it.effect("'auto' marks the last tool, first and last system parts, and final message boundary on Anthropic", () =>
     Effect.gen(function* () {
-      const prepared = yield* compileRequest(
+      const prepared = yield* LLMClient.prepare(
         LLM.request({
           model: anthropicModel,
           system: [
@@ -88,7 +87,7 @@ describe("applyCachePolicy", () => {
 
   it.effect("'auto' is a no-op on OpenAI (implicit caching protocol)", () =>
     Effect.gen(function* () {
-      const prepared = yield* compileRequest(
+      const prepared = yield* LLMClient.prepare(
         LLM.request({
           model: openaiModel,
           system: "Sys",
@@ -107,7 +106,7 @@ describe("applyCachePolicy", () => {
 
   it.effect("'auto' is a no-op on Gemini (out-of-band caching protocol)", () =>
     Effect.gen(function* () {
-      const prepared = yield* compileRequest(
+      const prepared = yield* LLMClient.prepare(
         LLM.request({
           model: geminiModel,
           system: "Sys",
@@ -124,7 +123,7 @@ describe("applyCachePolicy", () => {
 
   it.effect("'auto' on Bedrock emits cachePoint markers in the right places", () =>
     Effect.gen(function* () {
-      const prepared = yield* compileRequest(
+      const prepared = yield* LLMClient.prepare(
         LLM.request({
           model: bedrockModel,
           system: [
@@ -158,7 +157,7 @@ describe("applyCachePolicy", () => {
 
   it.effect("'none' disables auto placement even when manual hints exist", () =>
     Effect.gen(function* () {
-      const prepared = yield* compileRequest(
+      const prepared = yield* LLMClient.prepare(
         LLM.request({
           model: anthropicModel,
           system: "Sys",
@@ -177,7 +176,7 @@ describe("applyCachePolicy", () => {
 
   it.effect("granular object form: tools-only marks just tools", () =>
     Effect.gen(function* () {
-      const prepared = yield* compileRequest(
+      const prepared = yield* LLMClient.prepare(
         LLM.request({
           model: anthropicModel,
           system: "Sys",
@@ -196,7 +195,7 @@ describe("applyCachePolicy", () => {
 
   it.effect("auto policy preserves manual CacheHints on other parts", () =>
     Effect.gen(function* () {
-      const prepared = yield* compileRequest(
+      const prepared = yield* LLMClient.prepare(
         LLM.request({
           model: anthropicModel,
           system: [
@@ -242,7 +241,7 @@ describe("applyCachePolicy", () => {
       expect("cache" in tail ? tail.cache : undefined).toBeUndefined()
       expect(applyCachePolicy(applied)).toBe(applied)
 
-      const prepared = yield* compileRequest(request)
+      const prepared = yield* LLMClient.prepare(request)
 
       const body = prepared.body as {
         tools: Array<{ cache_control?: unknown }>
@@ -262,7 +261,7 @@ describe("applyCachePolicy", () => {
 
   it.effect("ttlSeconds in the policy flows through to wire markers", () =>
     Effect.gen(function* () {
-      const prepared = yield* compileRequest(
+      const prepared = yield* LLMClient.prepare(
         LLM.request({
           model: anthropicModel,
           system: "Sys",
@@ -279,7 +278,7 @@ describe("applyCachePolicy", () => {
 
   it.effect("messages: { tail: 2 } marks the last 2 message boundaries", () =>
     Effect.gen(function* () {
-      const prepared = yield* compileRequest(
+      const prepared = yield* LLMClient.prepare(
         LLM.request({
           model: anthropicModel,
           messages: [Message.user("u1"), Message.assistant("a1"), Message.user("u2"), Message.assistant("a2")],
@@ -297,7 +296,7 @@ describe("applyCachePolicy", () => {
 
   it.effect("'latest-assistant' marks the last assistant message", () =>
     Effect.gen(function* () {
-      const prepared = yield* compileRequest(
+      const prepared = yield* LLMClient.prepare(
         LLM.request({
           model: anthropicModel,
           messages: [Message.user("u1"), Message.assistant("a1"), Message.user("u2")],

packages/ai/test/executor.test.ts

@@ -85,17 +85,14 @@ describe("RequestExecutor", () => {
     ),
   )
 
-  it.effect("classifies generic HTTP 413 payload errors", () =>
+  it.effect("does not classify generic HTTP 413 payload errors as context overflow", () =>
     Effect.gen(function* () {
       const executor = yield* RequestExecutor.Service
       const error = yield* executor.execute(request).pipe(Effect.flip)
 
       expectLLMError(error)
-      expect(error.reason).toMatchObject({
+      expect(error.reason).toMatchObject({ _tag: "InvalidRequest" })
-        _tag: "InvalidRequest",
+      expect("classification" in error.reason ? error.reason.classification : undefined).toBeUndefined()
-        classification: "payload-too-large",
-        http: { response: { status: 413 } },
-      })
     }).pipe(Effect.provide(responsesLayer([new Response("request too large", { status: 413 })]))),
   )
 

packages/ai/test/exports.test.ts

@@ -53,7 +53,9 @@ describe("public exports", () => {
     expect(CloudflareWorkersAI.configure).toBeFunction()
     expect(CloudflareWorkersAI.configure({ accountId: "fixture", apiKey: "fixture" }).model).toBeFunction()
     expect(OpenRouter.model).toBeFunction()
+    expect(OpenRouter.provider.model).toBe(OpenRouter.model)
     expect(XAI.model).toBeFunction()
+    expect(XAI.provider.model).toBe(XAI.model)
     expect(XAI.provider.responses).toBe(XAI.responses)
     expect(XAI.provider.chat).toBe(XAI.chat)
     expect(XAI.configure({ apiKey: "fixture" }).responses("grok-4.3").route.id).toBe("openai-responses")

packages/ai/test/llm-option-types.types.ts

@@ -1,47 +0,0 @@
-import { Schema } from "effect"
-import { LLM, type Model, type ModelProviderOptions, type ProviderOptions } from "../src"
-import { OpenAIChat } from "../src/protocols"
-
-interface ExampleOptions {
-  readonly [key: string]: unknown
-  readonly mode?: "fast" | "thorough"
-}
-
-type ExampleProviderOptions = ProviderOptions & {
-  readonly example?: ExampleOptions
-}
-
-const model = OpenAIChat.route
-  .with({ endpoint: { baseURL: "https://example.com/v1" } })
-  .model<ExampleProviderOptions>({ id: "example" })
-
-LLM.request({ model, prompt: "Hello", providerOptions: { example: { mode: "fast" } } })
-LLM.request({ model, prompt: "Hello", providerOptions: { future: { option: true } } })
-
-LLM.request({
-  model,
-  prompt: "Hello",
-  // @ts-expect-error Known provider options preserve their value types.
-  providerOptions: { example: { mode: "slow" } },
-})
-
-LLM.generateObject({
-  model,
-  prompt: "Hello",
-  schema: Schema.Struct({ answer: Schema.String }),
-  providerOptions: { example: { mode: "thorough" } },
-})
-
-LLM.generateObject({
-  model,
-  prompt: "Hello",
-  jsonSchema: { type: "object" },
-  // @ts-expect-error Dynamic object generation uses the selected model's provider options.
-  providerOptions: { example: { mode: false } },
-})
-
-declare const generic: Model
-LLM.request({ model: generic, prompt: "Hello", providerOptions: { arbitrary: { option: true } } })
-
-const options: ModelProviderOptions<typeof model> = { example: { mode: "fast" } }
-void options

packages/ai/test/prepare.test.ts

@@ -1,10 +1,9 @@
 import { describe, expect, test } from "bun:test"
-import { Effect, Ref, Schema } from "effect"
+import { Effect, Schema } from "effect"
 import { HttpClientRequest } from "effect/unstable/http"
 import { LLM, mergeProviderOptions } from "../src"
 import { AnthropicMessages, OpenAIChat } from "../src/protocols"
 import { Auth, LLMClient } from "../src/route"
-import { compileRequest } from "../src/route/client"
 import { it } from "./lib/effect"
 import { dynamicResponse } from "./lib/http"
 import { deltaChunk } from "./lib/openai-chunks"
@@ -45,7 +44,7 @@ describe("request option precedence", () => {
     })
   })
 
-  it.effect("compiles bodies with route defaults, model defaults, and call options in order", () =>
+  it.effect("prepares bodies with route defaults, model defaults, and call options in order", () =>
     Effect.gen(function* () {
       const route = OpenAIChat.route.with({
         endpoint: { baseURL: "https://api.openai.test/v1/" },
@@ -60,7 +59,7 @@ describe("request option precedence", () => {
           providerOptions: { openai: { reasoningEffort: "medium" } },
         },
       })
-      const prepared = yield* compileRequest(
+      const prepared = yield* LLMClient.prepare<OpenAIChat.OpenAIChatBody>(
         LLM.request({
           model,
           prompt: "Say hello.",
@@ -137,149 +136,26 @@ describe("request option precedence", () => {
     ),
   )
 
-  it.effect("transforms the final HTTP request after serialization and authentication", () =>
+  it.effect("rejects raw body overlays for protocol-owned roots", () =>
-    LLMClient.generate(
-      LLM.request({
-        model: OpenAIChat.route
-          .with({ endpoint: { baseURL: "https://api.openai.test/v1/" }, auth: Auth.bearer("fresh-key") })
-          .model({ id: "gpt-4o-mini" }),
-        prompt: "Say hello.",
-      }),
-      {
-        http: (request, handler) =>
-          Effect.gen(function* () {
-            expect(request.headers.get("authorization")).toBe("Bearer fresh-key")
-            const headers = new Headers(request.headers)
-            headers.set("x-plugin", "transformed")
-            headers.set("content-type", "application/custom+json")
-            return yield* handler(
-              new Request("https://proxy.test/v1/chat/completions", {
-                method: "PUT",
-                headers,
-                body: JSON.stringify({ transformed: true }),
-              }),
-            )
-          }),
-      },
-    ).pipe(
-      Effect.provide(
-        dynamicResponse((input) =>
-          Effect.gen(function* () {
-            const web = yield* HttpClientRequest.toWeb(input.request).pipe(Effect.orDie)
-            expect(web.url).toBe("https://proxy.test/v1/chat/completions")
-            expect(web.method).toBe("PUT")
-            expect(web.headers.get("x-plugin")).toBe("transformed")
-            expect(web.headers.get("content-type")).toBe("application/custom+json")
-            expect(decodeJson(input.text)).toEqual({ transformed: true })
-            return input.respond(sseEvents(deltaChunk({}, "stop")), {
-              headers: { "content-type": "text/event-stream" },
-            })
-          }),
-        ),
-      ),
-    ),
-  )
-
-  it.effect("transforms the HTTP response before protocol decoding", () =>
     Effect.gen(function* () {
-      const response = yield* LLMClient.generate(
+      const model = OpenAIChat.route
+        .with({ endpoint: { baseURL: "https://api.openai.test/v1/" }, auth: Auth.bearer("test") })
+        .model({ id: "gpt-4o-mini" })
+      const error = yield* LLMClient.prepare(
         LLM.request({
-          model: OpenAIChat.route
+          model,
-            .with({ endpoint: { baseURL: "https://api.openai.test/v1/" }, auth: Auth.bearer("test") })
-            .model({ id: "gpt-4o-mini" }),
           prompt: "Say hello.",
+          http: { body: { model: "gpt-5", messages: [], tools: [] } },
         }),
-        {
+      ).pipe(Effect.flip)
-          http: (request, handler) =>
-            Effect.gen(function* () {
-              const response = yield* handler(request)
-              const body = yield* Effect.promise(() => response.text())
-              return new Response(body.replace("network", "hooked"), {
-                status: response.status,
-                headers: response.headers,
-              })
-            }),
-        },
-      ).pipe(
-        Effect.provide(
-          dynamicResponse((input) =>
-            Effect.succeed(
-              input.respond(sseEvents(deltaChunk({ content: "network" }, "stop")), {
-                headers: { "content-type": "text/event-stream" },
-              }),
-            ),
-          ),
-        ),
-      )
 
-      expect(response.text).toBe("hooked")
+      expect(error.reason).toMatchObject({
+        _tag: "InvalidRequest",
+        message: "http.body cannot overlay protocol-owned field(s): model, messages, tools",
+      })
     }),
   )
 
-  it.effect("can inspect an error response and retry the native request", () =>
-    Effect.gen(function* () {
-      const attempts = yield* Ref.make(0)
-      const response = yield* LLMClient.generate(
-        LLM.request({
-          model: OpenAIChat.route
-            .with({ endpoint: { baseURL: "https://api.openai.test/v1/" }, auth: Auth.bearer("stale") })
-            .model({ id: "gpt-4o-mini" }),
-          prompt: "Say hello.",
-        }),
-        {
-          http: (request, handler) =>
-            Effect.gen(function* () {
-              const retry = request.clone()
-              const response = yield* handler(request)
-              expect(response.status).toBe(401)
-              const headers = new Headers(retry.headers)
-              headers.set("authorization", "Bearer refreshed")
-              return yield* handler(new Request(retry, { headers }))
-            }),
-        },
-      ).pipe(
-        Effect.provide(
-          dynamicResponse((input) =>
-            Effect.gen(function* () {
-              yield* Ref.update(attempts, (value) => value + 1)
-              if (input.request.headers.authorization !== "Bearer refreshed")
-                return input.respond("unauthorized", { status: 401 })
-              return input.respond(sseEvents(deltaChunk({ content: "retried" }, "stop")), {
-                headers: { "content-type": "text/event-stream" },
-              })
-            }),
-          ),
-        ),
-      )
-
-      expect(response.text).toBe("retried")
-      expect(yield* Ref.get(attempts)).toBe(2)
-    }),
-  )
-
-  it.effect("applies raw body overlays after protocol lowering", () =>
-    LLMClient.generate(
-      LLM.request({
-        model: OpenAIChat.route
-          .with({ endpoint: { baseURL: "https://api.openai.test/v1/" }, auth: Auth.bearer("test") })
-          .model({ id: "gpt-4o-mini" }),
-        prompt: "Say hello.",
-        http: { body: { model: "gpt-5", messages: [], tools: [] } },
-      }),
-    ).pipe(
-      Effect.provide(
-        dynamicResponse((input) =>
-          Effect.gen(function* () {
-            expect(decodeJson(input.text)).toMatchObject({ model: "gpt-5", messages: [], tools: [] })
-            return input.respond(sseEvents(deltaChunk({}, "stop")), {
-              headers: { "content-type": "text/event-stream" },
-            })
-          }),
-        ),
-      ),
-    ),
-  )
-
   it.effect("uses model output limits after route limits and before call maxTokens", () =>
     Effect.gen(function* () {
       const route = AnthropicMessages.route.with({
@@ -288,8 +164,10 @@ describe("request option precedence", () => {
         limits: { output: 128 },
       })
       const model = route.model({ id: "claude-sonnet-4-5", defaults: { limits: { output: 64 } } })
-      const withoutMaxTokens = yield* compileRequest(LLM.request({ model, prompt: "Say hello.", cache: "none" }))
+      const withoutMaxTokens = yield* LLMClient.prepare<AnthropicMessages.AnthropicMessagesBody>(
-      const withMaxTokens = yield* compileRequest(
+        LLM.request({ model, prompt: "Say hello.", cache: "none" }),
+      )
+      const withMaxTokens = yield* LLMClient.prepare<AnthropicMessages.AnthropicMessagesBody>(
         LLM.request({ model, prompt: "Say hello.", cache: "none", generation: { maxTokens: 32 } }),
       )
 

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

@@ -6,6 +6,7 @@ describe("provider error classification", () => {
   test("classifies provider token limit messages as context overflow", () => {
     const messages = [
       "tokens in request more than max tokens allowed",
+      '{"error":{"type":"request_too_large","message":"Request exceeds the maximum size"}}',
       "Requested token count exceeds the model's maximum context length of 131072 tokens.",
       "Input length (265330) exceeds model's maximum context length (262144).",
       "Input length 131393 exceeds the maximum allowed input length of 131040 tokens.",
@@ -18,24 +19,6 @@ describe("provider error classification", () => {
     expect(messages.every(isContextOverflow)).toBe(true)
   })
 
-  test("classifies request size failures separately from context overflow", () => {
-    const failures = [
-      classifyProviderFailure({ message: "request too large", status: 413 }),
-      classifyProviderFailure({
-        message: '{"error":{"type":"request_too_large","message":"Request exceeds the maximum size"}}',
-        status: 400,
-      }),
-      classifyProviderFailure({ message: "upstream request entity too large", status: 502 }),
-    ]
-
-    expect(failures).toEqual(
-      failures.map((failure) =>
-        expect.objectContaining({ _tag: "InvalidRequest", classification: "payload-too-large" }),
-      ),
-    )
-    expect(isContextOverflow("413 status code (no body)")).toBe(false)
-  })
-
   test("does not classify rate limits as context overflow", () => {
     const messages = [
       "Throttling error: Too many tokens, please wait before trying again.",
@@ -76,10 +59,9 @@ describe("provider error classification", () => {
   })
 
   test("classifies transient client statuses as provider internal", () => {
-    expect([408, 409].map((status) => classifyProviderFailure({ message: `HTTP ${status}`, status })._tag)).toEqual([
+    expect(
-      "ProviderInternal",
+      [408, 409].map((status) => classifyProviderFailure({ message: `HTTP ${status}`, status })._tag),
-      "ProviderInternal",
+    ).toEqual(["ProviderInternal", "ProviderInternal"])
-    ])
   })
 
   test("classifies nested provider codes when a top-level code is also present", () => {

packages/ai/test/provider-options/anthropic-compatible.types.ts

@@ -1,13 +0,0 @@
-import { LLM } from "../../src"
-import { AnthropicCompatible } from "../../src/providers"
-
-const model = AnthropicCompatible.configure({ baseURL: "https://example.com" }).model("claude")
-
-LLM.request({ model, prompt: "Hello", providerOptions: { anthropic: { effort: "high" } } })
-
-LLM.request({
-  model,
-  prompt: "Hello",
-  // @ts-expect-error Anthropic effort must be a string.
-  providerOptions: { anthropic: { effort: 1 } },
-})

packages/ai/test/provider-options/anthropic.types.ts

@@ -1,13 +0,0 @@
-import { LLM } from "../../src"
-import { Anthropic } from "../../src/providers"
-
-const model = Anthropic.provider.model("claude-sonnet-4-5")
-
-LLM.request({ model, prompt: "Hello", providerOptions: { anthropic: { thinking: { type: "adaptive" } } } })
-
-LLM.request({
-  model,
-  prompt: "Hello",
-  // @ts-expect-error Anthropic thinking modes are a fixed union.
-  providerOptions: { anthropic: { thinking: { type: "automatic" } } },
-})

packages/ai/test/provider-options/azure.types.ts

@@ -1,13 +0,0 @@
-import { LLM } from "../../src"
-import { Azure } from "../../src/providers"
-
-const model = Azure.configure({ resourceName: "example" }).responses("deployment")
-
-LLM.request({ model, prompt: "Hello", providerOptions: { openai: { store: false } } })
-
-LLM.request({
-  model,
-  prompt: "Hello",
-  // @ts-expect-error Azure OpenAI store must be boolean.
-  providerOptions: { openai: { store: "false" } },
-})

packages/ai/test/provider-options/cloudflare.types.ts

@@ -1,13 +0,0 @@
-import { LLM } from "../../src"
-import { CloudflareWorkersAI } from "../../src/providers"
-
-const model = CloudflareWorkersAI.configure({ accountId: "account", apiKey: "test" }).model("model")
-
-LLM.request({ model, prompt: "Hello", providerOptions: { openai: { promptCacheKey: "cache" } } })
-
-LLM.request({
-  model,
-  prompt: "Hello",
-  // @ts-expect-error Cloudflare's OpenAI-compatible prompt cache key must be a string.
-  providerOptions: { openai: { promptCacheKey: 1 } },
-})

packages/ai/test/provider-options/github-copilot.types.ts

@@ -1,13 +0,0 @@
-import { LLM } from "../../src"
-import { GitHubCopilot } from "../../src/providers"
-
-const model = GitHubCopilot.configure({ baseURL: "https://example.com" }).model("gpt-5")
-
-LLM.request({ model, prompt: "Hello", providerOptions: { openai: { reasoningSummary: "auto" } } })
-
-LLM.request({
-  model,
-  prompt: "Hello",
-  // @ts-expect-error Copilot reasoning summaries use the OpenAI union.
-  providerOptions: { openai: { reasoningSummary: "full" } },
-})

packages/ai/test/provider-options/google-vertex-chat.types.ts

@@ -1,13 +0,0 @@
-import { LLM } from "../../src"
-import { GoogleVertexChat } from "../../src/providers"
-
-const model = GoogleVertexChat.configure({ accessToken: "test", project: "project" }).model("gemini")
-
-LLM.request({ model, prompt: "Hello", providerOptions: { openai: { serviceTier: "priority" } } })
-
-LLM.request({
-  model,
-  prompt: "Hello",
-  // @ts-expect-error Vertex OpenAI-compatible service tiers use the OpenAI union.
-  providerOptions: { openai: { serviceTier: "premium" } },
-})

packages/ai/test/provider-options/google-vertex-messages.types.ts

@@ -1,13 +0,0 @@
-import { LLM } from "../../src"
-import { GoogleVertexMessages } from "../../src/providers"
-
-const model = GoogleVertexMessages.configure({ accessToken: "test", project: "project" }).model("claude")
-
-LLM.request({ model, prompt: "Hello", providerOptions: { anthropic: { effort: "medium" } } })
-
-LLM.request({
-  model,
-  prompt: "Hello",
-  // @ts-expect-error Vertex Anthropic effort must be a string.
-  providerOptions: { anthropic: { effort: false } },
-})

packages/ai/test/provider-options/google-vertex-responses.types.ts

@@ -1,13 +0,0 @@
-import { LLM } from "../../src"
-import { GoogleVertexResponses } from "../../src/providers"
-
-const model = GoogleVertexResponses.configure({ accessToken: "test", project: "project" }).model("gemini")
-
-LLM.request({ model, prompt: "Hello", providerOptions: { openresponses: { textVerbosity: "high" } } })
-
-LLM.request({
-  model,
-  prompt: "Hello",
-  // @ts-expect-error Vertex Responses verbosity uses the Open Responses union.
-  providerOptions: { openresponses: { textVerbosity: "verbose" } },
-})

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

@@ -1,17 +0,0 @@
-import { LLM } from "../../src"
-import { GoogleVertex } from "../../src/providers"
-
-const model = GoogleVertex.provider.configure({ apiKey: "test" }).model("gemini-2.5-pro")
-
-LLM.request({
-  model,
-  prompt: "Hello",
-  providerOptions: { gemini: { thinkingConfig: { includeThoughts: true } } },
-})
-
-LLM.request({
-  model,
-  prompt: "Hello",
-  // @ts-expect-error Vertex Gemini includeThoughts must be boolean.
-  providerOptions: { gemini: { thinkingConfig: { includeThoughts: "yes" } } },
-})

packages/ai/test/provider-options/google.types.ts

@@ -1,47 +0,0 @@
-import { LLM } from "../../src"
-import { Google } from "../../src/providers"
-
-const model = Google.provider.model("gemini-2.5-pro")
-
-LLM.request({
-  model,
-  prompt: "Hello",
-  providerOptions: { gemini: { thinkingConfig: { thinkingBudget: 1024 } } },
-})
-
-LLM.request({
-  model,
-  prompt: "Hello",
-  providerOptions: {
-    gemini: {
-      // @ts-expect-error Gemini safety settings require a threshold for every category.
-      safetySettings: [{ category: "HARM_CATEGORY_HATE_SPEECH" }],
-    },
-  },
-})
-
-LLM.request({
-  model,
-  prompt: "Hello",
-  providerOptions: {
-    gemini: {
-      cachedContent: "cachedContents/example",
-      safetySettings: [{ category: "HARM_CATEGORY_HATE_SPEECH", threshold: "BLOCK_ONLY_HIGH" }],
-      serviceTier: "future-tier",
-      thinkingConfig: { thinkingLevel: "high", includeThoughts: true },
-    },
-  },
-})
-
-LLM.request({
-  model,
-  prompt: "Hello",
-  // @ts-expect-error Gemini thinking budgets must be numeric.
-  providerOptions: { gemini: { thinkingConfig: { thinkingBudget: "large" } } },
-})
-
-LLM.request({
-  model,
-  prompt: "Hello",
-  providerOptions: { gemini: { thinkingConfig: { thinkingLevel: "maximum" } } },
-})

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

@@ -1,13 +0,0 @@
-import { LLM } from "../../src"
-import { OpenAICompatibleResponses } from "../../src/providers"
-
-const model = OpenAICompatibleResponses.configure({ baseURL: "https://example.com" }).model("model")
-
-LLM.request({ model, prompt: "Hello", providerOptions: { openresponses: { reasoningSummary: "detailed" } } })
-
-LLM.request({
-  model,
-  prompt: "Hello",
-  // @ts-expect-error Open Responses reasoning summaries use a fixed union.
-  providerOptions: { openresponses: { reasoningSummary: "full" } },
-})

packages/ai/test/provider-options/openai-compatible.types.ts

@@ -1,13 +0,0 @@
-import { LLM } from "../../src"
-import { OpenAICompatible } from "../../src/providers"
-
-const model = OpenAICompatible.deepseek.model("deepseek-chat")
-
-LLM.request({ model, prompt: "Hello", providerOptions: { openai: { store: false } } })
-
-LLM.request({
-  model,
-  prompt: "Hello",
-  // @ts-expect-error OpenAI-compatible store must be boolean.
-  providerOptions: { openai: { store: "false" } },
-})

packages/ai/test/provider-options/openai.types.ts

@@ -1,13 +0,0 @@
-import { LLM } from "../../src"
-import { OpenAI } from "../../src/providers"
-
-const model = OpenAI.responses("gpt-5")
-
-LLM.request({ model, prompt: "Hello", providerOptions: { openai: { reasoningEffort: "high" } } })
-
-LLM.request({
-  model,
-  prompt: "Hello",
-  // @ts-expect-error OpenAI reasoning effort must be a string.
-  providerOptions: { openai: { reasoningEffort: 1 } },
-})

packages/ai/test/provider-options/openrouter.types.ts

@@ -1,35 +0,0 @@
-import { LLM } from "../../src"
-import { OpenRouter } from "../../src/providers"
-
-const model = OpenRouter.provider.model("anthropic/claude-sonnet-4.5")
-
-LLM.request({ model, prompt: "Hello", providerOptions: { openrouter: { usage: true } } })
-
-LLM.request({
-  model,
-  prompt: "Hello",
-  providerOptions: {
-    openrouter: {
-      models: ["google/gemini-3.1-pro"],
-      provider: {
-        order: ["anthropic"],
-        require_parameters: true,
-        data_collection: "future-policy",
-        sort: "future-sort",
-        max_price: { prompt: "0.50" },
-      },
-      reasoning: { effort: "future-effort", exclude: false },
-      plugins: [{ id: "future-plugin", enabled: true }],
-      web_search_options: { engine: "future-engine" },
-      debug: { echo_upstream_body: true },
-      user: "user_123",
-    },
-  },
-})
-
-LLM.request({
-  model,
-  prompt: "Hello",
-  // @ts-expect-error OpenRouter usage must be boolean or an option record.
-  providerOptions: { openrouter: { usage: "yes" } },
-})

packages/ai/test/provider-options/xai.types.ts

@@ -1,13 +0,0 @@
-import { LLM } from "../../src"
-import { XAI } from "../../src/providers"
-
-const model = XAI.provider.model("grok-4")
-
-LLM.request({ model, prompt: "Hello", providerOptions: { xai: { reasoningEffort: "high" } } })
-
-LLM.request({
-  model,
-  prompt: "Hello",
-  // @ts-expect-error xAI's OpenAI-compatible reasoning effort must be a string.
-  providerOptions: { xai: { reasoningEffort: true } },
-})

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

@@ -21,8 +21,6 @@ describe("provider package entrypoints", () => {
       import("@opencode-ai/ai/providers/google-vertex/chat"),
       import("@opencode-ai/ai/providers/google-vertex/responses"),
       import("@opencode-ai/ai/providers/google-vertex/messages"),
-      import("@opencode-ai/ai/providers/openrouter"),
-      import("@opencode-ai/ai/providers/xai"),
     ])
 
     for (const module of modules) expect(module.model).toBeFunction()
@@ -31,35 +29,6 @@ describe("provider package entrypoints", () => {
     expect(modules[12].model).toBe(modules[13].model)
   })
 
-  test("maps OpenRouter and xAI package settings onto executable models", async () => {
-    const OpenRouter = await import("@opencode-ai/ai/providers/openrouter")
-    const XAI = await import("@opencode-ai/ai/providers/xai")
-    const settings = {
-      apiKey: "fixture",
-      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,
-      providerOptions: { openrouter: { usage: true } },
-    })
-    const xai = XAI.model("grok-4", {
-      ...settings,
-      providerOptions: { xai: { reasoningEffort: "high" } },
-    })
-
-    for (const selected of [openrouter, xai]) {
-      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({ openrouter: { usage: true } })
-    expect(xai.route.defaults.providerOptions).toMatchObject({ xai: { reasoningEffort: "high", store: false } })
-  })
-
   test("maps package settings onto the executable model", () => {
     const selected = model("gpt-5", {
       apiKey: "fixture",

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

@@ -3,7 +3,6 @@ import { Effect } from "effect"
 import { HttpClientRequest } from "effect/unstable/http"
 import { CacheHint, LLM, LLMError, LLMRequest, Message, ToolCallPart, ToolDefinition, Usage } from "../../src"
 import { Auth, LLMClient } from "../../src/route"
-import { compileRequest } from "../../src/route/client"
 import * as AnthropicMessages from "../../src/protocols/anthropic-messages"
 import { continuationRequest, nativeAnthropicMessagesContinuation } from "../continuation-scenarios"
 import { it } from "../lib/effect"
@@ -45,7 +44,7 @@ const expectToolResult = (body: AnthropicMessages.AnthropicMessagesBody): Anthro
 describe("Anthropic Messages route", () => {
   it.effect("prepares Anthropic Messages target", () =>
     Effect.gen(function* () {
-      const prepared = yield* compileRequest(request)
+      const prepared = yield* LLMClient.prepare(request)
 
       expect(prepared.body).toEqual({
         model: "claude-sonnet-4-5",
@@ -60,7 +59,7 @@ describe("Anthropic Messages route", () => {
 
   it.effect("lowers adaptive thinking settings with effort", () =>
     Effect.gen(function* () {
-      const prepared = yield* compileRequest(
+      const prepared = yield* LLMClient.prepare<AnthropicMessages.AnthropicMessagesBody>(
         LLMRequest.update(request, {
           providerOptions: {
             anthropic: { thinking: { type: "adaptive", display: "summarized" }, effort: "low" },
@@ -77,17 +76,17 @@ describe("Anthropic Messages route", () => {
 
   it.effect("normalizes enabled and disabled thinking settings", () =>
     Effect.gen(function* () {
-      const enabled = yield* compileRequest(
+      const enabled = yield* LLMClient.prepare<AnthropicMessages.AnthropicMessagesBody>(
         LLMRequest.update(request, {
           providerOptions: { anthropic: { thinking: { type: "enabled", budgetTokens: 1_024 } } },
         }),
       )
-      const legacy = yield* compileRequest(
+      const legacy = yield* LLMClient.prepare<AnthropicMessages.AnthropicMessagesBody>(
         LLMRequest.update(request, {
           providerOptions: { anthropic: { thinking: { type: "enabled", budget_tokens: 2_048 } } },
         }),
       )
-      const disabled = yield* compileRequest(
+      const disabled = yield* LLMClient.prepare<AnthropicMessages.AnthropicMessagesBody>(
         LLMRequest.update(request, {
           providerOptions: { anthropic: { thinking: { type: "disabled" } } },
         }),
@@ -101,7 +100,7 @@ describe("Anthropic Messages route", () => {
 
   it.effect("rejects enabled thinking without a budget", () =>
     Effect.gen(function* () {
-      const error = yield* compileRequest(
+      const error = yield* LLMClient.prepare(
         LLMRequest.update(request, {
           providerOptions: { anthropic: { thinking: { type: "enabled" } } },
         }),
@@ -113,7 +112,7 @@ describe("Anthropic Messages route", () => {
 
   it.effect("lowers chronological system updates natively for Claude Opus 4.8 with cache hints", () =>
     Effect.gen(function* () {
-      const prepared = yield* compileRequest(
+      const prepared = yield* LLMClient.prepare<AnthropicMessages.AnthropicMessagesBody>(
         LLM.request({
           model: opus48,
           messages: [
@@ -138,7 +137,7 @@ describe("Anthropic Messages route", () => {
 
   it.effect("lowers chronological system updates to wrapped user text for unsupported Anthropic models", () =>
     Effect.gen(function* () {
-      const prepared = yield* compileRequest(
+      const prepared = yield* LLMClient.prepare<AnthropicMessages.AnthropicMessagesBody>(
         LLM.request({
           model,
           messages: [
@@ -165,7 +164,7 @@ describe("Anthropic Messages route", () => {
 
   it.effect("rejects non-text chronological system update content before send", () =>
     Effect.gen(function* () {
-      const error = yield* compileRequest(
+      const error = yield* LLMClient.prepare(
         LLM.request({
           model: opus48,
           messages: [
@@ -182,7 +181,7 @@ describe("Anthropic Messages route", () => {
   it.effect("falls back for unsupported native chronological system update placement", () =>
     Effect.gen(function* () {
       expect(
-        (yield* compileRequest(
+        (yield* LLMClient.prepare<AnthropicMessages.AnthropicMessagesBody>(
           LLM.request({
             model: opus48,
             messages: [Message.assistant("Plain."), Message.system("After plain assistant.")],
@@ -197,11 +196,12 @@ describe("Anthropic Messages route", () => {
         },
       ])
       expect(
-        (yield* compileRequest(LLM.request({ model: opus48, messages: [Message.system("First.")], cache: "none" })))
+        (yield* LLMClient.prepare<AnthropicMessages.AnthropicMessagesBody>(
-          .body.messages,
+          LLM.request({ model: opus48, messages: [Message.system("First.")], cache: "none" }),
+        )).body.messages,
       ).toEqual([{ role: "user", content: [{ type: "text", text: "<system-update>\nFirst.\n</system-update>" }] }])
       expect(
-        (yield* compileRequest(
+        (yield* LLMClient.prepare<AnthropicMessages.AnthropicMessagesBody>(
           LLM.request({
             model: opus48,
             messages: [Message.user("Before."), Message.system("One."), Message.system("Two.")],
@@ -223,7 +223,7 @@ describe("Anthropic Messages route", () => {
 
   it.effect("rejects a system update between a local tool call and its result", () =>
     Effect.gen(function* () {
-      const error = yield* compileRequest(
+      const error = yield* LLMClient.prepare(
         LLM.request({
           model: opus48,
           messages: [
@@ -242,7 +242,7 @@ describe("Anthropic Messages route", () => {
 
   it.effect("prepares tool call and tool result messages", () =>
     Effect.gen(function* () {
-      const prepared = yield* compileRequest(
+      const prepared = yield* LLMClient.prepare<AnthropicMessages.AnthropicMessagesBody>(
         LLM.request({
           id: "req_tool_result",
           model,
@@ -273,7 +273,7 @@ describe("Anthropic Messages route", () => {
 
   it.effect("keeps tools and sends tool_choice none", () =>
     Effect.gen(function* () {
-      const prepared = yield* compileRequest(
+      const prepared = yield* LLMClient.prepare<AnthropicMessages.AnthropicMessagesBody>(
         LLM.request({
           id: "req_tool_choice_none",
           model,
@@ -303,7 +303,7 @@ describe("Anthropic Messages route", () => {
   // not JSON-stringified into `tool_result.content`.
   it.effect("lowers media tool-result content as structured blocks", () =>
     Effect.gen(function* () {
-      const prepared = yield* compileRequest(
+      const prepared = yield* LLMClient.prepare<AnthropicMessages.AnthropicMessagesBody>(
         LLM.request({
           id: "req_tool_result_image",
           model,
@@ -335,7 +335,7 @@ describe("Anthropic Messages route", () => {
 
   it.effect("lowers single-image tool-result content as a structured image block", () =>
     Effect.gen(function* () {
-      const prepared = yield* compileRequest(
+      const prepared = yield* LLMClient.prepare<AnthropicMessages.AnthropicMessagesBody>(
         LLM.request({
           id: "req_tool_result_image_only",
           model,
@@ -360,7 +360,7 @@ describe("Anthropic Messages route", () => {
 
   it.effect("rejects unsupported media in tool-result content with a clear error", () =>
     Effect.gen(function* () {
-      const error = yield* compileRequest(
+      const error = yield* LLMClient.prepare(
         LLM.request({
           id: "req_tool_result_unsupported_media",
           model,
@@ -384,7 +384,7 @@ describe("Anthropic Messages route", () => {
 
   it.effect("prepares the composed native continuation request", () =>
     Effect.gen(function* () {
-      const prepared = yield* compileRequest(
+      const prepared = yield* LLMClient.prepare<AnthropicMessages.AnthropicMessagesBody>(
         continuationRequest({
           id: "req_native_continuation_anthropic",
           model,
@@ -428,7 +428,7 @@ describe("Anthropic Messages route", () => {
 
   it.effect("lowers preserved Anthropic reasoning signature metadata", () =>
     Effect.gen(function* () {
-      const prepared = yield* compileRequest(
+      const prepared = yield* LLMClient.prepare(
         LLM.request({
           model,
           messages: [
@@ -447,7 +447,7 @@ describe("Anthropic Messages route", () => {
 
   it.effect("round-trips redacted thinking as redacted_thinking blocks", () =>
     Effect.gen(function* () {
-      const prepared = yield* compileRequest(
+      const prepared = yield* LLMClient.prepare(
         LLM.request({
           model,
           messages: [
@@ -628,7 +628,9 @@ describe("Anthropic Messages route", () => {
         { type: "reasoning", text: "", providerMetadata: { anthropic: { signature: "sig_1" } } },
       ])
 
-      const prepared = yield* compileRequest(LLM.request({ model, messages: [response.message], cache: "none" }))
+      const prepared = yield* LLMClient.prepare<AnthropicMessages.AnthropicMessagesBody>(
+        LLM.request({ model, messages: [response.message], cache: "none" }),
+      )
       expect(prepared.body.messages).toEqual([
         { role: "assistant", content: [{ type: "thinking", thinking: "", signature: "sig_1" }] },
       ])
@@ -771,7 +773,7 @@ describe("Anthropic Messages route", () => {
           ),
         ),
       )
-      const prepared = yield* compileRequest(
+      const prepared = yield* LLMClient.prepare<AnthropicMessages.AnthropicMessagesBody>(
         LLM.request({
           model,
           messages: [
@@ -1131,7 +1133,7 @@ describe("Anthropic Messages route", () => {
 
   it.effect("round-trips provider-executed assistant content into server tool blocks", () =>
     Effect.gen(function* () {
-      const prepared = yield* compileRequest(
+      const prepared = yield* LLMClient.prepare(
         LLM.request({
           id: "req_round_trip",
           model,
@@ -1182,7 +1184,7 @@ describe("Anthropic Messages route", () => {
 
   it.effect("rejects round-trip for unknown server tool names", () =>
     Effect.gen(function* () {
-      const error = yield* compileRequest(
+      const error = yield* LLMClient.prepare(
         LLM.request({
           id: "req_unknown_server_tool",
           model,
@@ -1259,7 +1261,7 @@ describe("Anthropic Messages route", () => {
 
   it.effect("maps ttlSeconds >= 3600 to cache_control ttl: '1h'", () =>
     Effect.gen(function* () {
-      const prepared = yield* compileRequest(
+      const prepared = yield* LLMClient.prepare(
         LLM.request({
           model,
           system: { type: "text", text: "system", cache: new CacheHint({ type: "ephemeral", ttlSeconds: 3600 }) },
@@ -1275,7 +1277,7 @@ describe("Anthropic Messages route", () => {
 
   it.effect("emits cache_control on tool definitions and tool-result blocks", () =>
     Effect.gen(function* () {
-      const prepared = yield* compileRequest(
+      const prepared = yield* LLMClient.prepare(
         LLM.request({
           model,
           tools: [
@@ -1316,7 +1318,7 @@ describe("Anthropic Messages route", () => {
   it.effect("drops cache_control breakpoints past the 4-per-request cap", () =>
     Effect.gen(function* () {
       const hint = new CacheHint({ type: "ephemeral" })
-      const prepared = yield* compileRequest(
+      const prepared = yield* LLMClient.prepare(
         LLM.request({
           model,
           system: [
@@ -1342,7 +1344,7 @@ describe("Anthropic Messages route", () => {
   it.effect("spends breakpoint budget on tools before system before messages", () =>
     Effect.gen(function* () {
       const hint = new CacheHint({ type: "ephemeral" })
-      const prepared = yield* compileRequest(
+      const prepared = yield* LLMClient.prepare(
         LLM.request({
           model,
           tools: [

packages/ai/test/provider/bedrock-converse.test.ts

@@ -13,7 +13,6 @@ import {
   ToolDefinition,
 } from "../../src"
 import { LLMClient } from "../../src/route"
-import { compileRequest } from "../../src/route/client"
 import { AmazonBedrock } from "../../src/providers"
 import * as BedrockConverse from "../../src/protocols/bedrock-converse"
 import { it } from "../lib/effect"
@@ -102,7 +101,7 @@ const baseRequest = LLM.request({
 describe("Bedrock Converse route", () => {
   it.effect("prepares Converse target with system, inference config, and messages", () =>
     Effect.gen(function* () {
-      const prepared = yield* compileRequest(baseRequest)
+      const prepared = yield* LLMClient.prepare(baseRequest)
 
       expect(prepared.body).toEqual({
         modelId: "anthropic.claude-3-5-sonnet-20240620-v1:0",
@@ -115,7 +114,7 @@ describe("Bedrock Converse route", () => {
 
   it.effect("passes topK through additionalModelRequestFields as top_k", () =>
     Effect.gen(function* () {
-      const prepared = yield* compileRequest(
+      const prepared = yield* LLMClient.prepare<BedrockConverse.BedrockConverseBody>(
         LLMRequest.update(baseRequest, {
           generation: GenerationOptions.make({ maxTokens: 64, temperature: 0, topK: 40 }),
         }),
@@ -130,14 +129,14 @@ describe("Bedrock Converse route", () => {
 
   it.effect("omits additionalModelRequestFields when topK is unset", () =>
     Effect.gen(function* () {
-      const prepared = yield* compileRequest(baseRequest)
+      const prepared = yield* LLMClient.prepare<BedrockConverse.BedrockConverseBody>(baseRequest)
       expect(prepared.body.additionalModelRequestFields).toBeUndefined()
     }),
   )
 
   it.effect("lowers chronological system updates to wrapped user text in order", () =>
     Effect.gen(function* () {
-      const prepared = yield* compileRequest(
+      const prepared = yield* LLMClient.prepare<BedrockConverse.BedrockConverseBody>(
         LLM.request({
           model,
           messages: [Message.user("Before."), Message.system("Update."), Message.assistant("After.")],
@@ -154,7 +153,7 @@ describe("Bedrock Converse route", () => {
 
   it.effect("prepares tool config with toolSpec and toolChoice", () =>
     Effect.gen(function* () {
-      const prepared = yield* compileRequest(
+      const prepared = yield* LLMClient.prepare(
         LLMRequest.update(baseRequest, {
           tools: [
             ToolDefinition.make({
@@ -188,7 +187,7 @@ describe("Bedrock Converse route", () => {
 
   it.effect("keeps tools and omits the unsupported choice when tool choice is none", () =>
     Effect.gen(function* () {
-      const prepared = yield* compileRequest(
+      const prepared = yield* LLMClient.prepare(
         LLMRequest.update(baseRequest, {
           tools: [
             ToolDefinition.make({
@@ -218,7 +217,7 @@ describe("Bedrock Converse route", () => {
 
   it.effect("lowers assistant tool-call + tool-result message history", () =>
     Effect.gen(function* () {
-      const prepared = yield* compileRequest(
+      const prepared = yield* LLMClient.prepare(
         LLM.request({
           id: "req_history",
           model,
@@ -257,7 +256,7 @@ describe("Bedrock Converse route", () => {
 
   it.effect("lowers image content in tool-result messages", () =>
     Effect.gen(function* () {
-      const prepared = yield* compileRequest(
+      const prepared = yield* LLMClient.prepare(
         LLM.request({
           id: "req_tool_image",
           model,
@@ -492,7 +491,7 @@ describe("Bedrock Converse route", () => {
         providerMetadata: { bedrock: { signature: "sig_1" } },
       })
 
-      const prepared = yield* compileRequest(
+      const prepared = yield* LLMClient.prepare<BedrockConverse.BedrockConverseBody>(
         LLM.request({
           model,
           messages: [
@@ -547,7 +546,9 @@ describe("Bedrock Converse route", () => {
         },
       ])
 
-      const prepared = yield* compileRequest(LLM.request({ model, messages: [response.message], cache: "none" }))
+      const prepared = yield* LLMClient.prepare<BedrockConverse.BedrockConverseBody>(
+        LLM.request({ model, messages: [response.message], cache: "none" }),
+      )
       expect(prepared.body.messages).toEqual([
         {
           role: "assistant",
@@ -638,7 +639,7 @@ describe("Bedrock Converse route", () => {
         text: "",
         providerMetadata: { bedrock: { redactedData } },
       })
-      const prepared = yield* compileRequest(
+      const prepared = yield* LLMClient.prepare<BedrockConverse.BedrockConverseBody>(
         LLM.request({
           model,
           messages: [
@@ -753,7 +754,7 @@ describe("Bedrock Converse route", () => {
           secretAccessKey: "wJalrXUtnFEMI/K7MDENG/bPxRfiCYEXAMPLEKEY",
         },
       }).model("anthropic.claude-3-5-sonnet-20240620-v1:0")
-      const prepared = yield* compileRequest(LLMRequest.update(baseRequest, { model: signed }))
+      const prepared = yield* LLMClient.prepare(LLMRequest.update(baseRequest, { model: signed }))
 
       expect(prepared.route).toBe("bedrock-converse")
       expect(prepared.model).toBe(signed)
@@ -763,7 +764,7 @@ describe("Bedrock Converse route", () => {
   it.effect("emits cachePoint markers after system, user-text, and assistant-text with cache hints", () =>
     Effect.gen(function* () {
       const cache = new CacheHint({ type: "ephemeral" })
-      const prepared = yield* compileRequest(
+      const prepared = yield* LLMClient.prepare(
         LLM.request({
           id: "req_cache",
           model,
@@ -795,7 +796,7 @@ describe("Bedrock Converse route", () => {
 
   it.effect("does not emit cachePoint when no cache hint is set", () =>
     Effect.gen(function* () {
-      const prepared = yield* compileRequest(baseRequest)
+      const prepared = yield* LLMClient.prepare(baseRequest)
       expect(prepared.body).toMatchObject({
         system: [{ text: "You are concise." }],
         messages: [{ role: "user", content: [{ text: "Say hello." }] }],
@@ -805,7 +806,7 @@ describe("Bedrock Converse route", () => {
 
   it.effect("lowers image media into Bedrock image blocks", () =>
     Effect.gen(function* () {
-      const prepared = yield* compileRequest(
+      const prepared = yield* LLMClient.prepare(
         LLM.request({
           id: "req_image",
           model,
@@ -842,7 +843,7 @@ describe("Bedrock Converse route", () => {
 
   it.effect("base64-encodes Uint8Array image bytes", () =>
     Effect.gen(function* () {
-      const prepared = yield* compileRequest(
+      const prepared = yield* LLMClient.prepare(
         LLM.request({
           id: "req_image_bytes",
           model,
@@ -864,7 +865,7 @@ describe("Bedrock Converse route", () => {
 
   it.effect("lowers document media into Bedrock document blocks with format and name", () =>
     Effect.gen(function* () {
-      const prepared = yield* compileRequest(
+      const prepared = yield* LLMClient.prepare(
         LLM.request({
           id: "req_doc",
           model,
@@ -896,7 +897,7 @@ describe("Bedrock Converse route", () => {
 
   it.effect("requires names for document media", () =>
     Effect.gen(function* () {
-      const error = yield* compileRequest(
+      const error = yield* LLMClient.prepare(
         LLM.request({
           model,
           messages: [Message.user({ type: "media", mediaType: "application/pdf", data: "UERGREFUQQ==" })],
@@ -909,7 +910,7 @@ describe("Bedrock Converse route", () => {
 
   it.effect("passes named document-only messages through for provider validation", () =>
     Effect.gen(function* () {
-      const prepared = yield* compileRequest(
+      const prepared = yield* LLMClient.prepare<BedrockConverse.BedrockConverseBody>(
         LLM.request({
           model,
           cache: "none",
@@ -935,7 +936,7 @@ describe("Bedrock Converse route", () => {
 
   it.effect("lowers document media in tool results", () =>
     Effect.gen(function* () {
-      const prepared = yield* compileRequest(
+      const prepared = yield* LLMClient.prepare<BedrockConverse.BedrockConverseBody>(
         LLM.request({
           model,
           cache: "none",
@@ -987,7 +988,7 @@ describe("Bedrock Converse route", () => {
 
   it.effect("rejects unsupported image media types", () =>
     Effect.gen(function* () {
-      const error = yield* compileRequest(
+      const error = yield* LLMClient.prepare(
         LLM.request({
           id: "req_bad_image",
           model,
@@ -1001,7 +1002,7 @@ describe("Bedrock Converse route", () => {
 
   it.effect("rejects unsupported document media types", () =>
     Effect.gen(function* () {
-      const error = yield* compileRequest(
+      const error = yield* LLMClient.prepare(
         LLM.request({
           id: "req_bad_doc",
           model,
@@ -1016,7 +1017,7 @@ describe("Bedrock Converse route", () => {
   it.effect("maps ttlSeconds >= 3600 to cachePoint ttl: '1h'", () =>
     Effect.gen(function* () {
       const cache = new CacheHint({ type: "ephemeral", ttlSeconds: 3600 })
-      const prepared = yield* compileRequest(
+      const prepared = yield* LLMClient.prepare(
         LLM.request({
           model,
           system: [{ type: "text", text: "system", cache }],
@@ -1033,7 +1034,7 @@ describe("Bedrock Converse route", () => {
   it.effect("appends cachePoint after marked tool definitions and tool-result blocks", () =>
     Effect.gen(function* () {
       const cache = new CacheHint({ type: "ephemeral" })
-      const prepared = yield* compileRequest(
+      const prepared = yield* LLMClient.prepare(
         LLM.request({
           model,
           tools: [{ name: "lookup", description: "lookup", inputSchema: { type: "object", properties: {} }, cache }],
@@ -1065,7 +1066,7 @@ describe("Bedrock Converse route", () => {
   it.effect("drops cachePoint markers past the 4-per-request cap", () =>
     Effect.gen(function* () {
       const cache = new CacheHint({ type: "ephemeral" })
-      const prepared = yield* compileRequest(
+      const prepared = yield* LLMClient.prepare(
         LLM.request({
           model,
           system: [

packages/ai/test/provider/cloudflare.test.ts

@@ -3,7 +3,7 @@ import { ConfigProvider, Effect, Schema } from "effect"
 import { HttpClientRequest } from "effect/unstable/http"
 import { LLM, LLMEvent } from "../../src"
 import { CloudflareAIGateway, CloudflareWorkersAI } from "../../src/providers/cloudflare"
-import { compileRequest } from "../../src/route/client"
+import { LLMClient } from "../../src/route"
 import { it } from "../lib/effect"
 import { dynamicResponse } from "../lib/http"
 import { sseEvents } from "../lib/sse"
@@ -34,7 +34,7 @@ describe("Cloudflare", () => {
       })
       expect(model.route.endpoint.baseURL).toBe("https://gateway.ai.cloudflare.com/v1/test-account/test-gateway/compat")
 
-      const prepared = yield* compileRequest(LLM.request({ model, prompt: "Say hello." }))
+      const prepared = yield* LLMClient.prepare(LLM.request({ model, prompt: "Say hello." }))
 
       expect(prepared.route).toBe("cloudflare-ai-gateway")
       expect(prepared.body).toMatchObject({
@@ -129,7 +129,7 @@ describe("Cloudflare", () => {
         openai: { reasoningField: "reasoning", reasoningDetails: merged },
       })
 
-      const replay = yield* compileRequest(LLM.request({ model, messages: [response.message] }))
+      const replay = yield* LLMClient.prepare(LLM.request({ model, messages: [response.message] }))
       expect(replay.body.messages).toEqual([
         { role: "assistant", content: "Hello", reasoning: "Thinking", reasoning_details: merged },
       ])
@@ -180,7 +180,7 @@ describe("Cloudflare", () => {
 
   it.effect("allows a fully configured baseURL override", () =>
     Effect.gen(function* () {
-      const prepared = yield* compileRequest(
+      const prepared = yield* LLMClient.prepare(
         LLM.request({
           model: CloudflareAIGateway.configure({
             baseURL: "https://gateway.proxy.test/v1/custom/compat",
@@ -208,7 +208,7 @@ describe("Cloudflare", () => {
       })
       expect(model.route.endpoint.baseURL).toBe("https://api.cloudflare.com/client/v4/accounts/test-account/ai/v1")
 
-      const prepared = yield* compileRequest(LLM.request({ model, prompt: "Say hello." }))
+      const prepared = yield* LLMClient.prepare(LLM.request({ model, prompt: "Say hello." }))
 
       expect(prepared.route).toBe("cloudflare-workers-ai")
       expect(prepared.body).toMatchObject({

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

@@ -2,7 +2,6 @@ import { describe, expect } from "bun:test"
 import { Effect } from "effect"
 import { LLM, LLMError, LLMRequest, Message, ToolCallPart, ToolDefinition, Usage } from "../../src"
 import { Auth, LLMClient } from "../../src/route"
-import { compileRequest } from "../../src/route/client"
 import * as Gemini from "../../src/protocols/gemini"
 import { ProviderShared } from "../../src/protocols/shared"
 import { it } from "../lib/effect"
@@ -27,7 +26,7 @@ const request = LLM.request({
 describe("Gemini route", () => {
   it.effect("prepares Gemini target", () =>
     Effect.gen(function* () {
-      const prepared = yield* compileRequest(request)
+      const prepared = yield* LLMClient.prepare(request)
 
       expect(prepared.body).toEqual({
         contents: [{ role: "user", parts: [{ text: "Say hello." }] }],
@@ -39,56 +38,28 @@ describe("Gemini route", () => {
 
   it.effect("normalizes Gemini thinking options", () =>
     Effect.gen(function* () {
-      const prepared = yield* compileRequest(
+      const prepared = yield* LLMClient.prepare<Gemini.GeminiBody>(
         LLMRequest.update(request, {
-          providerOptions: {
+          providerOptions: { gemini: { thinkingConfig: { thinkingBudget: 0, includeThoughts: false } } },
-            gemini: {
-              cachedContent: "cachedContents/example",
-              safetySettings: [{ category: "HARM_CATEGORY_HATE_SPEECH", threshold: "BLOCK_ONLY_HIGH" }],
-              serviceTier: "priority",
-              thinkingConfig: { thinkingBudget: 0, includeThoughts: false, thinkingLevel: "high" },
-            },
-          },
         }),
       )
-      const filtered = yield* compileRequest(
+      const filtered = yield* LLMClient.prepare<Gemini.GeminiBody>(
         LLMRequest.update(request, {
           providerOptions: { gemini: { thinkingConfig: { thinkingBudget: "invalid", includeThoughts: false } } },
         }),
       )
-      const defaulted = yield* compileRequest(
-        LLMRequest.update(request, {
-          providerOptions: { gemini: { thinkingConfig: { thinkingLevel: "high" } } },
-        }),
-      )
-      const emptySafetySettings = yield* compileRequest(
-        LLMRequest.update(request, {
-          providerOptions: { gemini: { safetySettings: [] } },
-        }),
-      )
 
       expect(prepared.body.generationConfig?.thinkingConfig).toEqual({
         thinkingBudget: 0,
         includeThoughts: false,
-        thinkingLevel: "high",
       })
-      expect(prepared.body.cachedContent).toBe("cachedContents/example")
-      expect(prepared.body.safetySettings).toEqual([
-        { category: "HARM_CATEGORY_HATE_SPEECH", threshold: "BLOCK_ONLY_HIGH" },
-      ])
-      expect(prepared.body.serviceTier).toBe("priority")
       expect(filtered.body.generationConfig?.thinkingConfig).toEqual({ includeThoughts: false })
-      expect(defaulted.body.generationConfig?.thinkingConfig).toEqual({
-        includeThoughts: true,
-        thinkingLevel: "high",
-      })
-      expect(emptySafetySettings.body.safetySettings).toEqual([])
     }),
   )
 
   it.effect("lowers chronological system updates to wrapped user text in order", () =>
     Effect.gen(function* () {
-      const prepared = yield* compileRequest(
+      const prepared = yield* LLMClient.prepare<Gemini.GeminiBody>(
         LLM.request({
           model,
           messages: [Message.user("Before."), Message.system("Update."), Message.assistant("After.")],
@@ -104,7 +75,7 @@ describe("Gemini route", () => {
 
   it.effect("prepares multimodal user input and tool history", () =>
     Effect.gen(function* () {
-      const prepared = yield* compileRequest(
+      const prepared = yield* LLMClient.prepare(
         LLM.request({
           id: "req_tool_result",
           model,
@@ -172,7 +143,7 @@ describe("Gemini route", () => {
 
   it.effect("continues media tool results as inline model input without base64 text", () =>
     Effect.gen(function* () {
-      const prepared = yield* compileRequest(
+      const prepared = yield* LLMClient.prepare<Gemini.GeminiBody>(
         LLM.request({
           model,
           messages: [
@@ -217,7 +188,7 @@ describe("Gemini route", () => {
 
   it.effect("strips matching data URLs to raw base64 inlineData", () =>
     Effect.gen(function* () {
-      const prepared = yield* compileRequest(
+      const prepared = yield* LLMClient.prepare<Gemini.GeminiBody>(
         LLM.request({
           model,
           messages: [
@@ -258,7 +229,7 @@ describe("Gemini route", () => {
   ] as const)
     it.effect(`rejects ${name}`, () =>
       Effect.gen(function* () {
-        const error = yield* compileRequest(
+        const error = yield* LLMClient.prepare(
           LLM.request({ model, messages: [Message.user({ type: "media", ...media })] }),
         ).pipe(Effect.flip)
         expect(error.message).toMatch(/does not support|does not match|valid base64/)
@@ -267,7 +238,7 @@ describe("Gemini route", () => {
 
   it.effect("rejects oversized image input", () =>
     Effect.gen(function* () {
-      const error = yield* compileRequest(
+      const error = yield* LLMClient.prepare(
         LLM.request({
           model,
           messages: [
@@ -285,7 +256,7 @@ describe("Gemini route", () => {
 
   it.effect("keeps tools and sends function calling mode NONE", () =>
     Effect.gen(function* () {
-      const prepared = yield* compileRequest(
+      const prepared = yield* LLMClient.prepare(
         LLM.request({
           id: "req_tool_choice_none",
           model,
@@ -305,7 +276,7 @@ describe("Gemini route", () => {
 
   it.effect("sanitizes integer enums, dangling required, untyped arrays, and scalar object keys", () =>
     Effect.gen(function* () {
-      const prepared = yield* compileRequest(
+      const prepared = yield* LLMClient.prepare(
         LLM.request({
           id: "req_schema_patch",
           model,
@@ -486,7 +457,7 @@ describe("Gemini route", () => {
         response.events.findIndex((event) => event.type === "tool-call"),
       )
 
-      const prepared = yield* compileRequest(
+      const prepared = yield* LLMClient.prepare<Gemini.GeminiBody>(
         LLM.request({
           model,
           messages: [
@@ -720,7 +691,7 @@ describe("Gemini route", () => {
 
   it.effect("rejects unsupported assistant media content", () =>
     Effect.gen(function* () {
-      const error = yield* compileRequest(
+      const error = yield* LLMClient.prepare(
         LLM.request({
           id: "req_media",
           model,

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

@@ -4,7 +4,6 @@ import { HttpClientRequest } from "effect/unstable/http"
 import { LLM } from "../../src"
 import { GoogleVertex, GoogleVertexChat, GoogleVertexMessages, GoogleVertexResponses } from "../../src/providers"
 import { LLMClient } from "../../src/route"
-import { compileRequest } from "../../src/route/client"
 import { it } from "../lib/effect"
 import { dynamicResponse } from "../lib/http"
 import { deltaChunk, finishChunk } from "../lib/openai-chunks"
@@ -181,6 +180,23 @@ describe("Google Vertex providers", () => {
     }),
   )
 
+  it.effect("protects the Vertex Messages API version from body overlays", () =>
+    Effect.gen(function* () {
+      const error = yield* LLMClient.prepare(
+        LLM.request({
+          model: GoogleVertexMessages.configure({
+            accessToken: "vertex-token",
+            http: { body: { anthropic_version: "wrong" } },
+            project: "vertex-project",
+          }).model("claude-sonnet-4-6"),
+          prompt: "Say hello.",
+        }),
+      ).pipe(Effect.flip)
+
+      expect(error.message).toContain("http.body cannot overlay protocol-owned field(s): anthropic_version")
+    }),
+  )
+
   it.effect("routes tuned Gemini models through their deployed endpoint", () =>
     Effect.gen(function* () {
       const response = yield* LLMClient.generate(

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

@@ -5,7 +5,6 @@ import { OpenAIChat } from "../../src/protocols/openai-chat"
 import * as OpenAICompatible from "../../src/providers/openai-compatible"
 import * as OpenRouter from "../../src/providers/openrouter"
 import { LLMClient } from "../../src/route"
-import { compileRequest } from "../../src/route/client"
 import { recordedTests } from "../recorded-test"
 import { expectWeatherToolLoop, goldenWeatherToolLoopRequest, runWeatherToolLoop } from "../recorded-scenarios"
 
@@ -85,7 +84,9 @@ for (const item of cases) {
             ),
           ).toBe(true)
 
-          const replay = yield* compileRequest(LLM.request({ model: item.model, messages: [response.message] }))
+          const replay = yield* LLMClient.prepare<OpenAIChat.OpenAIChatBody>(
+            LLM.request({ model: item.model, messages: [response.message] }),
+          )
           expect(replay.body.messages).toMatchObject([
             { role: "assistant", content: response.text, reasoning: response.reasoning },
           ])

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

@@ -18,7 +18,6 @@ import * as OpenAI from "../../src/providers/openai"
 import * as OpenAIChat from "../../src/protocols/openai-chat"
 import { ProviderShared } from "../../src/protocols/shared"
 import { Auth, LLMClient } from "../../src/route"
-import { compileRequest } from "../../src/route/client"
 import { it } from "../lib/effect"
 import { dynamicResponse, fixedResponse, truncatedStream } from "../lib/http"
 import { deltaChunk, usageChunk } from "../lib/openai-chunks"
@@ -43,7 +42,11 @@ const request = LLM.request({
 describe("OpenAI Chat route", () => {
   it.effect("prepares OpenAI Chat payload", () =>
     Effect.gen(function* () {
-      const prepared = yield* compileRequest(request)
+      // Pass the OpenAIChat payload type so `prepared.body` is statically
+      // typed to the route's native shape — the assertions below read field
+      // names without `unknown` casts.
+      const prepared = yield* LLMClient.prepare<OpenAIChat.OpenAIChatBody>(request)
+      const _typed: { readonly model: string; readonly stream: true } = prepared.body
 
       expect(prepared.body).toEqual({
         model: "gpt-4o-mini",
@@ -61,7 +64,7 @@ describe("OpenAI Chat route", () => {
 
   it.effect("lowers chronological system updates to escaped user wrappers in order", () =>
     Effect.gen(function* () {
-      const prepared = yield* compileRequest(
+      const prepared = yield* LLMClient.prepare<OpenAIChat.OpenAIChatBody>(
         LLM.request({
           model,
           messages: [
@@ -84,7 +87,7 @@ describe("OpenAI Chat route", () => {
 
   it.effect("replays canonical reasoning as OpenAI-compatible reasoning_content", () =>
     Effect.gen(function* () {
-      const prepared = yield* compileRequest(
+      const prepared = yield* LLMClient.prepare<OpenAIChat.OpenAIChatBody>(
         LLM.request({
           model,
           messages: [
@@ -102,7 +105,7 @@ describe("OpenAI Chat route", () => {
 
   it.effect("writes reasoning to a configured custom field on every assistant message", () =>
     Effect.gen(function* () {
-      const prepared = yield* compileRequest(
+      const prepared = yield* LLMClient.prepare<OpenAIChat.OpenAIChatBody>(
         LLM.request({
           model: Model.update(model, { compatibility: { reasoningField: "vendor_reasoning" } }),
           messages: [
@@ -128,7 +131,7 @@ describe("OpenAI Chat route", () => {
 
   it.effect("rejects reasoning fields that conflict with assistant message fields", () =>
     Effect.gen(function* () {
-      const error = yield* compileRequest(
+      const error = yield* LLMClient.prepare(
         LLM.request({
           model: Model.update(model, { compatibility: { reasoningField: "content" } }),
           messages: [Message.assistant([{ type: "reasoning", text: "thinking" }])],
@@ -141,7 +144,7 @@ describe("OpenAI Chat route", () => {
 
   it.effect("maps OpenAI provider options to Chat options", () =>
     Effect.gen(function* () {
-      const prepared = yield* compileRequest(
+      const prepared = yield* LLMClient.prepare<OpenAIChat.OpenAIChatBody>(
         LLM.request({
           model: OpenAI.configure({ baseURL: "https://api.openai.test/v1/", apiKey: "test" }).chat("gpt-4o-mini"),
           prompt: "think",
@@ -156,7 +159,7 @@ describe("OpenAI Chat route", () => {
 
   it.effect("passes through custom OpenAI-compatible reasoning effort strings", () =>
     Effect.gen(function* () {
-      const prepared = yield* compileRequest(
+      const prepared = yield* LLMClient.prepare<OpenAIChat.OpenAIChatBody>(
         LLM.request({
           model,
           prompt: "think",
@@ -250,7 +253,7 @@ describe("OpenAI Chat route", () => {
 
   it.effect("prepares assistant tool-call and tool-result messages", () =>
     Effect.gen(function* () {
-      const prepared = yield* compileRequest(
+      const prepared = yield* LLMClient.prepare(
         LLM.request({
           id: "req_tool_result",
           model,
@@ -288,7 +291,7 @@ describe("OpenAI Chat route", () => {
   it.effect("preserves structured tool errors for the model", () =>
     Effect.gen(function* () {
       const error = { error: { type: "unknown", message: "Tool execution interrupted" } }
-      const prepared = yield* compileRequest(
+      const prepared = yield* LLMClient.prepare<OpenAIChat.OpenAIChatBody>(
         LLM.request({
           model,
           messages: [
@@ -308,7 +311,7 @@ describe("OpenAI Chat route", () => {
 
   it.effect("continues image tool results as vision input without base64 text", () =>
     Effect.gen(function* () {
-      const prepared = yield* compileRequest(
+      const prepared = yield* LLMClient.prepare<OpenAIChat.OpenAIChatBody>(
         LLM.request({
           model,
           messages: [
@@ -352,7 +355,7 @@ describe("OpenAI Chat route", () => {
 
   it.effect("orders parallel tool responses before one aggregated vision message", () =>
     Effect.gen(function* () {
-      const prepared = yield* compileRequest(
+      const prepared = yield* LLMClient.prepare<OpenAIChat.OpenAIChatBody>(
         LLM.request({
           model,
           messages: [
@@ -402,7 +405,7 @@ describe("OpenAI Chat route", () => {
 
   it.effect("aggregates consecutive tool images with a following system update", () =>
     Effect.gen(function* () {
-      const prepared = yield* compileRequest(
+      const prepared = yield* LLMClient.prepare<OpenAIChat.OpenAIChatBody>(
         LLM.request({
           model,
           messages: [
@@ -443,7 +446,7 @@ describe("OpenAI Chat route", () => {
 
   it.effect("appends system updates without replacing multipart user content", () =>
     Effect.gen(function* () {
-      const prepared = yield* compileRequest(
+      const prepared = yield* LLMClient.prepare<OpenAIChat.OpenAIChatBody>(
         LLM.request({
           model,
           messages: [
@@ -471,7 +474,7 @@ describe("OpenAI Chat route", () => {
   ] as const)
     it.effect(`rejects ${name}`, () =>
       Effect.gen(function* () {
-        const error = yield* compileRequest(
+        const error = yield* LLMClient.prepare(
           LLM.request({ model, messages: [Message.user({ type: "media", ...media })] }),
         ).pipe(Effect.flip)
         expect(error.message).toMatch(/does not support|does not match|valid base64/)
@@ -480,7 +483,7 @@ describe("OpenAI Chat route", () => {
 
   it.effect("rejects oversized image input", () =>
     Effect.gen(function* () {
-      const error = yield* compileRequest(
+      const error = yield* LLMClient.prepare(
         LLM.request({
           model,
           messages: [
@@ -498,7 +501,7 @@ describe("OpenAI Chat route", () => {
 
   it.effect("prepares raw and data URL image media as vision input", () =>
     Effect.gen(function* () {
-      const prepared = yield* compileRequest(
+      const prepared = yield* LLMClient.prepare<OpenAIChat.OpenAIChatBody>(
         LLM.request({
           id: "req_media",
           model,
@@ -525,7 +528,7 @@ describe("OpenAI Chat route", () => {
 
   it.effect("lowers reasoning-only assistant history", () =>
     Effect.gen(function* () {
-      const prepared = yield* compileRequest(
+      const prepared = yield* LLMClient.prepare<OpenAIChat.OpenAIChatBody>(
         LLM.request({
           id: "req_reasoning",
           model,
@@ -616,7 +619,9 @@ describe("OpenAI Chat route", () => {
           openai: { reasoningField: field },
         })
 
-        const replay = yield* compileRequest(LLM.request({ model, messages: [response.message] }))
+        const replay = yield* LLMClient.prepare<OpenAIChat.OpenAIChatBody>(
+          LLM.request({ model, messages: [response.message] }),
+        )
         expect(replay.body.messages).toEqual([{ role: "assistant", content: "Hello", [field]: "thinking" }])
       }
     }),
@@ -642,7 +647,9 @@ describe("OpenAI Chat route", () => {
         openai: { reasoningField: "vendor_reasoning" },
       })
 
-      const replay = yield* compileRequest(LLM.request({ model: custom, messages: [response.message] }))
+      const replay = yield* LLMClient.prepare<OpenAIChat.OpenAIChatBody>(
+        LLM.request({ model: custom, messages: [response.message] }),
+      )
       expect(replay.body.messages).toEqual([{ role: "assistant", content: "Hello", vendor_reasoning: "thinking" }])
     }),
   )
@@ -685,7 +692,9 @@ describe("OpenAI Chat route", () => {
         openai: { reasoningField: "reasoning", reasoningDetails: details },
       })
 
-      const replay = yield* compileRequest(LLM.request({ model, messages: [response.message] }))
+      const replay = yield* LLMClient.prepare<OpenAIChat.OpenAIChatBody>(
+        LLM.request({ model, messages: [response.message] }),
+      )
       expect(replay.body.messages).toEqual([
         {
           role: "assistant",
@@ -728,7 +737,9 @@ describe("OpenAI Chat route", () => {
         openai: { reasoningDetails: details },
       })
 
-      const replay = yield* compileRequest(LLM.request({ model, messages: [response.message] }))
+      const replay = yield* LLMClient.prepare<OpenAIChat.OpenAIChatBody>(
+        LLM.request({ model, messages: [response.message] }),
+      )
       expect(replay.body.messages).toEqual([{ role: "assistant", content: "Hello", reasoning_details: details }])
     }),
   )
@@ -753,7 +764,9 @@ describe("OpenAI Chat route", () => {
         openai: { reasoningField: "reasoning", reasoningDetails: details },
       })
 
-      const replay = yield* compileRequest(LLM.request({ model, messages: [response.message] }))
+      const replay = yield* LLMClient.prepare<OpenAIChat.OpenAIChatBody>(
+        LLM.request({ model, messages: [response.message] }),
+      )
       expect(replay.body.messages).toEqual([
         { role: "assistant", content: "Hello", reasoning: "thinking", reasoning_details: details },
       ])
@@ -826,7 +839,9 @@ describe("OpenAI Chat route", () => {
         openai: { reasoningDetails: [] },
       })
 
-      const replay = yield* compileRequest(LLM.request({ model, messages: [response.message] }))
+      const replay = yield* LLMClient.prepare<OpenAIChat.OpenAIChatBody>(
+        LLM.request({ model, messages: [response.message] }),
+      )
       expect(replay.body.messages).toEqual([{ role: "assistant", content: "Hello", reasoning_details: [] }])
     }),
   )
@@ -874,7 +889,9 @@ describe("OpenAI Chat route", () => {
         response.events.findIndex(LLMEvent.is.textStart),
       )
 
-      const replay = yield* compileRequest(LLM.request({ model, messages: [response.message] }))
+      const replay = yield* LLMClient.prepare<OpenAIChat.OpenAIChatBody>(
+        LLM.request({ model, messages: [response.message] }),
+      )
       expect(replay.body.messages).toEqual([
         { role: "assistant", content: "Hello", reasoning: "thinking", reasoning_details: merged },
       ])
@@ -901,7 +918,9 @@ describe("OpenAI Chat route", () => {
       expect(response.events.filter(LLMEvent.is.reasoningStart)).toHaveLength(1)
       expect(response.events.filter(LLMEvent.is.reasoningEnd)).toHaveLength(1)
 
-      const replay = yield* compileRequest(LLM.request({ model, messages: [response.message] }))
+      const replay = yield* LLMClient.prepare<OpenAIChat.OpenAIChatBody>(
+        LLM.request({ model, messages: [response.message] }),
+      )
       expect(replay.body.messages).toEqual([{ role: "assistant", content: null, reasoning_details: details }])
     }),
   )
@@ -931,7 +950,7 @@ describe("OpenAI Chat route", () => {
     Effect.gen(function* () {
       const first = { type: "reasoning.text", text: "first", signature: "signed-0", index: 0 }
       const second = { type: "reasoning.text", text: "second", signature: "signed-1", index: 1 }
-      const replay = yield* compileRequest(
+      const replay = yield* LLMClient.prepare<OpenAIChat.OpenAIChatBody>(
         LLM.request({
           model,
           messages: [
@@ -960,7 +979,7 @@ describe("OpenAI Chat route", () => {
   it.effect("retains scalar replay for mixed structured reasoning parts", () =>
     Effect.gen(function* () {
       const detail = { type: "reasoning.encrypted", data: "opaque", index: 0 }
-      const replay = yield* compileRequest(
+      const replay = yield* LLMClient.prepare<OpenAIChat.OpenAIChatBody>(
         LLM.request({
           model,
           messages: [
@@ -985,7 +1004,7 @@ describe("OpenAI Chat route", () => {
   it.effect("replays native scalar reasoning alongside native details", () =>
     Effect.gen(function* () {
       const details = [{ type: "reasoning.encrypted", data: "opaque", index: 0 }]
-      const replay = yield* compileRequest(
+      const replay = yield* LLMClient.prepare<OpenAIChat.OpenAIChatBody>(
         LLM.request({
           model,
           messages: [

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

@@ -3,7 +3,6 @@ import { Effect, Schema } from "effect"
 import { HttpClientRequest } from "effect/unstable/http"
 import { LLM, LLMRequest, Message, ToolCallPart, ToolChoice, ToolDefinition } from "../../src"
 import { Auth, LLMClient } from "../../src/route"
-import { compileRequest } from "../../src/route/client"
 import * as OpenAICompatible from "../../src/providers/openai-compatible"
 import * as OpenAICompatibleChat from "../../src/protocols/openai-compatible-chat"
 import { it } from "../lib/effect"
@@ -53,7 +52,7 @@ const providerFamilies = [
 describe("OpenAI-compatible Chat route", () => {
   it.effect("prepares generic Chat target", () =>
     Effect.gen(function* () {
-      const prepared = yield* compileRequest(
+      const prepared = yield* LLMClient.prepare(
         LLMRequest.update(request, {
           tools: [ToolDefinition.make({ name: "lookup", description: "Lookup data", inputSchema: { type: "object" } })],
           toolChoice: ToolChoice.make({ type: "required" }),
@@ -128,7 +127,7 @@ describe("OpenAI-compatible Chat route", () => {
 
   it.effect("matches AI SDK compatible basic request body fixture", () =>
     Effect.gen(function* () {
-      const prepared = yield* compileRequest(request)
+      const prepared = yield* LLMClient.prepare(request)
 
       expect(prepared.body).toEqual({
         model: "deepseek-chat",
@@ -144,23 +143,9 @@ describe("OpenAI-compatible Chat route", () => {
     }),
   )
 
-  it.effect("configures the max tokens request field", () =>
-    Effect.gen(function* () {
-      const compatible = OpenAICompatibleChat.route
-        .with({ provider: "custom", endpoint: { baseURL: "https://api.custom.test/v1" } })
-        .model({ id: "custom-model", compatibility: { maxTokensField: "max_completion_tokens" } })
-      const prepared = yield* compileRequest(
-        LLM.request({ model: compatible, prompt: "Say hello.", generation: { maxTokens: 20 } }),
-      )
-
-      expect(prepared.body).toMatchObject({ max_completion_tokens: 20 })
-      expect(prepared.body).not.toHaveProperty("max_tokens")
-    }),
-  )
-
   it.effect("matches AI SDK compatible tool request body fixture", () =>
     Effect.gen(function* () {
-      const prepared = yield* compileRequest(
+      const prepared = yield* LLMClient.prepare(
         LLM.request({
           id: "req_tool_parity",
           model,

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

@@ -7,7 +7,6 @@ import { OpenResponses } from "../../src/protocols/open-responses"
 import { OpenAICompatibleResponses } from "../../src/protocols/openai-compatible-responses"
 import { OpenAIResponses } from "../../src/protocols/openai-responses"
 import { LLMClient } from "../../src/route"
-import { compileRequest } from "../../src/route/client"
 import { it } from "../lib/effect"
 import { fixedResponse } from "../lib/http"
 import { sseEvents } from "../lib/sse"
@@ -24,7 +23,7 @@ describe("Open Responses-compatible route", () => {
         baseURL: "https://responses.example.test/v1",
         provider: "example",
       }).model("example-model")
-      const prepared = yield* compileRequest(
+      const prepared = yield* LLMClient.prepare(
         LLM.request({
           model,
           system: "You are concise.",
@@ -62,7 +61,7 @@ describe("Open Responses-compatible route", () => {
         apiKey: "test-key",
         baseURL: "https://responses.example.test/v1",
       }).model("example-model")
-      const error = yield* compileRequest(
+      const error = yield* LLMClient.prepare(
         LLM.request({ model, prompt: "Draw.", tools: [OpenAI.imageGeneration()] }),
       ).pipe(Effect.flip)
 
@@ -77,7 +76,7 @@ describe("Open Responses-compatible route", () => {
         apiKey: "test-key",
         baseURL: "https://responses.example.test/v1",
       }).model("example-model")
-      const prepared = yield* compileRequest(
+      const prepared = yield* LLMClient.prepare(
         LLM.request({
           model,
           messages: [
@@ -103,7 +102,7 @@ describe("Open Responses-compatible route", () => {
         baseURL: "https://responses.example.test/v1",
         providerOptions: { openresponses: { reasoningEffort: "low", store: true } },
       }).model("example-model")
-      const prepared = yield* compileRequest(LLM.request({ model, prompt: "Think." }))
+      const prepared = yield* LLMClient.prepare(LLM.request({ model, prompt: "Think." }))
 
       expect(prepared.body).toMatchObject({
         reasoning: { effort: "low" },

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

@@ -14,7 +14,6 @@ import {
   Usage,
 } from "../../src"
 import { Auth, LLMClient, RequestExecutor, WebSocketExecutor } from "../../src/route"
-import { compileRequest } from "../../src/route/client"
 import * as Azure from "../../src/providers/azure"
 import * as OpenAI from "../../src/providers/openai"
 import * as XAI from "../../src/providers/xai"
@@ -57,7 +56,7 @@ const expectToolOutput = (body: OpenAIResponses.OpenAIResponsesBody): OpenAITool
 describe("OpenAI Responses route", () => {
   it.effect("prepares OpenAI Responses target", () =>
     Effect.gen(function* () {
-      const prepared = yield* compileRequest(request)
+      const prepared = yield* LLMClient.prepare(request)
 
       expect(prepared.body).toEqual({
         model: "gpt-4.1-mini",
@@ -75,7 +74,7 @@ describe("OpenAI Responses route", () => {
 
   it.effect("lowers the hosted OpenAI image generation tool", () =>
     Effect.gen(function* () {
-      const prepared = yield* compileRequest(
+      const prepared = yield* LLMClient.prepare<OpenAIResponses.OpenAIResponsesBody>(
         LLM.request({
           model,
           prompt: "Show me a rooftop garden.",
@@ -93,7 +92,7 @@ describe("OpenAI Responses route", () => {
 
   it.effect("rejects invalid hosted image generation options locally", () =>
     Effect.gen(function* () {
-      const error = yield* compileRequest(
+      const error = yield* LLMClient.prepare(
         LLM.request({
           model,
           prompt: "Show me a rooftop garden.",
@@ -110,7 +109,7 @@ describe("OpenAI Responses route", () => {
     Effect.gen(function* () {
       const input = LLMRequest.update(request, { providerOptions: { openai: { serviceTier: "priority" } } })
       expect(input.providerOptions).toEqual({ openai: { serviceTier: "priority" } })
-      const prepared = yield* compileRequest(input)
+      const prepared = yield* LLMClient.prepare(input)
 
       expect(prepared.body).toMatchObject({ service_tier: "priority" })
       expect(prepared.body).not.toHaveProperty("serviceTier")
@@ -119,7 +118,7 @@ describe("OpenAI Responses route", () => {
 
   it.effect("passes through custom OpenAI reasoning effort strings", () =>
     Effect.gen(function* () {
-      const prepared = yield* compileRequest(
+      const prepared = yield* LLMClient.prepare<OpenAIResponses.OpenAIResponsesBody>(
         LLMRequest.update(request, { providerOptions: { openai: { reasoningEffort: "experimental" } } }),
       )
 
@@ -129,7 +128,7 @@ describe("OpenAI Responses route", () => {
 
   it.effect("omits unsupported semantic service tiers", () =>
     Effect.gen(function* () {
-      const prepared = yield* compileRequest(
+      const prepared = yield* LLMClient.prepare(
         LLMRequest.update(request, { providerOptions: { openai: { serviceTier: "unsupported" } } }),
       )
 
@@ -139,7 +138,7 @@ describe("OpenAI Responses route", () => {
 
   it.effect("flattens top-level object unions in function schemas", () =>
     Effect.gen(function* () {
-      const prepared = yield* compileRequest(
+      const prepared = yield* LLMClient.prepare<OpenAIResponses.OpenAIResponsesBody>(
         LLMRequest.update(request, {
           tools: [
             ToolDefinition.make({
@@ -192,7 +191,7 @@ describe("OpenAI Responses route", () => {
 
   it.effect("lowers chronological system updates to escaped user wrappers in order", () =>
     Effect.gen(function* () {
-      const prepared = yield* compileRequest(
+      const prepared = yield* LLMClient.prepare<OpenAIResponses.OpenAIResponsesBody>(
         LLM.request({
           model,
           messages: [
@@ -218,7 +217,7 @@ describe("OpenAI Responses route", () => {
 
   it.effect("prepares OpenAI Responses WebSocket target", () =>
     Effect.gen(function* () {
-      const prepared = yield* compileRequest(
+      const prepared = yield* LLMClient.prepare(
         LLMRequest.update(request, {
           model: OpenAIResponses.webSocketRoute
             .with({ endpoint: { baseURL: "https://api.openai.test/v1/" }, auth: Auth.bearer("test") })
@@ -396,7 +395,7 @@ describe("OpenAI Responses route", () => {
 
   it.effect("prepares function call and function output input items", () =>
     Effect.gen(function* () {
-      const prepared = yield* compileRequest(
+      const prepared = yield* LLMClient.prepare(
         LLM.request({
           id: "req_tool_result",
           model,
@@ -433,7 +432,7 @@ describe("OpenAI Responses route", () => {
         content: [],
         structured: {},
       }
-      const prepared = yield* compileRequest(
+      const prepared = yield* LLMClient.prepare<OpenAIResponses.OpenAIResponsesBody>(
         LLM.request({
           model,
           messages: [
@@ -454,7 +453,7 @@ describe("OpenAI Responses route", () => {
 
   it.effect("keeps primitive tool errors as plain text", () =>
     Effect.gen(function* () {
-      const prepared = yield* compileRequest(
+      const prepared = yield* LLMClient.prepare<OpenAIResponses.OpenAIResponsesBody>(
         LLM.request({
           model,
           messages: [
@@ -470,7 +469,7 @@ describe("OpenAI Responses route", () => {
 
   it.effect("keeps non-JSON tool errors as plain text", () =>
     Effect.gen(function* () {
-      const prepared = yield* compileRequest(
+      const prepared = yield* LLMClient.prepare<OpenAIResponses.OpenAIResponsesBody>(
         LLM.request({
           model,
           messages: [
@@ -488,7 +487,7 @@ describe("OpenAI Responses route", () => {
   // image data is not JSON-stringified into `function_call_output.output`.
   it.effect("lowers image tool-result content as structured input_image items", () =>
     Effect.gen(function* () {
-      const prepared = yield* compileRequest(
+      const prepared = yield* LLMClient.prepare<OpenAIResponses.OpenAIResponsesBody>(
         LLM.request({
           id: "req_tool_result_image",
           model,
@@ -517,7 +516,7 @@ describe("OpenAI Responses route", () => {
 
   it.effect("lowers single-image tool-result content as structured input_image array", () =>
     Effect.gen(function* () {
-      const prepared = yield* compileRequest(
+      const prepared = yield* LLMClient.prepare<OpenAIResponses.OpenAIResponsesBody>(
         LLM.request({
           id: "req_tool_result_image_only",
           model,
@@ -541,7 +540,7 @@ describe("OpenAI Responses route", () => {
 
   it.effect("lowers PDF tool-result content as structured input_file array", () =>
     Effect.gen(function* () {
-      const prepared = yield* compileRequest(
+      const prepared = yield* LLMClient.prepare<OpenAIResponses.OpenAIResponsesBody>(
         LLM.request({
           id: "req_tool_result_pdf",
           model,
@@ -576,7 +575,7 @@ describe("OpenAI Responses route", () => {
 
   it.effect("uses xAI inline file encoding for PDF tool results", () =>
     Effect.gen(function* () {
-      const prepared = yield* compileRequest(
+      const prepared = yield* LLMClient.prepare<OpenAIResponses.OpenAIResponsesBody>(
         LLM.request({
           model: xaiModel,
           messages: [
@@ -611,7 +610,7 @@ describe("OpenAI Responses route", () => {
 
   it.effect("rejects unsupported media in tool-result content with a clear error", () =>
     Effect.gen(function* () {
-      const error = yield* compileRequest(
+      const error = yield* LLMClient.prepare(
         LLM.request({
           id: "req_tool_result_unsupported_media",
           model,
@@ -634,7 +633,7 @@ describe("OpenAI Responses route", () => {
 
   it.effect("prepares the composed native continuation request", () =>
     Effect.gen(function* () {
-      const prepared = yield* compileRequest(
+      const prepared = yield* LLMClient.prepare<OpenAIResponses.OpenAIResponsesBody>(
         continuationRequest({
           id: "req_native_continuation_openai",
           model,
@@ -676,7 +675,7 @@ describe("OpenAI Responses route", () => {
 
   it.effect("maps OpenAI provider options to Responses options", () =>
     Effect.gen(function* () {
-      const prepared = yield* compileRequest(
+      const prepared = yield* LLMClient.prepare<OpenAIResponses.OpenAIResponsesBody>(
         LLM.request({
           model: OpenAI.configure({ baseURL: "https://api.openai.test/v1/", apiKey: "test" }).model("gpt-5.2"),
           prompt: "think",
@@ -701,7 +700,7 @@ describe("OpenAI Responses route", () => {
 
   it.effect("accepts the full ResponseIncludable union", () =>
     Effect.gen(function* () {
-      const prepared = yield* compileRequest(
+      const prepared = yield* LLMClient.prepare<OpenAIResponses.OpenAIResponsesBody>(
         LLM.request({
           model,
           prompt: "hi",
@@ -723,7 +722,7 @@ describe("OpenAI Responses route", () => {
 
   it.effect("filters unknown includable values out of the include array", () =>
     Effect.gen(function* () {
-      const prepared = yield* compileRequest(
+      const prepared = yield* LLMClient.prepare<OpenAIResponses.OpenAIResponsesBody>(
         LLM.request({
           model,
           prompt: "hi",
@@ -740,7 +739,7 @@ describe("OpenAI Responses route", () => {
 
   it.effect("treats an explicit empty include as no include at all", () =>
     Effect.gen(function* () {
-      const prepared = yield* compileRequest(
+      const prepared = yield* LLMClient.prepare<OpenAIResponses.OpenAIResponsesBody>(
         LLM.request({ model, prompt: "hi", providerOptions: { openai: { include: [] } } }),
       )
 
@@ -750,7 +749,7 @@ describe("OpenAI Responses route", () => {
 
   it.effect("treats an all-invalid include as no include at all", () =>
     Effect.gen(function* () {
-      const prepared = yield* compileRequest(
+      const prepared = yield* LLMClient.prepare<OpenAIResponses.OpenAIResponsesBody>(
         LLM.request({ model, prompt: "hi", providerOptions: { openai: { include: ["bogus.thing"] } } }),
       )
 
@@ -760,7 +759,7 @@ describe("OpenAI Responses route", () => {
 
   it.effect("omits include when no include is set", () =>
     Effect.gen(function* () {
-      const prepared = yield* compileRequest(
+      const prepared = yield* LLMClient.prepare<OpenAIResponses.OpenAIResponsesBody>(
         LLM.request({ model, prompt: "hi", providerOptions: { openai: { store: false } } }),
       )
 
@@ -774,7 +773,7 @@ describe("OpenAI Responses route", () => {
       // reasoningSummary: "auto" by default. Without `include`, a follow-up
       // turn cannot replay reasoning state, so the facade also opts into
       // `reasoning.encrypted_content` automatically.
-      const prepared = yield* compileRequest(
+      const prepared = yield* LLMClient.prepare<OpenAIResponses.OpenAIResponsesBody>(
         LLM.request({
           model: OpenAI.configure({ baseURL: "https://api.openai.test/v1/", apiKey: "test" }).responses("gpt-5.2"),
           prompt: "hi",
@@ -789,7 +788,7 @@ describe("OpenAI Responses route", () => {
 
   it.effect("lets callers opt out of the GPT-5 default include", () =>
     Effect.gen(function* () {
-      const prepared = yield* compileRequest(
+      const prepared = yield* LLMClient.prepare<OpenAIResponses.OpenAIResponsesBody>(
         LLM.request({
           model: OpenAI.configure({ baseURL: "https://api.openai.test/v1/", apiKey: "test" }).responses("gpt-5.2"),
           prompt: "hi",
@@ -803,7 +802,7 @@ describe("OpenAI Responses route", () => {
 
   it.effect("request OpenAI provider options override route defaults", () =>
     Effect.gen(function* () {
-      const prepared = yield* compileRequest(
+      const prepared = yield* LLMClient.prepare<OpenAIResponses.OpenAIResponsesBody>(
         LLM.request({
           model: OpenAI.configure({
             baseURL: "https://api.openai.test/v1/",
@@ -935,7 +934,9 @@ describe("OpenAI Responses route", () => {
         },
       ])
 
-      const prepared = yield* compileRequest(LLM.request({ model, messages: [response.message] }))
+      const prepared = yield* LLMClient.prepare<OpenAIResponses.OpenAIResponsesBody>(
+        LLM.request({ model, messages: [response.message] }),
+      )
       expect(prepared.body.input).toEqual([
         {
           role: "assistant",
@@ -1269,7 +1270,7 @@ describe("OpenAI Responses route", () => {
 
   it.effect("preserves assistant content order around reasoning items", () =>
     Effect.gen(function* () {
-      const prepared = yield* compileRequest(
+      const prepared = yield* LLMClient.prepare<OpenAIResponses.OpenAIResponsesBody>(
         LLM.request({
           id: "req_reasoning_order",
           model,
@@ -1307,7 +1308,7 @@ describe("OpenAI Responses route", () => {
 
   it.effect("references stored reasoning items by id", () =>
     Effect.gen(function* () {
-      const prepared = yield* compileRequest(
+      const prepared = yield* LLMClient.prepare<OpenAIResponses.OpenAIResponsesBody>(
         LLM.request({
           model,
           messages: [
@@ -1329,7 +1330,7 @@ describe("OpenAI Responses route", () => {
 
   it.effect("references stored provider-executed hosted tool results by id", () =>
     Effect.gen(function* () {
-      const prepared = yield* compileRequest(
+      const prepared = yield* LLMClient.prepare<OpenAIResponses.OpenAIResponsesBody>(
         LLM.request({
           model,
           messages: [
@@ -1366,7 +1367,7 @@ describe("OpenAI Responses route", () => {
   it.effect("continues stateless hosted image generation with the generated image", () =>
     Effect.gen(function* () {
       const imageTool = OpenAI.imageGeneration({ action: "edit" })
-      const prepared = yield* compileRequest(
+      const prepared = yield* LLMClient.prepare<OpenAIResponses.OpenAIResponsesBody>(
         LLM.request({
           model,
           messages: [
@@ -1407,7 +1408,7 @@ describe("OpenAI Responses route", () => {
 
   it.effect("joins streamed summary blocks into one continuation reasoning item", () =>
     Effect.gen(function* () {
-      const prepared = yield* compileRequest(
+      const prepared = yield* LLMClient.prepare<OpenAIResponses.OpenAIResponsesBody>(
         LLM.request({
           id: "req_multi_summary_continuation",
           model,
@@ -1444,7 +1445,7 @@ describe("OpenAI Responses route", () => {
 
   it.effect("skips non-persisted reasoning ids without encrypted state", () =>
     Effect.gen(function* () {
-      const prepared = yield* compileRequest(
+      const prepared = yield* LLMClient.prepare(
         LLM.request({
           id: "req_reasoning_without_encrypted_state",
           model,
@@ -1761,7 +1762,7 @@ describe("OpenAI Responses route", () => {
 
   it.effect("lowers user image and PDF content", () =>
     Effect.gen(function* () {
-      const prepared = yield* compileRequest(
+      const prepared = yield* LLMClient.prepare<OpenAIResponses.OpenAIResponsesBody>(
         LLM.request({
           id: "req_media",
           model,
@@ -1792,7 +1793,7 @@ describe("OpenAI Responses route", () => {
 
   it.effect("uses xAI inline file encoding for user PDFs", () =>
     Effect.gen(function* () {
-      const prepared = yield* compileRequest(
+      const prepared = yield* LLMClient.prepare<OpenAIResponses.OpenAIResponsesBody>(
         LLM.request({
           model: xaiModel,
           messages: [
@@ -1824,7 +1825,7 @@ describe("OpenAI Responses route", () => {
 
   it.effect("rejects unsupported user media content", () =>
     Effect.gen(function* () {
-      const error = yield* compileRequest(
+      const error = yield* LLMClient.prepare(
         LLM.request({
           id: "req_media",
           model,

packages/ai/test/provider/openrouter.test.ts

@@ -1,8 +1,7 @@
 import { describe, expect } from "bun:test"
 import { Effect } from "effect"
-import { CacheHint, LLM, Message } from "../../src"
+import { LLM, Message } from "../../src"
 import { LLMClient } from "../../src/route"
-import { compileRequest } from "../../src/route/client"
 import * as OpenRouter from "../../src/providers/openrouter"
 import { it } from "../lib/effect"
 import { fixedResponse } from "../lib/http"
@@ -20,141 +19,20 @@ describe("OpenRouter", () => {
       })
       expect(model.route.endpoint.baseURL).toBe("https://openrouter.ai/api/v1")
 
-      const prepared = yield* compileRequest(LLM.request({ model, prompt: "Say hello." }))
+      const prepared = yield* LLMClient.prepare(LLM.request({ model, prompt: "Say hello." }))
 
       expect(prepared.route).toBe("openrouter")
       expect(prepared.body).toMatchObject({
         model: "openai/gpt-4o-mini",
         messages: [{ role: "user", content: "Say hello." }],
         stream: true,
-        usage: { include: true },
       })
     }),
   )
 
-  it.effect("lowers the native cache policy to OpenRouter cache controls", () =>
-    Effect.gen(function* () {
-      const prepared = yield* compileRequest(
-        LLM.request({
-          model: OpenRouter.configure({ apiKey: "test-key" }).model("anthropic/claude-sonnet-4.6"),
-          system: [
-            { type: "text", text: "Base agent", cache: new CacheHint({ type: "ephemeral", ttlSeconds: 3_600 }) },
-            { type: "text", text: "Project instructions" },
-          ],
-          tools: [{ name: "lookup", description: "Lookup", inputSchema: { type: "object", properties: {} } }],
-          prompt: "Hello",
-          cache: { tools: true, system: true, messages: { tail: 1 } },
-        }),
-      )
-
-      expect(prepared.body).toMatchObject({
-        tools: [{ cache_control: { type: "ephemeral" } }],
-        messages: [
-          {
-            role: "system",
-            content: [
-              { text: "Base agent", cache_control: { type: "ephemeral", ttl: "1h" } },
-              { text: "Project instructions", cache_control: { type: "ephemeral" } },
-            ],
-          },
-          {
-            role: "user",
-            content: [{ text: "Hello", cache_control: { type: "ephemeral" } }],
-          },
-        ],
-      })
-    }),
-  )
-
-  it.effect("lowers manual assistant and tool-result cache hints", () =>
-    Effect.gen(function* () {
-      const prepared = yield* compileRequest(
-        LLM.request({
-          model: OpenRouter.configure({ apiKey: "test-key" }).model("anthropic/claude-sonnet-4.6"),
-          cache: "none",
-          messages: [
-            Message.user("Call the tool"),
-            Message.assistant([
-              { type: "text", text: "Calling", cache: new CacheHint({ type: "ephemeral" }) },
-              { type: "tool-call", id: "call_1", name: "lookup", input: {} },
-            ]),
-            Message.tool({
-              id: "call_1",
-              name: "lookup",
-              result: "Done",
-              cache: new CacheHint({ type: "ephemeral", ttlSeconds: 3_600 }),
-            }),
-          ],
-        }),
-      )
-
-      expect(prepared.body.messages).toMatchObject([
-        { role: "user", content: "Call the tool" },
-        { role: "assistant", content: "Calling", cache_control: { type: "ephemeral" } },
-        { role: "tool", content: '"Done"', cache_control: { type: "ephemeral", ttl: "1h" } },
-      ])
-    }),
-  )
-
-  it.effect("caps manual cache controls at four breakpoints", () =>
-    Effect.gen(function* () {
-      const cache = new CacheHint({ type: "ephemeral" })
-      const prepared = yield* compileRequest(
-        LLM.request({
-          model: OpenRouter.configure({ apiKey: "test-key" }).model("anthropic/claude-sonnet-4.6"),
-          cache: "none",
-          system: [1, 2, 3, 4, 5].map((index) => ({ type: "text" as const, text: `System ${index}`, cache })),
-          prompt: "Hello",
-        }),
-      )
-
-      const system = prepared.body.messages[0]
-      expect(system?.role).toBe("system")
-      expect(
-        system && Array.isArray(system.content)
-          ? system.content.filter((part) => "cache_control" in part && part.cache_control !== undefined)
-          : [],
-      ).toHaveLength(4)
-    }),
-  )
-
-  it.effect("preserves cache policy hints on reasoning-only assistant messages", () =>
-    Effect.gen(function* () {
-      const prepared = yield* compileRequest(
-        LLM.request({
-          model: OpenRouter.configure({ apiKey: "test-key" }).model("anthropic/claude-sonnet-4.6"),
-          cache: { messages: "latest-assistant" },
-          messages: [Message.user("Think"), Message.assistant([{ type: "reasoning", text: "Reasoning" }])],
-        }),
-      )
-
-      expect(prepared.body.messages).toMatchObject([
-        { role: "user", content: "Think" },
-        { role: "assistant", cache_control: { type: "ephemeral" } },
-      ])
-    }),
-  )
-
-  it.effect("allows usage accounting to be disabled explicitly", () =>
-    Effect.gen(function* () {
-      const prepared = yield* compileRequest(
-        LLM.request({
-          model: OpenRouter.configure({
-            apiKey: "test-key",
-            providerOptions: { openrouter: { usage: false } },
-          }).model("openai/gpt-4o-mini"),
-          cache: "none",
-          prompt: "Hello",
-        }),
-      )
-
-      expect(prepared.body.usage).toEqual({ include: false })
-    }),
-  )
-
   it.effect("applies OpenRouter payload options from the model helper", () =>
     Effect.gen(function* () {
-      const prepared = yield* compileRequest(
+      const prepared = yield* LLMClient.prepare(
         LLM.request({
           model: OpenRouter.configure({
             apiKey: "test-key",
@@ -163,13 +41,6 @@ describe("OpenRouter", () => {
                 usage: true,
                 reasoning: { effort: "high" },
                 promptCacheKey: "session_123",
-                models: ["anthropic/claude-sonnet-4.6", "google/gemini-3.1-pro"],
-                provider: { order: ["anthropic", "google"], require_parameters: true },
-                plugins: [{ id: "response-healing" }],
-                web_search_options: { engine: "native", max_results: 3 },
-                debug: { echo_upstream_body: true },
-                user: "user_123",
-                future_option: { enabled: true },
               },
             },
           }).model("anthropic/claude-3.7-sonnet:thinking"),
@@ -181,54 +52,10 @@ describe("OpenRouter", () => {
         usage: { include: true },
         reasoning: { effort: "high" },
         prompt_cache_key: "session_123",
-        models: ["anthropic/claude-sonnet-4.6", "google/gemini-3.1-pro"],
-        provider: { order: ["anthropic", "google"], require_parameters: true },
-        plugins: [{ id: "response-healing" }],
-        web_search_options: { engine: "native", max_results: 3 },
-        debug: { echo_upstream_body: true },
-        user: "user_123",
-        future_option: { enabled: true },
       })
     }),
   )
 
-  it.effect("filters invalid known OpenRouter options while preserving extensions", () =>
-    Effect.gen(function* () {
-      const invalid: Record<string, unknown> = {
-        usage: "yes",
-        models: "anthropic/claude-sonnet-4.6",
-        provider: [],
-        plugins: {},
-        web_search_options: [],
-        debug: [],
-        user: 123,
-        reasoning: [],
-        promptCacheKey: 123,
-        future_option: { enabled: true },
-      }
-      const prepared = yield* compileRequest(
-        LLM.request({
-          model: OpenRouter.configure({
-            apiKey: "test-key",
-            providerOptions: { openrouter: invalid },
-          }).model("openai/gpt-4o-mini"),
-          prompt: "Hello",
-        }),
-      )
-
-      expect(prepared.body).toMatchObject({ future_option: { enabled: true } })
-      expect(prepared.body).not.toHaveProperty("usage")
-      expect(prepared.body).not.toHaveProperty("models")
-      expect(prepared.body).not.toHaveProperty("provider")
-      expect(prepared.body).not.toHaveProperty("plugins")
-      expect(prepared.body).not.toHaveProperty("web_search_options")
-      expect(prepared.body).not.toHaveProperty("debug")
-      expect(prepared.body).not.toHaveProperty("user")
-      expect(prepared.body).not.toHaveProperty("reasoning")
-      expect(prepared.body).not.toHaveProperty("prompt_cache_key")
-    }),
-  )
-
   it.effect("preserves the upstream provider finish reason", () =>
     Effect.gen(function* () {
       const model = OpenRouter.configure({ apiKey: "test-key" }).model("anthropic/claude-sonnet-4.6")
@@ -273,10 +100,9 @@ describe("OpenRouter", () => {
         { type: "reasoning.text", signature: "signed", format: "anthropic-claude-v1", index: 0 },
         { type: "reasoning.encrypted", data: "opaque", format: "openai-responses-v1", index: 1 },
       ]
-      const prepared = yield* compileRequest(
+      const prepared = yield* LLMClient.prepare<OpenRouter.OpenRouterBody>(
         LLM.request({
           model: OpenRouter.configure({ apiKey: "test-key" }).model("anthropic/claude-sonnet-4.6"),
-          cache: "none",
           messages: [
             Message.assistant([
               {
@@ -307,10 +133,9 @@ describe("OpenRouter", () => {
         { type: "reasoning.encrypted", id: "state", data: "opaque" },
         { type: "reasoning.encrypted", id: "state", data: "opaque" },
       ]
-      const prepared = yield* compileRequest(
+      const prepared = yield* LLMClient.prepare<OpenRouter.OpenRouterBody>(
         LLM.request({
           model: OpenRouter.configure({ apiKey: "test-key" }).model("anthropic/claude-sonnet-4.6"),
-          cache: "none",
           messages: [
             Message.assistant({
               type: "reasoning",
@@ -333,10 +158,9 @@ describe("OpenRouter", () => {
         { type: "reasoning.text", id: "first", index: 0, text: "A", opaque: "first" },
         { type: "reasoning.text", id: "second", index: 1, text: "B", opaque: "second" },
       ]
-      const prepared = yield* compileRequest(
+      const prepared = yield* LLMClient.prepare<OpenRouter.OpenRouterBody>(
         LLM.request({
           model: OpenRouter.configure({ apiKey: "test-key" }).model("anthropic/claude-sonnet-4.6"),
-          cache: "none",
           messages: [
             Message.assistant({
               type: "reasoning",
@@ -355,10 +179,9 @@ describe("OpenRouter", () => {
 
   it.effect("omits scalar reasoning without continuation details", () =>
     Effect.gen(function* () {
-      const prepared = yield* compileRequest(
+      const prepared = yield* LLMClient.prepare<OpenRouter.OpenRouterBody>(
         LLM.request({
           model: OpenRouter.configure({ apiKey: "test-key" }).model("anthropic/claude-sonnet-4.6"),
-          cache: "none",
           messages: [Message.assistant({ type: "reasoning", text: "Thinking" })],
         }),
       )

packages/ai/test/tool-schema-projection.test.ts

@@ -3,8 +3,7 @@ import { Effect } from "effect"
 import { LLM } from "../src"
 import { OpenAIChat } from "../src/protocols"
 import { ToolSchemaProjection } from "../src/protocols/utils/tool-schema"
-import { Auth } from "../src/route"
+import { Auth, LLMClient } from "../src/route"
-import { compileRequest } from "../src/route/client"
 import { it } from "./lib/effect"
 
 describe("tool schema projections", () => {
@@ -80,7 +79,7 @@ describe("tool schema projections", () => {
       const model = OpenAIChat.route
         .with({ endpoint: { baseURL: "https://api.openai.test/v1/" }, auth: Auth.bearer("test") })
         .model({ id: "kimi-k2", compatibility: { toolSchema: "moonshot" } })
-      const prepared = yield* compileRequest(
+      const prepared = yield* LLMClient.prepare<OpenAIChat.OpenAIChatBody>(
         LLM.request({
           model,
           prompt: "Use the tool.",

packages/app/package.json

@@ -59,7 +59,6 @@
     "@opencode-ai/sdk": "file:vendor/opencode-ai-sdk-1.18.8-dev.tgz",
     "@opencode-ai/session-ui": "workspace:*",
     "@opencode-ai/ui": "workspace:*",
-    "@opencode-ai/util": "workspace:*",
     "@pierre/trees": "1.0.0-beta.4",
     "@sentry/solid": "catalog:",
     "@shikijs/transformers": "3.9.2",

packages/app/src/components/dialog-fork.tsx

@@ -69,7 +69,7 @@ export const DialogFork: Component = () => {
     const dir = base64Encode(sdk().directory)
 
     sdk()
-      .api.session.fork({ sessionID, boundary: { type: "before", messageID: item.id } })
+      .api.session.fork({ sessionID, messageID: item.id })
       .then((forked) => {
         dialog.close()
         prompt.set(restored, undefined, { dir, id: forked.id })

packages/app/src/components/titlebar-tab-nav.tsx

@@ -4,7 +4,6 @@ import { createResizeObserver } from "@solid-primitives/resize-observer"
 import { createMutation } from "@tanstack/solid-query"
 import { IconButtonV2 } from "@opencode-ai/ui/v2/icon-button-v2"
 import { Icon as IconV2 } from "@opencode-ai/ui/v2/icon"
-import { displayLabel } from "@opencode-ai/util/session-title-fallback"
 import { useGlobal } from "@/context/global"
 import { ServerConnection, serverName } from "@/context/server"
 import { displayName, projectForSession } from "@/pages/layout/helpers"
@@ -55,10 +54,7 @@ export function TabNavItem(props: {
     if (!session) return
     return projectForSession(session, serverCtx()?.projects.list() ?? [])
   })
-  const title = createMemo(() => {
+  const title = createMemo(() => props.session()?.title ?? props.fallbackTitle)
-    const session = props.session()
-    return session ? displayLabel(session) : props.fallbackTitle
-  })
 
   const projectName = createMemo(() => {
     const session = props.session()
@@ -147,7 +143,7 @@ export function TabNavItem(props: {
     if (!canOpenTabRename(props.dragging, editing(), rename.isPending)) return
     const session = props.session()
     if (!session) return
-    titleEl.textContent = session.title ?? ""
+    titleEl.textContent = session.title
     setEditing(true)
 
     requestAnimationFrame(() => {
@@ -306,7 +302,7 @@ export function TabNavItem(props: {
       }}
       data={{
         projectName: projectName(),
-        title: title(),
+        title: props.session()?.title,
         path: previewPath(),
         serverName: serverLabel(),
       }}

packages/app/src/context/global-sync/bootstrap.ts

@@ -129,8 +129,8 @@ export const loadProjectsQuery = (scope: ServerScope, api: ProjectApi) =>
         api.list().then((projects) => {
           return projects
             .filter((p) => !!p?.id)
-            .map(normalizeProjectInfo)
             .filter((p) => !!p.worktree && !p.worktree.includes("opencode-test"))
+            .map(normalizeProjectInfo)
             .slice()
             .sort((a, b) => cmp(a.id, b.id))
         }),

packages/app/src/context/global-sync/utils.ts

@@ -168,7 +168,6 @@ export function sanitizeProject(project: Project) {
 export function normalizeProjectInfo(project: Project | CurrentProject): Project {
   return {
     ...project,
-    worktree: "canonical" in project ? project.canonical : project.worktree,
     vcs: project.vcs === "git" ? "git" : undefined,
   }
 }

packages/app/src/pages/home/home-session-search-controller.ts

@@ -3,7 +3,6 @@ import { useLanguage } from "@/context/language"
 import { serverName } from "@/context/server"
 import { displayName } from "@/pages/layout/helpers"
 import { makeEventListener } from "@solid-primitives/event-listener"
-import { displayLabel } from "@opencode-ai/util/session-title-fallback"
 import { createMemo, onCleanup } from "solid-js"
 import { createStore } from "solid-js/store"
 import type { HomeController } from "./home-controller"
@@ -24,7 +23,7 @@ export function createHomeSessionSearchController(home: HomeController, sessions
     if (!value) return []
     return sessions.data
       .searchRecords()
-      .filter((record) => `${displayLabel(record.session)} ${record.projectName}`.toLowerCase().includes(value))
+      .filter((record) => `${record.session.title} ${record.projectName}`.toLowerCase().includes(value))
   })
   const active = createMemo(() => {
     const records = results()

packages/app/src/pages/home/home-sessions-view.tsx

@@ -6,10 +6,10 @@ import { ButtonV2 } from "@opencode-ai/ui/v2/button-v2"
 import { Icon as IconV2 } from "@opencode-ai/ui/v2/icon"
 import { IconButtonV2 } from "@opencode-ai/ui/v2/icon-button-v2"
 import { TooltipV2 } from "@opencode-ai/ui/v2/tooltip-v2"
-import { displayLabel } from "@opencode-ai/util/session-title-fallback"
 import { useLanguage } from "@/context/language"
 import { ServerConnection } from "@/context/server"
 import { SessionTabAvatarView } from "@/pages/layout/session-tab-avatar"
+import { sessionTitle } from "@/utils/session-title"
 import { shouldOpenSessionInBackground } from "../home-session-open"
 import {
   HomeSessionStatusController,
@@ -344,7 +344,7 @@ function HomeSessionSearchResultRow(
     selected: boolean
   },
 ) {
-  const title = createMemo(() => displayLabel(props.record.session))
+  const title = createMemo(() => sessionTitle(props.record.session.title) || props.record.session.id)
   const showProjectName = () => props.showProjectName() && props.record.projectName
   const key = () => homeSessionSearchKey(props.record)
 
@@ -415,7 +415,7 @@ function HomeSessionGroupHeader(props: {
 }
 
 function HomeSessionRow(props: HomeSessionsViewProps & { record: HomeSessionRecord }) {
-  const title = createMemo(() => displayLabel(props.record.session))
+  const title = createMemo(() => sessionTitle(props.record.session.title) || props.record.session.id)
   const showProjectName = () => props.showProjectName() && props.record.projectName
 
   return (

packages/app/src/pages/layout/sidebar-items.tsx

@@ -5,7 +5,6 @@ import { Icon as IconV2 } from "@opencode-ai/ui/v2/icon"
 import { IconButton } from "@opencode-ai/ui/icon-button"
 import { Spinner } from "@opencode-ai/ui/spinner"
 import { Tooltip } from "@opencode-ai/ui/tooltip"
-import { displayLabel } from "@opencode-ai/util/session-title-fallback"
 import { getFilename } from "@opencode-ai/core/util/path"
 import { A, useParams } from "@solidjs/router"
 import { type Accessor, createMemo, For, type JSX, Match, Show, Switch } from "solid-js"
@@ -15,6 +14,7 @@ import { getAvatarColors, type LocalProject, useLayout } from "@/context/layout"
 import { useNotification } from "@/context/notification"
 import { usePermission } from "@/context/permission"
 import { messageAgentColor } from "@/utils/agent"
+import { sessionTitle } from "@/utils/session-title"
 import { sessionPermissionRequest } from "../session/composer/session-request-tree"
 import { childSessionOnPath, getProjectAvatarSource, hasProjectPermissions } from "./helpers"
 
@@ -104,7 +104,7 @@ const SessionRow = (props: {
   warmPress: () => void
   warmFocus: () => void
 }): JSX.Element => {
-  const title = () => displayLabel(props.session)
+  const title = () => sessionTitle(props.session.title)
 
   return (
     <A
@@ -229,7 +229,7 @@ export const SessionItem = (props: SessionItemProps): JSX.Element => {
               fallback={
                 <Tooltip
                   placement={props.mobile ? "bottom" : "right"}
-                  value={displayLabel(props.session)}
+                  value={sessionTitle(props.session.title)}
                   gutter={10}
                   class="min-w-0 w-full"
                 >

packages/app/src/pages/session.tsx

@@ -57,7 +57,7 @@ import { ServerConnection, serverName, useServer } from "@/context/server"
 import { useSettings } from "@/context/settings"
 import { useSync } from "@/context/sync"
 import { useTabs } from "@/context/tabs"
-import { TerminalProvider, useTerminal } from "@/context/terminal"
+import { TerminalProvider } from "@/context/terminal"
 import { PromptInput } from "@/components/prompt-input"
 import { PromptInputV2Composer, usePromptInputV2Controller } from "@/components/prompt-input-v2"
 import { useSettingsCommand } from "@/components/settings-dialog"
@@ -70,11 +70,11 @@ import {
   createSessionComposerRegionController,
   SessionComposerRegion,
 } from "@/pages/session/composer"
-import { createOpenReviewFile, createSessionTabs, createSizing, shouldShowFileTree } from "@/pages/session/helpers"
+import { createOpenReviewFile, createSizing, shouldShowFileTree } from "@/pages/session/helpers"
 import { MessageTimeline } from "@/pages/session/timeline/message-timeline"
 import { createTimelineModel } from "@/pages/session/timeline/model"
 import { type DiffStyle, SessionReviewTab, type SessionReviewTabProps } from "@/pages/session/review-tab"
-import { useSessionLayout } from "@/pages/session/session-layout"
+import { createSessionController } from "@/pages/session/session-controller"
 import { restorePromptModel, syncPromptModel, syncSessionModel } from "@/pages/session/session-model-helpers"
 import {
   clampSessionPanelWidth,
@@ -101,7 +101,6 @@ import { extractPromptFromParts } from "@/utils/prompt"
 import { formatServerError, isLocalSessionNotFoundError, isSessionNotFoundError } from "@/utils/server-errors"
 import { legacySessionHref, requireServerKey, sessionHref } from "@/utils/session-route"
 import { useUsageExceededDialogs } from "./session/usage-exceeded-dialogs"
-import { createSessionOwnership } from "./session/session-ownership"
 import { createSessionLineage } from "./session/session-lineage"
 
 type FollowupItem = FollowupDraft & { id: string }
@@ -366,15 +365,28 @@ export default function Page() {
   const prompt = usePrompt()
   const comments = useComments()
   const command = useCommand()
-  const terminal = useTerminal()
   const [searchParams, setSearchParams] = useSearchParams<{ prompt?: string }>()
   const location = useLocation()
   const navigate = useNavigate()
-  const { params, sessionKey, workspaceKey, tabs, view } = useSessionLayout()
+  const isDesktop = createMediaQuery("(min-width: 768px)")
+  const newSessionDesign = createMemo(() => settings.general.newLayoutDesigns())
+  const canReview = createMemo(() => !!sync().project)
+  const session = createSessionController({
+    review: isDesktop,
+    hasReview: canReview,
+    fileBrowser: (sessionID) => newSessionDesign() && isDesktop() && !!sessionID,
+  })
+  const params = session.identity.params
+  const sessionKey = session.identity.sessionKey
+  const workspaceKey = session.identity.workspaceKey
+  const tabs = session.layout.tabs
+  const view = session.layout.view
   const reviewMode = () => view().review.mode() ?? "git"
   const reviewFile = () => view().review.file()
-  const sessionOwnership = createSessionOwnership(sessionKey)
+  const sessionOwnership = session.ownership
-  const newSessionDesign = createMemo(() => settings.general.newLayoutDesigns())
+  const info = session.data.info
+  const isChildSession = session.data.isChild
+  const revertMessageID = session.data.revertMessageID
 
   createEffect(() => {
     if (!prompt.ready()) return
@@ -433,8 +445,8 @@ export default function Page() {
         const current = tabs().tabs()
         if (current.all.length > 0 || current.active) return
 
-        const all = normalizeTabs(from.all)
+        const all = session.tabs.normalizeAll(from.all)
-        const active = from.active ? normalizeTab(from.active) : undefined
+        const active = from.active ? session.tabs.normalize(from.active) : undefined
         tabs().setAll(all)
         tabs().setActive(active && all.includes(active) ? active : all[0])
 
@@ -445,7 +457,6 @@ export default function Page() {
     ),
   )
 
-  const isDesktop = createMediaQuery("(min-width: 768px)")
   const size = createSizing()
   const desktopReviewOpen = createMemo(() => isDesktop() && view().reviewPanel.opened())
   const desktopV2ReviewOpen = createMemo(() => newSessionDesign() && desktopReviewOpen() && !!params.id)
@@ -510,46 +521,16 @@ export default function Page() {
     }),
   )
 
-  function normalizeTab(tab: string) {
-    if (!tab.startsWith("file://")) return tab
-    return file.tab(tab)
-  }
-
-  function normalizeTabs(list: string[]) {
-    const seen = new Set<string>()
-    const next: string[] = []
-    for (const item of list) {
-      const value = normalizeTab(item)
-      if (seen.has(value)) continue
-      seen.add(value)
-      next.push(value)
-    }
-    return next
-  }
-
   const openReviewPanel = () => {
     if (!view().reviewPanel.opened()) view().reviewPanel.open()
   }
 
-  const info = createMemo(() => (params.id ? sync().session.get(params.id) : undefined))
+  const activeTab = session.tabs.activeTab
-  const isChildSession = createMemo(() => !!info()?.parentID)
+  const activeFileTab = session.tabs.activeFileTab
-  const canReview = createMemo(() => !!sync().project)
+  const timeline = createTimelineModel({ session })
-  const reviewTab = createMemo(() => isDesktop())
-  const tabState = createSessionTabs({
-    tabs,
-    pathFromTab: file.pathFromTab,
-    normalizeTab,
-    review: reviewTab,
-    hasReview: canReview,
-  })
-  const activeTab = tabState.activeTab
-  const activeFileTab = tabState.activeFileTab
-  const revertMessageID = createMemo(() => info()?.revert?.messageID)
-  const timeline = createTimelineModel({ sessionID: () => params.id, revertMessageID })
   const historyLoading = timeline.history.loading
   const historyMore = timeline.history.more
   const lastUserMessage = timeline.lastUserMessage
-  const messages = timeline.messages
   const messagesReady = timeline.ready
   const sessionSync = timeline.resource
   const userMessages = timeline.userMessages
@@ -1138,11 +1119,10 @@ export default function Page() {
 
   useComposerCommands()
   useSessionCommands({
+    session,
     navigateMessageByOffset,
     setActiveMessage,
     focusInput,
-    review: reviewTab,
-    fileBrowser: () => newSessionDesign() && isDesktop() && !!params.id,
   })
   command.register("session-palette", () => [
     {
@@ -1692,8 +1672,6 @@ export default function Page() {
     })
   }
 
-  const merge = (next: NonNullable<ReturnType<typeof info>>, target = sync()) => target.session.remember(next)
-
   const roll = (sessionID: string, next: NonNullable<ReturnType<typeof info>>["revert"], target = sync()) => {
     const session = target.session.get(sessionID)
     if (!session) return
@@ -1754,7 +1732,7 @@ export default function Page() {
   const queueEnabled = createMemo(() => {
     const id = params.id
     if (!id) return false
-    return settings.general.followup() === "queue" && busy(id) && !composer.blocked() && !isChildSession()
+    return settings.general.followup() === "queue" && session.data.working() && !composer.blocked() && !isChildSession()
   })
 
   const followupText = (item: FollowupDraft) => {
@@ -1932,7 +1910,7 @@ export default function Page() {
     if (followup.paused[sessionID]) return
     if (isChildSession()) return
     if (composer.blocked()) return
-    if (busy(sessionID)) return
+    if (session.data.working()) return
 
     void sendFollowup(sessionID, item.id)
   })
@@ -2080,6 +2058,7 @@ export default function Page() {
             <Show when={messagesReady() ? params.id : undefined} keyed>
               {(_id) => (
                 <MessageTimeline
+                  session={session}
                   actions={actions}
                   scroll={ui.scroll}
                   onResumeScroll={resumeScroll}
@@ -2157,7 +2136,7 @@ export default function Page() {
                 : undefined,
             onResponseSubmit: resumeScroll,
             openParent: () => {
-              const id = info()?.parentID
+              const id = session.data.parentID()
               if (!id) return
               navigate(
                 params.serverKey

packages/app/src/pages/session/session-controller.test.ts

@@ -0,0 +1,71 @@
+import { describe, expect, test } from "bun:test"
+import type { AssistantMessage, Message, UserMessage } from "@opencode-ai/sdk/v2"
+import { createRoot, createSignal } from "solid-js"
+import {
+  normalizeSessionTab,
+  normalizeSessionTabs,
+  selectSessionUserMessages,
+  selectVisibleSessionUserMessages,
+} from "./session-domain"
+import { createSessionOwnership } from "./session-ownership"
+
+const user = (id: string): UserMessage => ({
+  id,
+  sessionID: "session",
+  role: "user",
+  time: { created: 0 },
+  agent: "build",
+  model: { providerID: "provider", modelID: "model" },
+})
+
+const assistant: AssistantMessage = {
+  id: "msg_2",
+  sessionID: "session",
+  role: "assistant",
+  time: { created: 0 },
+  parentID: "msg_1",
+  modelID: "model",
+  providerID: "provider",
+  mode: "build",
+  agent: "build",
+  path: { cwd: "/workspace", root: "/workspace" },
+  cost: 0,
+  tokens: { input: 0, output: 0, reasoning: 0, cache: { read: 0, write: 0 } },
+}
+
+describe("session controller invariants", () => {
+  test("normalizes file tabs once while preserving non-file tabs and order", () => {
+    const normalize = (tab: string) => normalizeSessionTab(tab, (value) => value.toLowerCase())
+
+    expect(normalizeSessionTabs(["review", "file://SRC/A.TS", "file://src/a.ts", "context"], normalize)).toEqual([
+      "review",
+      "file://src/a.ts",
+      "context",
+    ])
+  })
+
+  test("selects user history strictly before the revert boundary", () => {
+    const messages: Message[] = [user("msg_1"), assistant, user("msg_3"), user("msg_5")]
+    const users = selectSessionUserMessages(messages)
+
+    expect(users.map((message) => message.id)).toEqual(["msg_1", "msg_3", "msg_5"])
+    expect(selectVisibleSessionUserMessages(users, "msg_3").map((message) => message.id)).toEqual(["msg_1"])
+    expect(selectVisibleSessionUserMessages(users)).toBe(users)
+  })
+
+  test("rejects work captured by a previous session", () => {
+    createRoot((dispose) => {
+      const [key, setKey] = createSignal("session-a")
+      const ownership = createSessionOwnership(key)
+      const captured = ownership.capture()
+      let ran = false
+
+      setKey("session-b")
+
+      expect(captured.current()).toBe(false)
+      expect(captured.run(() => (ran = true))).toBeUndefined()
+      expect(ran).toBe(false)
+      dispose()
+    })
+  })
+})

packages/app/src/pages/session/session-controller.ts

@@ -0,0 +1,101 @@
+import type { Message, UserMessage } from "@opencode-ai/sdk/v2"
+import { createMemo, type Accessor } from "solid-js"
+import { useFile } from "@/context/file"
+import { useSync } from "@/context/sync"
+import { same } from "@/utils/same"
+import { createSessionTabs } from "./helpers"
+import {
+  normalizeSessionTab,
+  normalizeSessionTabs,
+  selectSessionUserMessages,
+  selectVisibleSessionUserMessages,
+} from "./session-domain"
+import { useSessionLayout } from "./session-layout"
+import { createSessionOwnership } from "./session-ownership"
+
+const emptyMessages: Message[] = []
+const emptyUserMessages: UserMessage[] = []
+const idle = { type: "idle" as const }
+
+export function createSessionController(input: {
+  review?: Accessor<boolean>
+  hasReview?: Accessor<boolean>
+  fileBrowser?: (sessionID: string | undefined) => boolean
+}) {
+  const file = useFile()
+  const sync = useSync()
+  const layout = useSessionLayout()
+  const sessionID = createMemo(() => layout.params.id)
+  const info = createMemo(() => {
+    const id = sessionID()
+    return id ? sync().session.get(id) : undefined
+  })
+  const parentID = createMemo(() => info()?.parentID)
+  const parent = createMemo(() => {
+    const id = parentID()
+    return id ? sync().session.get(id) : undefined
+  })
+  const status = createMemo(() => {
+    const id = sessionID()
+    return id ? (sync().data.session_status[id] ?? idle) : idle
+  })
+  const messages = createMemo(() => {
+    const id = sessionID()
+    return id ? (sync().data.message[id] ?? emptyMessages) : emptyMessages
+  })
+  const userMessages = createMemo(() => selectSessionUserMessages(messages()), emptyUserMessages, { equals: same })
+  const revertMessageID = createMemo(() => info()?.revert?.messageID)
+  const visibleUserMessages = createMemo(
+    () => selectVisibleSessionUserMessages(userMessages(), revertMessageID()),
+    emptyUserMessages,
+    { equals: same },
+  )
+  const normalizeTab = (tab: string) => normalizeSessionTab(tab, file.tab)
+  const tabs = createSessionTabs({
+    tabs: layout.tabs,
+    pathFromTab: file.pathFromTab,
+    normalizeTab,
+    review: input.review,
+    hasReview: input.hasReview,
+    fileBrowser: input.fileBrowser ? () => input.fileBrowser?.(sessionID()) ?? false : undefined,
+  })
+
+  return {
+    identity: {
+      params: layout.params,
+      sessionID,
+      sessionKey: layout.sessionKey,
+      workspaceKey: layout.workspaceKey,
+    },
+    data: {
+      info,
+      parent,
+      parentID,
+      isChild: createMemo(() => !!parentID()),
+      status,
+      working: createMemo(() => {
+        const id = sessionID()
+        return id ? sync().data.session_working(id) : false
+      }),
+      revertMessageID,
+    },
+    history: {
+      messages,
+      userMessages,
+      visibleUserMessages,
+      lastUserMessage: createMemo(() => visibleUserMessages().at(-1)),
+    },
+    layout: {
+      tabs: layout.tabs,
+      view: layout.view,
+    },
+    ownership: createSessionOwnership(layout.sessionKey),
+    tabs: {
+      ...tabs,
+      normalize: normalizeTab,
+      normalizeAll: (values: string[]) => normalizeSessionTabs(values, normalizeTab),
+    },
+  }
+}
+
+export type SessionController = ReturnType<typeof createSessionController>

packages/app/src/pages/session/session-domain.ts

@@ -0,0 +1,19 @@
+import type { Message, UserMessage } from "@opencode-ai/sdk/v2"
+
+export function normalizeSessionTab(tab: string, normalizeFileTab: (tab: string) => string) {
+  if (!tab.startsWith("file://")) return tab
+  return normalizeFileTab(tab)
+}
+
+export function normalizeSessionTabs(tabs: string[], normalize: (tab: string) => string) {
+  return [...new Set(tabs.map(normalize))]
+}
+
+export function selectSessionUserMessages(messages: Message[]) {
+  return messages.filter((message): message is UserMessage => message.role === "user")
+}
+
+export function selectVisibleSessionUserMessages(messages: UserMessage[], revertMessageID?: string) {
+  if (!revertMessageID) return messages
+  return messages.filter((message) => message.id < revertMessageID)
+}

packages/app/src/pages/session/timeline/controller-projection.ts

@@ -0,0 +1,26 @@
+export function timelineChildTitle(input: {
+  parentID?: string
+  taskDescription?: string
+  title?: string
+  fallback: string
+}) {
+  if (!input.parentID) return input.title ?? ""
+  if (input.taskDescription) return input.taskDescription
+  return input.title?.replace(/\s+\(@[^)]+ subagent\)$/, "") || input.fallback
+}
+
+export function timelineRemovedSessionIDs(sessions: readonly { id: string; parentID?: string }[], sessionID: string) {
+  const removed = new Set([sessionID])
+  const byParent = Map.groupBy(
+    sessions.filter((session) => session.parentID),
+    (session) => session.parentID!,
+  )
+  const visit = (id: string) =>
+    byParent.get(id)?.forEach((child) => {
+      if (removed.has(child.id)) return
+      removed.add(child.id)
+      visit(child.id)
+    })
+  visit(sessionID)
+  return removed
+}

packages/app/src/pages/session/timeline/controller.test.ts

@@ -0,0 +1,30 @@
+import { describe, expect, test } from "bun:test"
+import { timelineChildTitle, timelineRemovedSessionIDs } from "./controller-projection"
+
+describe("timeline controller", () => {
+  test("projects child titles from task descriptions and session fallbacks", () => {
+    expect(timelineChildTitle({ title: "Root", fallback: "New session" })).toBe("Root")
+    expect(
+      timelineChildTitle({ parentID: "parent", taskDescription: "Investigate timeline", fallback: "New session" }),
+    ).toBe("Investigate timeline")
+    expect(
+      timelineChildTitle({ parentID: "parent", title: "Fallback (@build subagent)", fallback: "New session" }),
+    ).toBe("Fallback")
+    expect(timelineChildTitle({ parentID: "parent", fallback: "New session" })).toBe("New session")
+  })
+
+  test("collects the removed session and all descendants", () => {
+    const removed = timelineRemovedSessionIDs(
+      [
+        { id: "root" },
+        { id: "child", parentID: "root" },
+        { id: "grandchild", parentID: "child" },
+        { id: "sibling", parentID: "root" },
+        { id: "unrelated" },
+      ],
+      "root",
+    )
+
+    expect([...removed]).toEqual(["root", "child", "grandchild", "sibling"])
+  })
+})

packages/app/src/pages/session/timeline/controller.tsx

@@ -0,0 +1,328 @@
+import type { Message, Part, UserMessage } from "@opencode-ai/sdk/v2"
+import { Button } from "@opencode-ai/ui/button"
+import { Dialog } from "@opencode-ai/ui/dialog"
+import { DialogFooter, DialogHeader, DialogTitleGroup, DialogV2 } from "@opencode-ai/ui/v2/dialog-v2"
+import { ButtonV2 } from "@opencode-ai/ui/v2/button-v2"
+import { useNavigate } from "@solidjs/router"
+import { createEffect, createMemo, on, type Accessor } from "solid-js"
+import { createStore, produce } from "solid-js/store"
+import { notifySessionTabsRemoved } from "@/components/titlebar-session-events"
+import { useDialog } from "@opencode-ai/ui/context/dialog"
+import { useLanguage } from "@/context/language"
+import { usePlatform } from "@/context/platform"
+import { useServerSDK } from "@/context/server-sdk"
+import { useSettings } from "@/context/settings"
+import { useSDK } from "@/context/sdk"
+import { useSync } from "@/context/sync"
+import { useTabs } from "@/context/tabs"
+import type { SessionController } from "@/pages/session/session-controller"
+import { legacySessionHref, requireServerKey, sessionHref } from "@/utils/session-route"
+import { sessionTitle } from "@/utils/session-title"
+import { showToast } from "@/utils/toast"
+import { timelineChildTitle, timelineRemovedSessionIDs } from "./controller-projection"
+import { createTimelineProjection } from "./projection"
+
+const emptyMessages: Message[] = []
+const emptyParts: Part[] = []
+const taskDescription = (part: Part, sessionID: string): string | undefined => {
+  if (part.type !== "tool" || part.tool !== "task") return undefined
+  const metadata = "metadata" in part.state ? part.state.metadata : undefined
+  if (metadata?.sessionId !== sessionID) return undefined
+  const value = part.state.input?.description
+  if (typeof value === "string" && value) return value
+  return undefined
+}
+
+export type TimelineSessionSource = {
+  identity: Pick<SessionController["identity"], "params" | "sessionID" | "sessionKey">
+  data: Pick<SessionController["data"], "info" | "parent" | "parentID" | "status">
+  history: Pick<SessionController["history"], "messages">
+}
+
+export function createTimelineController(input: {
+  session: TimelineSessionSource
+  userMessages: Accessor<UserMessage[]>
+}) {
+  const navigate = useNavigate()
+  const serverSDK = useServerSDK()
+  const sdk = useSDK()
+  const sync = useSync()
+  const settings = useSettings()
+  const tabs = useTabs()
+  const dialog = useDialog()
+  const language = useLanguage()
+  const platform = usePlatform()
+  const params = input.session.identity.params
+  const sessionKey = input.session.identity.sessionKey
+  const sessionID = input.session.identity.sessionID
+  const status = input.session.data.status
+  const messages = input.session.history.messages
+  const projectedMessages = createMemo(() => {
+    const id = sessionID()
+    if (!id) return []
+    const visible = new Set(input.userMessages().map((message) => message.id))
+    const boundary = messages().find((message) => message.role === "user" && !visible.has(message.id))?.id
+    const projected = sync().data.session_message[id] ?? []
+    return boundary ? projected.filter((message) => message.id < boundary) : projected
+  })
+  const info = input.session.data.info
+  const titleValue = createMemo(() => info()?.title)
+  const titleLabel = createMemo(() => sessionTitle(titleValue()))
+  const shareUrl = createMemo(() => info()?.share?.url)
+  const shareEnabled = createMemo(() => sync().data.config.share !== "disabled")
+  const parentID = input.session.data.parentID
+  const parent = input.session.data.parent
+  const parentMessages = createMemo(() => {
+    const id = parentID()
+    return id ? (sync().data.message[id] ?? emptyMessages) : emptyMessages
+  })
+  const parentTitle = createMemo(() => sessionTitle(parent()?.title) ?? language.t("command.session.new"))
+  const parts = (messageID: string) => sync().data.part[messageID] ?? emptyParts
+  const part = (messageID: string, partID: string) => parts(messageID).find((item) => item.id === partID)
+  const childTaskDescription = createMemo(() => {
+    const id = sessionID()
+    if (!id) return undefined
+    return parentMessages()
+      .flatMap((message) => parts(message.id))
+      .map((item) => taskDescription(item, id))
+      .findLast((value): value is string => !!value)
+  })
+  const childTitle = createMemo(() => {
+    return timelineChildTitle({
+      parentID: parentID(),
+      taskDescription: childTaskDescription(),
+      title: titleLabel(),
+      fallback: language.t("command.session.new"),
+    })
+  })
+  const showHeader = createMemo(() => !!(titleValue() || parentID()))
+  const projection = createTimelineProjection({
+    messages,
+    userMessages: input.userMessages,
+    sessionMessages: projectedMessages,
+    parts,
+    status,
+    showReasoningSummaries: settings.general.showReasoningSummaries,
+    inlineComments: settings.general.newLayoutDesigns,
+  })
+  const [pending, setPending] = createStore({ rename: false, share: false, unshare: false })
+
+  const errorMessage = (error: unknown) => {
+    if (error && typeof error === "object" && "data" in error) {
+      const data = error.data
+      if (data && typeof data === "object" && "message" in data && typeof data.message === "string") return data.message
+    }
+    if (error instanceof Error) return error.message
+    return language.t("common.requestFailed")
+  }
+  const rename = async (title: string) => {
+    const id = sessionID()
+    if (!id || pending.rename) return false
+    const next = title.trim()
+    if (!next || next === (titleLabel() ?? "")) return true
+    setPending("rename", true)
+    const success = await sdk()
+      .api.session.rename({ sessionID: id, title: next })
+      .then(() => true)
+      .catch((error) => {
+        showToast({ title: language.t("common.requestFailed"), description: errorMessage(error) })
+        return false
+      })
+    setPending("rename", false)
+    if (!success) return false
+    sync().set(
+      produce((draft) => {
+        const index = draft.session.findIndex((session) => session.id === id)
+        if (index !== -1) draft.session[index].title = next
+      }),
+    )
+    return true
+  }
+  const share = async () => {
+    const id = sessionID()
+    if (!id || pending.share || !shareEnabled()) return
+    setPending("share", true)
+    await serverSDK()
+      .client.session.share({ sessionID: id })
+      .catch((error) => console.error("Failed to share session", error))
+    setPending("share", false)
+  }
+  const unshare = async () => {
+    const id = sessionID()
+    if (!id || pending.unshare || !shareEnabled()) return
+    setPending("unshare", true)
+    await serverSDK()
+      .client.session.unshare({ sessionID: id })
+      .catch((error) => console.error("Failed to unshare session", error))
+    setPending("unshare", false)
+  }
+  const href = (id: string) =>
+    params.serverKey ? sessionHref(requireServerKey(params.serverKey), id) : legacySessionHref(sdk().directory, id)
+  const navigateAfterRemoval = (id: string, parent?: string, next?: string) => {
+    if (params.id !== id) return
+    if (parent) return navigate(href(parent))
+    if (next) return navigate(href(next))
+    if (params.serverKey)
+      return tabs.newDraft({ server: requireServerKey(params.serverKey), directory: sdk().directory })
+    navigate(`/${params.dir}/session`)
+  }
+  const archive = async (id: string) => {
+    const session = sync().session.get(id)
+    if (!session || (await sdk().protocol) !== "v1") return
+    const index = sync().data.session.findIndex((item) => item.id === id)
+    const next = index === -1 ? undefined : (sync().data.session[index + 1] ?? sync().data.session[index - 1])
+    await sdk()
+      .client.session.update({ sessionID: id, directory: sdk().directory, time: { archived: Date.now() } })
+      .then(() => {
+        sync().set(
+          produce((draft) => {
+            const index = draft.session.findIndex((item) => item.id === id)
+            if (index !== -1) draft.session.splice(index, 1)
+          }),
+        )
+        sync().session.evict(id)
+        void navigateAfterRemoval(id, session.parentID, next?.id)
+        notifySessionTabsRemoved({ directory: sdk().directory, sessionIDs: [id] })
+      })
+      .catch((error) => showToast({ title: language.t("common.requestFailed"), description: errorMessage(error) }))
+  }
+  const remove = async (id: string) => {
+    const session = sync().session.get(id)
+    if (!session) return false
+    const sessions = sync().data.session.filter((item) => !item.parentID && !item.time?.archived)
+    const index = sessions.findIndex((item) => item.id === id)
+    const next = index === -1 ? undefined : (sessions[index + 1] ?? sessions[index - 1])
+    const success = await sdk()
+      .api.session.remove({ sessionID: id })
+      .then(() => true)
+      .catch((error) => {
+        showToast({ title: language.t("session.delete.failed.title"), description: errorMessage(error) })
+        return false
+      })
+    if (!success) return false
+    const removed = timelineRemovedSessionIDs(sync().data.session, id)
+    void navigateAfterRemoval(id, session.parentID, next?.id)
+    sync().set(produce((draft) => void (draft.session = draft.session.filter((item) => !removed.has(item.id)))))
+    removed.forEach((sessionID) => sync().session.evict(sessionID))
+    notifySessionTabsRemoved({ directory: sdk().directory, sessionIDs: [...removed] })
+    return true
+  }
+
+  function DeleteDialog(props: { sessionID: string }) {
+    const name = createMemo(
+      () => sessionTitle(sync().session.get(props.sessionID)?.title) ?? language.t("command.session.new"),
+    )
+    const confirm = async () => {
+      await remove(props.sessionID)
+      dialog.close()
+    }
+    if (settings.general.newLayoutDesigns())
+      return (
+        <DialogV2 fit>
+          <DialogHeader hideClose>
+            <DialogTitleGroup
+              title={language.t("session.delete.title")}
+              description={language.t("session.delete.confirm", { name: name() })}
+            />
+          </DialogHeader>
+          <DialogFooter>
+            <ButtonV2 variant="ghost" onClick={() => dialog.close()}>
+              {language.t("common.cancel")}
+            </ButtonV2>
+            <ButtonV2 variant="danger" onClick={confirm}>
+              {language.t("session.delete.button")}
+            </ButtonV2>
+          </DialogFooter>
+        </DialogV2>
+      )
+    return (
+      <Dialog title={language.t("session.delete.title")} fit>
+        <div class="flex flex-col gap-4 pl-6 pr-2.5 pb-3">
+          <div class="flex flex-col gap-1">
+            <span class="text-14-regular text-text-strong">
+              {language.t("session.delete.confirm", { name: name() })}
+            </span>
+          </div>
+          <div class="flex justify-end gap-2">
+            <Button variant="ghost" size="large" onClick={() => dialog.close()}>
+              {language.t("common.cancel")}
+            </Button>
+            <Button variant="primary" size="large" onClick={confirm}>
+              {language.t("session.delete.button")}
+            </Button>
+          </div>
+        </div>
+      </Dialog>
+    )
+  }
+
+  createEffect(
+    on(
+      () => [parentID(), childTaskDescription()] as const,
+      ([id, description]) => {
+        if (!id || description || sync().data.message[id] !== undefined) return
+        void sync().session.sync(id)
+      },
+      { defer: true },
+    ),
+  )
+
+  return {
+    data: {
+      sessionKey,
+      sessionID,
+      status,
+      titleValue,
+      titleLabel,
+      shareUrl,
+      shareEnabled,
+      parentID,
+      parentTitle,
+      childTitle,
+      showHeader,
+      parts,
+      part,
+      projection,
+      newLayoutDesigns: settings.general.newLayoutDesigns,
+      showReasoningSummaries: settings.general.showReasoningSummaries,
+      shellToolPartsExpanded: settings.general.shellToolPartsExpanded,
+      editToolPartsExpanded: settings.general.editToolPartsExpanded,
+    },
+    pending: {
+      rename: () => pending.rename,
+      share: () => pending.share,
+      unshare: () => pending.unshare,
+    },
+    action: {
+      rename,
+      share,
+      unshare,
+      archive,
+      showDelete: (id: string) => dialog.show(() => <DeleteDialog sessionID={id} />),
+      navigateParent: () => {
+        const id = parentID()
+        if (id) navigate(href(id))
+      },
+      viewShare: () => {
+        const url = shareUrl()
+        if (url) platform.openLink(url)
+      },
+      copyShareUrl: async () => {
+        const url = shareUrl()
+        if (!url) return
+        await navigator.clipboard.writeText(url).then(
+          () =>
+            showToast({
+              variant: "success",
+              icon: "circle-check",
+              title: language.t("session.share.copy.copied"),
+              description: url,
+            }),
+          (error) => showToast({ title: language.t("common.requestFailed"), description: errorMessage(error) }),
+        )
+      },
+    },
+  }
+}
+
+export type TimelineController = ReturnType<typeof createTimelineController>

packages/app/src/pages/session/timeline/message-timeline.tsx

@@ -11,10 +11,8 @@ import {
   type Accessor,
   type JSX,
 } from "solid-js"
-import { createStore, produce } from "solid-js/store"
+import { createStore } from "solid-js/store"
 import { Dynamic } from "solid-js/web"
-import { useNavigate } from "@solidjs/router"
-import { useMutation } from "@tanstack/solid-query"
 import { createVirtualizer, defaultRangeExtractor, elementScroll, type VirtualItem } from "@tanstack/solid-virtual"
 import { Accordion } from "@opencode-ai/ui/accordion"
 import { Button } from "@opencode-ai/ui/button"
@@ -35,8 +33,6 @@ import { Icon as IconV2 } from "@opencode-ai/ui/v2/icon"
 import { IconButtonV2 } from "@opencode-ai/ui/v2/icon-button-v2"
 import { DropdownMenu } from "@opencode-ai/ui/dropdown-menu"
 import { MenuV2 } from "@opencode-ai/ui/v2/menu-v2"
-import { Dialog } from "@opencode-ai/ui/dialog"
-import { DialogFooter, DialogHeader, DialogTitleGroup, DialogV2 } from "@opencode-ai/ui/v2/dialog-v2"
 import { InlineInput } from "@opencode-ai/ui/inline-input"
 import { ButtonV2 } from "@opencode-ai/ui/v2/button-v2"
 import { SessionRetry } from "@opencode-ai/session-ui/session-retry"
@@ -45,43 +41,22 @@ import { StickyAccordionHeader } from "@opencode-ai/ui/sticky-accordion-header"
 import { TextField } from "@opencode-ai/ui/text-field"
 import { TextReveal } from "@opencode-ai/ui/text-reveal"
 import { TextShimmer } from "@opencode-ai/ui/text-shimmer"
-import type {
+import type { AssistantMessage, ToolPart, UserMessage } from "@opencode-ai/sdk/v2"
-  AssistantMessage,
-  Message as MessageType,
-  Part as PartType,
-  ToolPart,
-  UserMessage,
-} from "@opencode-ai/sdk/v2"
-import { showToast } from "@/utils/toast"
 import { getDirectory, getFilename } from "@opencode-ai/core/util/path"
 import { Popover as KobaltePopover } from "@kobalte/core/popover"
 import { normalize } from "@opencode-ai/session-ui/session-diff"
 import { useFileComponent } from "@opencode-ai/ui/context/file"
 import { shouldMarkBoundaryGesture, normalizeWheelDelta } from "@/pages/session/message-gesture"
 import { SessionContextUsage } from "@/components/session-context-usage"
-import { useDialog } from "@opencode-ai/ui/context/dialog"
 import { useLanguage } from "@/context/language"
-import { useSessionKey } from "@/pages/session/session-layout"
-import { useServerSDK } from "@/context/server-sdk"
-import { usePlatform } from "@/context/platform"
-import { useSettings } from "@/context/settings"
-import { useTabs } from "@/context/tabs"
-import { legacySessionHref, requireServerKey, sessionHref } from "@/utils/session-route"
-import { useSDK } from "@/context/sdk"
-import { useSync } from "@/context/sync"
-import { notifySessionTabsRemoved } from "@/components/titlebar-session-events"
-import { displayLabel } from "@opencode-ai/util/session-title-fallback"
 import { scheduleConnectedMeasure } from "./measure"
 import { observeElementOffsetReconnectAware } from "./observe-element-offset"
-import { createTimelineProjection } from "./projection"
 import { MessageComment, SummaryDiff, TimelineRow, TimelineRowMap } from "./rows"
 import { filterVirtualIndexes } from "./virtual-items"
+import { createTimelineController, type TimelineController, type TimelineSessionSource } from "./controller"
 
-const emptyMessages: MessageType[] = []
-const emptyParts: PartType[] = []
 const emptyTools: ToolPart[] = []
 const emptyAssistantMessages: AssistantMessage[] = []
-const idle = { type: "idle" as const }
 
 type FramedTimelineRow = Exclude<TimelineRow.TimelineRow, { _tag: "TurnGap" }>
 type TimelineRowByTag<T extends TimelineRow.TimelineRow["_tag"]> = Extract<TimelineRow.TimelineRow, { _tag: T }>
@@ -89,14 +64,6 @@ type TimelineRowByTag<T extends TimelineRow.TimelineRow["_tag"]> = Extract<Timel
 const timelineFallbackItemSize = 60
 const timelineCache = new Map<string, { measurements: VirtualItem[]; toolOpen: Record<string, boolean | undefined> }>()
 
-const taskDescription = (part: PartType, sessionID: string) => {
-  if (part.type !== "tool" || part.tool !== "task") return
-  const metadata = "metadata" in part.state ? part.state.metadata : undefined
-  if (metadata?.sessionId !== sessionID) return
-  const value = part.state.input?.description
-  if (typeof value === "string" && value) return value
-}
-
 const boundaryTarget = (root: HTMLElement, target: EventTarget | null) => {
   const current = target instanceof Element ? target : undefined
   const nested = current?.closest("[data-scrollable]")
@@ -237,7 +204,8 @@ function TimelineDiffView(props: { diff: SummaryDiff }) {
   )
 }
 
-export function MessageTimeline(props: {
+type MessageTimelineProps = {
+  session: TimelineSessionSource
   actions?: UserActions
   scroll: { overflow: boolean; bottom: boolean; jump: boolean }
   onResumeScroll: () => void
@@ -257,94 +225,42 @@ export function MessageTimeline(props: {
   setRevealMessage?: (fn: (id: string) => void) => void
   setScrollToEnd?: (fn: () => void) => void
   setHistoryAnchor?: (handlers: { capture: () => void; restore: (done: boolean) => void }) => void
-}) {
+}
-  let touchGesture: number | undefined
 
-  const navigate = useNavigate()
+export function MessageTimeline(props: MessageTimelineProps) {
-  const serverSDK = useServerSDK()
+  const controller = createTimelineController({ session: props.session, userMessages: () => props.userMessages })
-  const sdk = useSDK()
+  return (
-  const sync = useSync()
+    <MessageTimelineView {...props} data={controller.data} action={controller.action} pending={controller.pending} />
-  const settings = useSettings()
+  )
-  const tabs = useTabs()
+}
-  const dialog = useDialog()
+
+function MessageTimelineView(
+  props: MessageTimelineProps & {
+    data: TimelineController["data"]
+    action: TimelineController["action"]
+    pending: TimelineController["pending"]
+  },
+) {
+  let touchGesture: number | undefined
   const language = useLanguage()
-  const { params, sessionKey } = useSessionKey()
+  const ownerSessionKey = props.data.sessionKey()
-  const ownerSessionKey = sessionKey()
   const cached = timelineCache.get(ownerSessionKey)
   const initialMeasurements = cached?.measurements
   const coldBottomMount = !initialMeasurements?.length && props.shouldAnchorBottom()
-  const platform = usePlatform()
 
   const [listRoot, setListRoot] = createSignal<HTMLDivElement>()
-  const sessionID = createMemo(() => params.id)
+  const sessionID = props.data.sessionID
-  const sessionStatus = createMemo(() => {
+  const sessionStatus = props.data.status
-    const id = sessionID()
+  const titleLabel = props.data.titleLabel
-    if (!id) return idle
+  const shareUrl = props.data.shareUrl
-    return sync().data.session_status[id] ?? idle
+  const shareEnabled = props.data.shareEnabled
-  })
+  const parentID = props.data.parentID
-  const sessionMessages = createMemo(() => (sessionID() ? (sync().data.message[sessionID()!] ?? []) : []))
+  const parentTitle = props.data.parentTitle
-  const projectedMessages = createMemo(() => {
+  const childTitle = props.data.childTitle
-    const id = sessionID()
+  const showHeader = props.data.showHeader
-    if (!id) return []
+  const getMsgParts = props.data.parts
-    const visible = new Set(props.userMessages.map((message) => message.id))
+  const getMsgPart = props.data.part
-    const boundary = sessionMessages().find((message) => message.role === "user" && !visible.has(message.id))?.id
+  const projection = props.data.projection
-    const messages = sync().data.session_message[id] ?? []
-    return boundary ? messages.filter((message) => message.id < boundary) : messages
-  })
-  const info = createMemo(() => {
-    const id = sessionID()
-    if (!id) return
-    return sync().session.get(id)
-  })
-  const titleLabel = createMemo(() => {
-    const session = info()
-    if (!session) return
-    return displayLabel(session)
-  })
-  const shareUrl = createMemo(() => info()?.share?.url)
-  const shareEnabled = createMemo(() => sync().data.config.share !== "disabled")
-  const parentID = createMemo(() => info()?.parentID)
-  const parent = createMemo(() => {
-    const id = parentID()
-    if (!id) return
-    return sync().session.get(id)
-  })
-  const parentMessages = createMemo(() => {
-    const id = parentID()
-    if (!id) return emptyMessages
-    return sync().data.message[id] ?? emptyMessages
-  })
-  const parentTitle = createMemo(() => {
-    const session = parent()
-    return session ? displayLabel(session) : language.t("command.session.new")
-  })
-  const getMsgParts = (msgId: string) => sync().data.part[msgId] ?? emptyParts
-  const getMsgPart = (messageID: string, partID: string) => getMsgParts(messageID).find((part) => part.id === partID)
-  const childTaskDescription = createMemo(() => {
-    const id = sessionID()
-    if (!id) return
-    return parentMessages()
-      .flatMap((message) => getMsgParts(message.id))
-      .map((part) => taskDescription(part, id))
-      .findLast((value): value is string => !!value)
-  })
-  const childTitle = createMemo(() => {
-    if (!parentID()) return titleLabel() ?? ""
-    if (childTaskDescription()) return childTaskDescription()
-    const value = titleLabel()?.replace(/\s+\(@[^)]+ subagent\)$/, "")
-    if (value) return value
-    return language.t("command.session.new")
-  })
-  const showHeader = createMemo(() => !!(titleLabel() || parentID()))
-  const projection = createTimelineProjection({
-    messages: sessionMessages,
-    userMessages: () => props.userMessages,
-    sessionMessages: projectedMessages,
-    parts: getMsgParts,
-    status: sessionStatus,
-    showReasoningSummaries: settings.general.showReasoningSummaries,
-    inlineComments: settings.general.newLayoutDesigns,
-  })
   const activeMessageID = projection.activeMessageID
   const assistantMessagesByParent = projection.assistantMessagesByParent
   const lastAssistantGroupKey = projection.lastAssistantGroupKey
@@ -534,9 +450,9 @@ export function MessageTimeline(props: {
     virtualizer.scrollToEnd()
   }
 
-  let measuredSessionKey = sessionKey()
+  let measuredSessionKey = props.data.sessionKey()
   createEffect(() => {
-    const key = sessionKey()
+    const key = props.data.sessionKey()
     timelineRows().length
     if (measuredSessionKey !== key) {
       measuredSessionKey = key
@@ -649,88 +565,6 @@ export function MessageTimeline(props: {
     props.setScrollRef(undefined)
   })
 
-  const viewShare = () => {
-    const url = shareUrl()
-    if (!url) return
-    platform.openLink(url)
-  }
-
-  const errorMessage = (err: unknown) => {
-    if (err && typeof err === "object" && "data" in err) {
-      const data = (err as { data?: { message?: string } }).data
-      if (data?.message) return data.message
-    }
-    if (err instanceof Error) return err.message
-    return language.t("common.requestFailed")
-  }
-
-  const shareMutation = useMutation(() => ({
-    mutationFn: (id: string) => serverSDK().client.session.share({ sessionID: id }),
-    onError: (err) => {
-      console.error("Failed to share session", err)
-    },
-  }))
-
-  const unshareMutation = useMutation(() => ({
-    mutationFn: (id: string) => serverSDK().client.session.unshare({ sessionID: id }),
-    onError: (err) => {
-      console.error("Failed to unshare session", err)
-    },
-  }))
-
-  const titleMutation = useMutation(() => ({
-    mutationFn: (input: { id: string; title: string }) =>
-      sdk().api.session.rename({ sessionID: input.id, title: input.title }),
-    onSuccess: (_, input) => {
-      sync().set(
-        produce((draft) => {
-          const index = draft.session.findIndex((s) => s.id === input.id)
-          if (index !== -1) draft.session[index].title = input.title
-        }),
-      )
-      setTitle("editing", false)
-    },
-    onError: (err) => {
-      showToast({
-        title: language.t("common.requestFailed"),
-        description: errorMessage(err),
-      })
-    },
-  }))
-
-  const shareSession = () => {
-    const id = sessionID()
-    if (!id || shareMutation.isPending) return
-    if (!shareEnabled()) return
-    shareMutation.mutate(id)
-  }
-
-  const unshareSession = () => {
-    const id = sessionID()
-    if (!id || unshareMutation.isPending) return
-    if (!shareEnabled()) return
-    unshareMutation.mutate(id)
-  }
-  const copyShareUrl = () => {
-    const url = shareUrl()
-    if (!url) return
-    void navigator.clipboard
-      .writeText(url)
-      .then(() =>
-        showToast({
-          variant: "success",
-          icon: "circle-check",
-          title: language.t("session.share.copy.copied"),
-          description: url,
-        }),
-      )
-      .catch((err: unknown) =>
-        showToast({
-          title: language.t("common.requestFailed"),
-          description: errorMessage(err),
-        }),
-      )
-  }
   const selectShareUrlText: JSX.EventHandler<HTMLDivElement, MouseEvent> = (event) => {
     const selection = window.getSelection()
     if (!selection) return
@@ -742,7 +576,7 @@ export function MessageTimeline(props: {
 
   createEffect(
     on(
-      sessionKey,
+      props.data.sessionKey,
       () =>
         setTitle({
           draft: "",
@@ -755,18 +589,6 @@ export function MessageTimeline(props: {
     ),
   )
 
-  createEffect(
-    on(
-      () => [parentID(), childTaskDescription()] as const,
-      ([id, description]) => {
-        if (!id || description) return
-        if (sync().data.message[id] !== undefined) return
-        void sync().session.sync(id)
-      },
-      { defer: true },
-    ),
-  )
-
   const openTitleEditor = () => {
     if (!sessionID() || parentID()) return
     setTitle({ editing: true, draft: titleLabel() ?? "" })
@@ -778,194 +600,12 @@ export function MessageTimeline(props: {
   }
 
   const closeTitleEditor = () => {
-    if (titleMutation.isPending) return
+    if (props.pending.rename()) return
     setTitle("editing", false)
   }
 
-  const saveTitleEditor = () => {
+  const saveTitleEditor = async () => {
-    const id = sessionID()
+    if (await props.action.rename(title.draft)) setTitle("editing", false)
-    if (!id) return
-    if (titleMutation.isPending) return
-
-    const next = title.draft.trim()
-    if (!next || next === (titleLabel() ?? "")) {
-      setTitle("editing", false)
-      return
-    }
-
-    titleMutation.mutate({ id, title: next })
-  }
-
-  const navigateAfterSessionRemoval = (sessionID: string, parentID?: string, nextSessionID?: string) => {
-    if (params.id !== sessionID) return
-    const href = (id: string) =>
-      params.serverKey ? sessionHref(requireServerKey(params.serverKey), id) : legacySessionHref(sdk().directory, id)
-    if (parentID) {
-      navigate(href(parentID))
-      return
-    }
-    if (nextSessionID) {
-      navigate(href(nextSessionID))
-      return
-    }
-    if (params.serverKey) {
-      tabs.newDraft({ server: requireServerKey(params.serverKey), directory: sdk().directory })
-      return
-    }
-    navigate(`/${params.dir}/session`)
-  }
-
-  const archiveSession = async (sessionID: string) => {
-    const session = sync().session.get(sessionID)
-    if (!session) return
-    if ((await sdk().protocol) !== "v1") return
-
-    const sessions = sync().data.session ?? []
-    const index = sessions.findIndex((s) => s.id === sessionID)
-    const nextSession = index === -1 ? undefined : (sessions[index + 1] ?? sessions[index - 1])
-
-    await sdk()
-      .client.session.update({ sessionID, directory: sdk().directory, time: { archived: Date.now() } })
-      .then(() => {
-        sync().set(
-          produce((draft) => {
-            const index = draft.session.findIndex((s) => s.id === sessionID)
-            if (index !== -1) draft.session.splice(index, 1)
-          }),
-        )
-        sync().session.evict(sessionID)
-        navigateAfterSessionRemoval(sessionID, session.parentID, nextSession?.id)
-        notifySessionTabsRemoved({ directory: sdk().directory, sessionIDs: [sessionID] })
-      })
-      .catch((err) => {
-        showToast({
-          title: language.t("common.requestFailed"),
-          description: errorMessage(err),
-        })
-      })
-  }
-
-  const deleteSession = async (sessionID: string) => {
-    const session = sync().session.get(sessionID)
-    if (!session) return false
-
-    const sessions = (sync().data.session ?? []).filter((s) => !s.parentID && !s.time?.archived)
-    const index = sessions.findIndex((s) => s.id === sessionID)
-    const nextSession = index === -1 ? undefined : (sessions[index + 1] ?? sessions[index - 1])
-
-    const result = await sdk()
-      .api.session.remove({ sessionID })
-      .then(() => true)
-      .catch((err) => {
-        showToast({
-          title: language.t("session.delete.failed.title"),
-          description: errorMessage(err),
-        })
-        return false
-      })
-
-    if (!result) return false
-
-    const removed = new Set<string>([sessionID])
-    const byParent = new Map<string, string[]>()
-    for (const item of sync().data.session) {
-      const parentID = item.parentID
-      if (!parentID) continue
-      const existing = byParent.get(parentID)
-      if (existing) {
-        existing.push(item.id)
-        continue
-      }
-      byParent.set(parentID, [item.id])
-    }
-
-    const stack = [sessionID]
-    while (stack.length) {
-      const parentID = stack.pop()
-      if (!parentID) continue
-
-      const children = byParent.get(parentID)
-      if (!children) continue
-
-      for (const child of children) {
-        if (removed.has(child)) continue
-        removed.add(child)
-        stack.push(child)
-      }
-    }
-
-    navigateAfterSessionRemoval(sessionID, session.parentID, nextSession?.id)
-
-    sync().set(
-      produce((draft) => {
-        draft.session = draft.session.filter((s) => !removed.has(s.id))
-      }),
-    )
-
-    for (const id of removed) {
-      sync().session.evict(id)
-    }
-    notifySessionTabsRemoved({ directory: sdk().directory, sessionIDs: [...removed] })
-    return true
-  }
-
-  const navigateParent = () => {
-    const id = parentID()
-    if (!id) return
-    navigate(
-      params.serverKey ? sessionHref(requireServerKey(params.serverKey), id) : legacySessionHref(sdk().directory, id),
-    )
-  }
-
-  function DialogDeleteSession(props: { sessionID: string }) {
-    const name = createMemo(() => {
-      const session = sync().session.get(props.sessionID)
-      return session ? displayLabel(session) : language.t("command.session.new")
-    })
-    const handleDelete = async () => {
-      await deleteSession(props.sessionID)
-      dialog.close()
-    }
-
-    if (settings.general.newLayoutDesigns())
-      return (
-        <DialogV2 fit>
-          <DialogHeader hideClose>
-            <DialogTitleGroup
-              title={language.t("session.delete.title")}
-              description={language.t("session.delete.confirm", { name: name() })}
-            />
-          </DialogHeader>
-          <DialogFooter>
-            <ButtonV2 variant="ghost" onClick={() => dialog.close()}>
-              {language.t("common.cancel")}
-            </ButtonV2>
-            <ButtonV2 variant="danger" onClick={handleDelete}>
-              {language.t("session.delete.button")}
-            </ButtonV2>
-          </DialogFooter>
-        </DialogV2>
-      )
-
-    return (
-      <Dialog title={language.t("session.delete.title")} fit>
-        <div class="flex flex-col gap-4 pl-6 pr-2.5 pb-3">
-          <div class="flex flex-col gap-1">
-            <span class="text-14-regular text-text-strong">
-              {language.t("session.delete.confirm", { name: name() })}
-            </span>
-          </div>
-          <div class="flex justify-end gap-2">
-            <Button variant="ghost" size="large" onClick={() => dialog.close()}>
-              {language.t("common.cancel")}
-            </Button>
-            <Button variant="primary" size="large" onClick={handleDelete}>
-              {language.t("session.delete.button")}
-            </Button>
-          </div>
-        </div>
-      </Dialog>
-    )
   }
 
   const workingTurn = (userMessageID: string) => sessionStatus().type !== "idle" && activeMessageID() === userMessageID
@@ -1044,7 +684,7 @@ export function MessageTimeline(props: {
     const defaultOpen = createMemo(() => {
       const item = part()
       if (!item) return
-      return partDefaultOpen(item, settings.general.shellToolPartsExpanded(), settings.general.editToolPartsExpanded())
+      return partDefaultOpen(item, props.data.shellToolPartsExpanded(), props.data.editToolPartsExpanded())
     })
 
     return (
@@ -1057,7 +697,7 @@ export function MessageTimeline(props: {
                 message={message()}
                 showAssistantCopyPartID={assistantCopyPartID(row().userMessageID)}
                 turnDurationMs={turnDurationMs(row().userMessageID)}
-                useV2Actions={settings.general.newLayoutDesigns()}
+                useV2Actions={props.data.newLayoutDesigns()}
                 defaultOpen={defaultOpen()}
                 toolOpen={toolOpen[part().id] ?? defaultOpen()}
                 onToolOpenChange={(open) => setToolOpen(part().id, open)}
@@ -1120,8 +760,8 @@ export function MessageTimeline(props: {
                       <div
                         classList={{
                           "shrink-0 max-w-[260px] rounded-[6px] border-border-weak-base bg-background-stronger px-2.5 py-2": true,
-                          "border-[0.5px]": settings.general.newLayoutDesigns(),
+                          "border-[0.5px]": props.data.newLayoutDesigns(),
-                          border: !settings.general.newLayoutDesigns(),
+                          border: !props.data.newLayoutDesigns(),
                         }}
                       >
                         <div class="flex items-center gap-1.5 min-w-0 text-11-medium text-text-strong">
@@ -1156,7 +796,7 @@ export function MessageTimeline(props: {
           if (m?.role === "user") return m
         })
         const messageComments = createMemo(() => {
-          if (!settings.general.newLayoutDesigns()) return []
+          if (!props.data.newLayoutDesigns()) return []
           return getMsgParts(userMessageRow().userMessageID).flatMap((part) => MessageComment.fromPart(part) ?? [])
         })
         return (
@@ -1169,7 +809,7 @@ export function MessageTimeline(props: {
                       message={message()}
                       parts={getMsgParts(userMessageRow().userMessageID)}
                       actions={props.actions}
-                      useV2Actions={settings.general.newLayoutDesigns()}
+                      useV2Actions={props.data.newLayoutDesigns()}
                       comments={messageComments()}
                     />
                   </div>
@@ -1217,7 +857,7 @@ export function MessageTimeline(props: {
             <div data-slot="session-turn-message-container" class="w-full px-4 md:px-5">
               <TimelineThinkingRow
                 reasoningHeading={thinkingRow().reasoningHeading}
-                showReasoningSummaries={settings.general.showReasoningSummaries()}
+                showReasoningSummaries={props.data.showReasoningSummaries()}
               />
             </div>
           </TimelineRowFrame>
@@ -1333,16 +973,16 @@ export function MessageTimeline(props: {
       <div
         class="absolute left-1/2 -translate-x-1/2 z-[60] pointer-events-none transition-all duration-200 ease-out"
         classList={{
-          "bottom-8": settings.general.newLayoutDesigns(),
+          "bottom-8": props.data.newLayoutDesigns(),
-          "bottom-6": !settings.general.newLayoutDesigns(),
+          "bottom-6": !props.data.newLayoutDesigns(),
           "opacity-100 translate-y-0 scale-100": props.scroll.overflow && props.scroll.jump,
           "opacity-0 translate-y-2 pointer-events-none": !props.scroll.overflow || !props.scroll.jump,
-          "scale-[0.8]": (!props.scroll.overflow || !props.scroll.jump) && settings.general.newLayoutDesigns(),
+          "scale-[0.8]": (!props.scroll.overflow || !props.scroll.jump) && props.data.newLayoutDesigns(),
-          "scale-95": (!props.scroll.overflow || !props.scroll.jump) && !settings.general.newLayoutDesigns(),
+          "scale-95": (!props.scroll.overflow || !props.scroll.jump) && !props.data.newLayoutDesigns(),
         }}
       >
         <Show
-          when={settings.general.newLayoutDesigns()}
+          when={props.data.newLayoutDesigns()}
           fallback={
             <button
               type="button"
@@ -1405,22 +1045,22 @@ export function MessageTimeline(props: {
             classList={{
               "sticky top-0 z-30": true,
               "bg-[linear-gradient(to_bottom,var(--v2-background-bg-base)_48px,transparent)]":
-                settings.general.newLayoutDesigns(),
+                props.data.newLayoutDesigns(),
               "bg-[linear-gradient(to_bottom,var(--background-stronger)_48px,transparent)]":
-                !settings.general.newLayoutDesigns(),
+                !props.data.newLayoutDesigns(),
               "w-full": true,
               "pb-4": true,
               "pr-3": true,
-              "pl-2.5": settings.general.newLayoutDesigns(),
+              "pl-2.5": props.data.newLayoutDesigns(),
-              "pl-2 md:pl-4": !settings.general.newLayoutDesigns(),
+              "pl-2 md:pl-4": !props.data.newLayoutDesigns(),
-              "md:max-w-200 md:mx-auto 2xl:max-w-[1000px]": props.centered && !settings.general.newLayoutDesigns(),
+              "md:max-w-200 md:mx-auto 2xl:max-w-[1000px]": props.centered && !props.data.newLayoutDesigns(),
             }}
           >
             <div class="h-12 w-full flex items-center justify-between gap-2">
               <div
                 classList={{
                   "flex items-center gap-1 min-w-0 flex-1": true,
-                  "pr-3": !settings.general.newLayoutDesigns(),
+                  "pr-3": !props.data.newLayoutDesigns(),
                 }}
               >
                 <div class="flex items-center min-w-0 flex-1 w-full">
@@ -1429,7 +1069,7 @@ export function MessageTimeline(props: {
                       type="button"
                       data-slot="session-title-parent"
                       class="min-w-0 max-w-[40%] truncate pl-2 text-[13px] font-[530] leading-4 tracking-[-0.04px] text-v2-text-text-faint transition-colors hover:text-v2-text-text-muted"
-                      onClick={navigateParent}
+                      onClick={props.action.navigateParent}
                     >
                       {parentTitle()}
                     </button>
@@ -1450,8 +1090,8 @@ export function MessageTimeline(props: {
                           classList={{
                             "truncate text-[13px] font-[530] leading-4 tracking-[-0.04px] text-v2-text-text-base": true,
                             "w-fit rounded-[6px] px-2 py-1 hover:bg-v2-overlay-simple-overlay-hover":
-                              settings.general.newLayoutDesigns(),
+                              props.data.newLayoutDesigns(),
-                            "grow-1 min-w-0": !settings.general.newLayoutDesigns(),
+                            "grow-1 min-w-0": !props.data.newLayoutDesigns(),
                           }}
                           onClick={openTitleEditor}
                         >
@@ -1465,15 +1105,14 @@ export function MessageTimeline(props: {
                         }}
                         data-slot="session-title-child"
                         value={title.draft}
-                        disabled={titleMutation.isPending}
+                        disabled={props.pending.rename()}
                         classList={{
                           "block text-[13px] font-[530] leading-4 tracking-[-0.04px] text-v2-text-text-base": true,
-                          "w-full flex-1 grow-1 min-w-0 pl-1 -ml-1 rounded-[6px]": !settings.general.newLayoutDesigns(),
+                          "w-full flex-1 grow-1 min-w-0 pl-1 -ml-1 rounded-[6px]": !props.data.newLayoutDesigns(),
-                          "field-sizing-content self-start rounded-[6px] px-2 py-1 ":
+                          "field-sizing-content self-start rounded-[6px] px-2 py-1 ": props.data.newLayoutDesigns(),
-                            settings.general.newLayoutDesigns(),
                         }}
                         style={{
-                          "--inline-input-shadow": settings.general.newLayoutDesigns()
+                          "--inline-input-shadow": props.data.newLayoutDesigns()
                             ? "none"
                             : "var(--shadow-xs-border-select)",
                         }}
@@ -1501,17 +1140,17 @@ export function MessageTimeline(props: {
                   <div
                     classList={{
                       "shrink-0 flex items-center": true,
-                      "gap-2": settings.general.newLayoutDesigns(),
+                      "gap-2": props.data.newLayoutDesigns(),
-                      "gap-3": !settings.general.newLayoutDesigns(),
+                      "gap-3": !props.data.newLayoutDesigns(),
                     }}
                   >
                     <SessionContextUsage
                       placement="bottom"
-                      buttonAppearance={settings.general.newLayoutDesigns() ? "v2" : "default"}
+                      buttonAppearance={props.data.newLayoutDesigns() ? "v2" : "default"}
                     />
                     <Show when={!parentID()}>
                       <Show
-                        when={settings.general.newLayoutDesigns()}
+                        when={props.data.newLayoutDesigns()}
                         fallback={
                           <DropdownMenu
                             gutter={4}
@@ -1574,13 +1213,11 @@ export function MessageTimeline(props: {
                                     </DropdownMenu.ItemLabel>
                                   </DropdownMenu.Item>
                                 </Show>
-                                <DropdownMenu.Item onSelect={() => void archiveSession(id)}>
+                                <DropdownMenu.Item onSelect={() => void props.action.archive(id)}>
                                   <DropdownMenu.ItemLabel>{language.t("common.archive")}</DropdownMenu.ItemLabel>
                                 </DropdownMenu.Item>
                                 <DropdownMenu.Separator />
-                                <DropdownMenu.Item
+                                <DropdownMenu.Item onSelect={() => props.action.showDelete(id)}>
-                                  onSelect={() => dialog.show(() => <DialogDeleteSession sessionID={id} />)}
-                                >
                                   <DropdownMenu.ItemLabel>{language.t("common.delete")}</DropdownMenu.ItemLabel>
                                 </DropdownMenu.Item>
                               </DropdownMenu.Content>
@@ -1645,11 +1282,11 @@ export function MessageTimeline(props: {
                                   {language.t("session.share.action.share")}...
                                 </MenuV2.Item>
                               </Show>
-                              <MenuV2.Item onSelect={() => void archiveSession(id)}>
+                              <MenuV2.Item onSelect={() => void props.action.archive(id)}>
                                 {language.t("common.archive")}
                               </MenuV2.Item>
                               <MenuV2.Separator />
-                              <MenuV2.Item onSelect={() => dialog.show(() => <DialogDeleteSession sessionID={id} />)}>
+                              <MenuV2.Item onSelect={() => props.action.showDelete(id)}>
                                 {language.t("common.delete")}...
                               </MenuV2.Item>
                             </MenuV2.Content>
@@ -1661,7 +1298,7 @@ export function MessageTimeline(props: {
                         open={share.open}
                         anchorRef={() => more}
                         placement="bottom-end"
-                        gutter={settings.general.newLayoutDesigns() ? 6 : 4}
+                        gutter={props.data.newLayoutDesigns() ? 6 : 4}
                         modal={false}
                         onOpenChange={(open) => {
                           if (open) setShare("dismiss", null)
@@ -1673,7 +1310,7 @@ export function MessageTimeline(props: {
                             data-component="popover-content"
                             classList={{
                               "flex w-80 max-w-none flex-col items-start gap-3 rounded-[10px] border-0 bg-v2-background-bg-layer-01 p-3 shadow-[var(--v2-elevation-floating)]":
-                                settings.general.newLayoutDesigns(),
+                                props.data.newLayoutDesigns(),
                             }}
                             style={{ "min-width": "320px" }}
                             onEscapeKeyDown={(event) => {
@@ -1693,7 +1330,7 @@ export function MessageTimeline(props: {
                             }}
                           >
                             <Show
-                              when={settings.general.newLayoutDesigns()}
+                              when={props.data.newLayoutDesigns()}
                               fallback={
                                 <div class="flex flex-col p-3">
                                   <div class="flex flex-col gap-1">
@@ -1714,10 +1351,10 @@ export function MessageTimeline(props: {
                                           size="large"
                                           variant="primary"
                                           class="w-full"
-                                          onClick={shareSession}
+                                          onClick={() => void props.action.share()}
-                                          disabled={shareMutation.isPending}
+                                          disabled={props.pending.share()}
                                         >
-                                          {shareMutation.isPending
+                                          {props.pending.share()
                                             ? language.t("session.share.action.publishing")
                                             : language.t("session.share.action.publish")}
                                         </Button>
@@ -1737,10 +1374,10 @@ export function MessageTimeline(props: {
                                             size="large"
                                             variant="secondary"
                                             class="w-full shadow-none border border-border-weak-base"
-                                            onClick={unshareSession}
+                                            onClick={() => void props.action.unshare()}
-                                            disabled={unshareMutation.isPending}
+                                            disabled={props.pending.unshare()}
                                           >
-                                            {unshareMutation.isPending
+                                            {props.pending.unshare()
                                               ? language.t("session.share.action.unpublishing")
                                               : language.t("session.share.action.unpublish")}
                                           </Button>
@@ -1748,8 +1385,8 @@ export function MessageTimeline(props: {
                                             size="large"
                                             variant="primary"
                                             class="w-full"
-                                            onClick={viewShare}
+                                            onClick={props.action.viewShare}
-                                            disabled={unshareMutation.isPending}
+                                            disabled={props.pending.unshare()}
                                           >
                                             {language.t("session.share.action.view")}
                                           </Button>
@@ -1777,10 +1414,10 @@ export function MessageTimeline(props: {
                                     <ButtonV2
                                       variant="contrast"
                                       class="w-full"
-                                      onClick={shareSession}
+                                      onClick={() => void props.action.share()}
-                                      disabled={shareMutation.isPending}
+                                      disabled={props.pending.share()}
                                     >
-                                      {shareMutation.isPending
+                                      {props.pending.share()
                                         ? language.t("session.share.action.publishing")
                                         : language.t("session.share.action.publish")}
                                     </ButtonV2>
@@ -1806,7 +1443,7 @@ export function MessageTimeline(props: {
                                         variant="ghost-muted"
                                         icon={<IconV2 name="outline-copy" />}
                                         aria-label={language.t("session.share.copy.copyLink")}
-                                        onClick={copyShareUrl}
+                                        onClick={() => void props.action.copyShareUrl()}
                                       />
                                       <IconButtonV2
                                         type="button"
@@ -1814,18 +1451,18 @@ export function MessageTimeline(props: {
                                         variant="ghost-muted"
                                         icon={<IconV2 name="outline-square-arrow" />}
                                         aria-label={language.t("session.share.action.view")}
-                                        onClick={viewShare}
+                                        onClick={props.action.viewShare}
-                                        disabled={unshareMutation.isPending}
+                                        disabled={props.pending.unshare()}
                                       />
                                     </div>
                                     <div class="flex w-full">
                                       <ButtonV2
                                         variant="outline"
                                         class="w-full"
-                                        onClick={unshareSession}
+                                        onClick={() => void props.action.unshare()}
-                                        disabled={unshareMutation.isPending}
+                                        disabled={props.pending.unshare()}
                                       >
-                                        {unshareMutation.isPending
+                                        {props.pending.unshare()
                                           ? language.t("session.share.action.unpublishing")
                                           : language.t("session.share.action.unpublish")}
                                       </ButtonV2>

packages/app/src/pages/session/timeline/model.ts

@@ -1,23 +1,24 @@
-import type { Message, UserMessage } from "@opencode-ai/sdk/v2"
+import type { Message } from "@opencode-ai/sdk/v2"
 import { createMemo, createResource, onCleanup, untrack, type Accessor } from "solid-js"
 import { useServerSync } from "@/context/server-sync"
 import { useSync } from "@/context/sync"
-import { same } from "@/utils/same"
+import type { SessionController } from "../session-controller"
+
+export {
+  selectSessionUserMessages as selectUserMessages,
+  selectVisibleSessionUserMessages as selectVisibleUserMessages,
+} from "../session-domain"
 
-const emptyUserMessages: UserMessage[] = []
 const sessionFreshness = 15_000
 
-export function createTimelineModel(input: {
+export function createTimelineModel(input: { session: Pick<SessionController, "identity" | "history"> }) {
-  sessionID: Accessor<string | undefined>
-  revertMessageID: Accessor<string | undefined>
-}) {
   const serverSync = useServerSync()
   const sync = useSync()
   let refreshFrame: number | undefined
   let refreshTimer: number | undefined
 
   const [resource] = createResource(
-    () => input.sessionID(),
+    () => input.session.identity.sessionID(),
     (id) => {
       clearRefresh()
       if (!id) return
@@ -29,7 +30,7 @@ export function createTimelineModel(input: {
         refreshFrame = undefined
         refreshTimer = window.setTimeout(() => {
           refreshTimer = undefined
-          if (input.sessionID() !== id) return
+          if (input.session.identity.sessionID() !== id) return
           untrack(() => {
             if (stale) void sync().session.sync(id, { force: true })
           })
@@ -39,33 +40,21 @@ export function createTimelineModel(input: {
       return sync().session.sync(id)
     },
   )
-  const messages = createMemo(() => {
-    const id = input.sessionID()
-    return id ? (sync().data.message[id] ?? []) : []
-  })
   const ready = createMemo(() => {
-    const id = input.sessionID()
+    const id = input.session.identity.sessionID()
     return !id || isTimelineReady(sync().data.message[id], serverSync().session.history.loading(id))
   })
-  const userMessages = createMemo(() => selectUserMessages(messages()), emptyUserMessages, { equals: same })
-  const visibleUserMessages = createMemo(
-    () => {
-      return selectVisibleUserMessages(userMessages(), input.revertMessageID())
-    },
-    emptyUserMessages,
-    { equals: same },
-  )
   const more = createMemo(() => {
-    const id = input.sessionID()
+    const id = input.session.identity.sessionID()
     return id ? sync().session.history.more(id) : false
   })
   const loading = createMemo(() => {
-    const id = input.sessionID()
+    const id = input.session.identity.sessionID()
     return id ? sync().session.history.loading(id) : false
   })
   const loadOlder = async (options?: { before?: () => void; after?: (done: boolean) => void }) => {
     return loadOlderTimeline({
-      sessionID: input.sessionID,
+      sessionID: input.session.identity.sessionID,
       more,
       loading,
       loadMore: (sessionID) => sync().session.history.loadMore(sessionID),
@@ -78,12 +67,12 @@ export function createTimelineModel(input: {
 
   return {
     history: { loadOlder, loading, more },
-    lastUserMessage: createMemo(() => visibleUserMessages().at(-1)),
+    lastUserMessage: input.session.history.lastUserMessage,
-    messages,
+    messages: input.session.history.messages,
     ready,
     resource,
-    userMessages,
+    userMessages: input.session.history.userMessages,
-    visibleUserMessages,
+    visibleUserMessages: input.session.history.visibleUserMessages,
   }
 
   function clearRefresh() {
@@ -94,19 +83,10 @@ export function createTimelineModel(input: {
   }
 }
 
-export function selectUserMessages(messages: Message[]) {
-  return messages.filter((message): message is UserMessage => message.role === "user")
-}
-
 export function isTimelineReady(messages: Message[] | undefined, loading: boolean) {
   return messages !== undefined && (messages.some((message) => message.role === "user") || !loading)
 }
 
-export function selectVisibleUserMessages(messages: UserMessage[], revertMessageID?: string) {
-  if (!revertMessageID) return messages
-  return messages.filter((message) => message.id < revertMessageID)
-}
-
 export async function loadOlderTimeline(input: {
   sessionID: Accessor<string | undefined>
   more: Accessor<boolean>

packages/app/src/pages/session/timeline/observe-element-offset.test.ts

@@ -38,19 +38,17 @@ test("reports a divergent native offset once and ignores equal offsets and unrel
     instance.scrollOffset = offset
   })
 
+  document.body.append(unrelated)
+  unrelated.remove()
+  await frames(2)
+  expect(calls).toEqual([])
+
   route.remove()
   document.body.append(route)
   await new Promise((resolve) => setTimeout(resolve, 0))
   await frames(3)
   expect(calls).toEqual([[0, false]])
 
-  instance.scrollOffset = 79_400
-  document.body.append(unrelated)
-  unrelated.remove()
-  await frames(2)
-  expect(calls).toEqual([[0, false]])
-
-  instance.scrollOffset = 0
   route.remove()
   document.body.append(route)
   await new Promise((resolve) => setTimeout(resolve, 0))

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

@@ -13,19 +13,25 @@ import { useSync } from "@/context/sync"
 import { useTerminal } from "@/context/terminal"
 import { showToast } from "@/utils/toast"
 import { findLast } from "@opencode-ai/core/util/array"
-import { createSessionTabs } from "@/pages/session/helpers"
 import { extractPromptFromParts } from "@/utils/prompt"
-import { UserMessage } from "@opencode-ai/sdk/v2"
+import type { UserMessage } from "@opencode-ai/sdk/v2"
-import { useSessionLayout } from "@/pages/session/session-layout"
-import { createSessionOwnership } from "./session-ownership"
 import { useLocal } from "@/context/local"
+import type { SessionController } from "./session-controller"
+
+type SessionCommandSource = {
+  identity: SessionController["identity"]
+  data: Pick<SessionController["data"], "info" | "revertMessageID">
+  history: Pick<SessionController["history"], "userMessages" | "visibleUserMessages">
+  layout: SessionController["layout"]
+  ownership: SessionController["ownership"]
+  tabs: Pick<SessionController["tabs"], "activeFileTab" | "closableTab">
+}
 
 export type SessionCommandContext = {
+  session: SessionCommandSource
   navigateMessageByOffset: (offset: number) => void
   setActiveMessage: (message: UserMessage | undefined) => void
   focusInput: () => void
-  review?: () => boolean
-  fileBrowser?: () => boolean
 }
 
 const withCategory = (category: string) => {
@@ -49,15 +55,17 @@ export const useSessionCommands = (actions: SessionCommandContext) => {
   const layout = useLayout()
   const local = useLocal()
   const navigate = useNavigate()
-  const { params, sessionKey, tabs, view } = useSessionLayout()
+  const params = actions.session.identity.params
-  const sessionOwnership = createSessionOwnership(sessionKey)
+  const tabs = actions.session.layout.tabs
+  const view = actions.session.layout.view
+  const sessionOwnership = actions.session.ownership
   const openDialog = async <T,>(load: () => Promise<T>, show: (value: T) => void) => {
     const owner = sessionOwnership.capture()
     const value = await load()
     owner.run(() => show(value))
   }
   const runCommand = async <T,>(input: {
-    owner: ReturnType<ReturnType<typeof createSessionOwnership>["capture"]>
+    owner: ReturnType<SessionController["ownership"]["capture"]>
     prompt: T
     request: () => Promise<unknown>
     updatePrompt: (prompt: T) => void
@@ -68,39 +76,13 @@ export const useSessionCommands = (actions: SessionCommandContext) => {
     input.owner.run(input.updateViewport)
   }
 
-  const info = () => {
+  const info = actions.session.data.info
-    const id = params.id
+  const activeFileTab = actions.session.tabs.activeFileTab
-    if (!id) return
+  const closableTab = actions.session.tabs.closableTab
-    return sync().session.get(id)
-  }
-  const hasReview = () => !!params.id
-  const normalizeTab = (tab: string) => {
-    if (!tab.startsWith("file://")) return tab
-    return file.tab(tab)
-  }
-  const tabState = createSessionTabs({
-    tabs,
-    pathFromTab: file.pathFromTab,
-    normalizeTab,
-    review: actions.review,
-    hasReview,
-    fileBrowser: actions.fileBrowser,
-  })
-  const activeFileTab = tabState.activeFileTab
-  const closableTab = tabState.closableTab
   const shown = settings.visibility.fileTree
 
-  const messages = () => {
+  const userMessages = actions.session.history.userMessages
-    const id = params.id
+  const visibleUserMessages = actions.session.history.visibleUserMessages
-    if (!id) return []
-    return sync().data.message[id] ?? []
-  }
-  const userMessages = () => messages().filter((m) => m.role === "user") as UserMessage[]
-  const visibleUserMessages = () => {
-    const revert = info()?.revert?.messageID
-    if (!revert) return userMessages()
-    return userMessages().filter((m) => m.id < revert)
-  }
 
   const showAllFiles = () => {
     if (layout.fileTree.tab() !== "changes") return
@@ -309,7 +291,7 @@ export const useSessionCommands = (actions: SessionCommandContext) => {
     const session = sdk().api.session
     const directory = sdk().directory
     const promptSession = prompt.capture()
-    const revert = info()?.revert?.messageID
+    const revert = actions.session.data.revertMessageID()
     const messages = userMessages()
     const message = findLast(messages, (x) => !revert || x.id < revert)
     if (!message) return
@@ -338,7 +320,7 @@ export const useSessionCommands = (actions: SessionCommandContext) => {
     const messages = userMessages()
     const promptSession = prompt.capture()
 
-    const revertMessageID = info()?.revert?.messageID
+    const revertMessageID = actions.session.data.revertMessageID()
     if (!revertMessageID) return
 
     const next = messages.find((x) => x.id > revertMessageID)

packages/app/src/utils/server-compat.test.ts

@@ -26,6 +26,7 @@ function setup(
         return new Response(undefined, { status: 204 })
       if (request.method === "POST" && request.url.endsWith("/prompt")) {
         return Response.json({
+          admittedSeq: 1,
           id: "msg_1",
           sessionID: "ses_1",
           timeCreated: 1,

packages/app/src/utils/server-compat.ts

@@ -128,7 +128,7 @@ function createV1Api(input: CompatibleInput): CompatibleApi {
   const located = <T>(data: T, value?: { directory?: string }) => ({
     location: {
       directory: directory(value) ?? "",
-      project: { id: "", directory: directory(value) ?? "", canonical: directory(value) ?? "" },
+      project: { id: "", directory: directory(value) ?? "" },
     },
     data,
   })
@@ -229,6 +229,7 @@ function createV1Api(input: CompatibleInput): CompatibleApi {
           ],
         })
         return {
+          admittedSeq: 0,
           id: value.id ?? "",
           sessionID: value.sessionID,
           timeCreated: Date.now(),
@@ -254,6 +255,7 @@ function createV1Api(input: CompatibleInput): CompatibleApi {
           })),
         })
         return {
+          admittedSeq: 0,
           id: value.id ?? "",
           sessionID: value.sessionID,
           timeCreated: Date.now(),
@@ -278,6 +280,7 @@ function createV1Api(input: CompatibleInput): CompatibleApi {
           modelID: value.model.modelID,
         })
         return {
+          admittedSeq: 0,
           id: value.id ?? "",
           sessionID: value.sessionID,
           timeCreated: Date.now(),
@@ -298,19 +301,12 @@ function createV1Api(input: CompatibleInput): CompatibleApi {
     project: {
       ...input.current.project,
       async list() {
-        return ((await legacy().project.list()).data ?? []).map((project) => ({
+        return ((await legacy().project.list()).data ?? []) as Project[]
-          ...project,
-          canonical: project.worktree,
-        }))
       },
       async current(value?: Parameters<ServerApi["project"]["current"]>[0]) {
         const result = await legacy(value?.location).project.current()
         if (!result.data) throw new Error("Project not found")
-        return {
+        return { id: result.data.id, directory: result.data.worktree } satisfies ProjectCurrent
-          id: result.data.id,
-          directory: result.data.worktree,
-          canonical: result.data.worktree,
-        } satisfies ProjectCurrent
       },
       // async update(value: Parameters<ServerApi["project"]["update"]>[0]) {
       //   const project = (await legacy().project.list()).data?.find((item) => item.id === value.projectID)

packages/app/src/utils/session-title.ts

@@ -0,0 +1,7 @@
+const pattern = /^(New session|Child session) - \d{4}-\d{2}-\d{2}T\d{2}:\d{2}:\d{2}\.\d{3}Z$/
+
+export function sessionTitle(title?: string) {
+  if (!title) return title
+  const match = title.match(pattern)
+  return match?.[1] ?? title
+}

packages/app/src/utils/session.ts

@@ -1,6 +1,5 @@
 import type { SessionApi, SessionInfo, SessionListInput } from "@opencode-ai/client/promise"
 import type { Session } from "@opencode-ai/sdk/v2/client"
-import { withTimestampedFallback } from "@opencode-ai/util/session-title-fallback"
 
 export function normalizeSessionInfo(input: SessionInfo | Session): Session {
   if (!("location" in input)) return input
@@ -14,7 +13,7 @@ export function normalizeSessionInfo(input: SessionInfo | Session): Session {
     parentID: input.parentID,
     cost: input.cost,
     tokens: input.tokens,
-    title: withTimestampedFallback(input),
+    title: input.title,
     agent: input.agent,
     model: input.model,
     version: "",

packages/cli/package.json

@@ -40,9 +40,6 @@
     "open": "10.1.2",
     "semver": "catalog:",
     "solid-js": "catalog:",
-    "tree-sitter-bash": "0.25.0",
-    "tree-sitter-powershell": "0.25.10",
-    "web-tree-sitter": "0.25.10",
     "uqr": "0.1.3",
     "ws": "8.21.0"
   },

packages/cli/script/build-node.ts

@@ -62,8 +62,8 @@ for (const target of targets) {
   await rm("dist-node", { recursive: true, force: true })
   const assetHash = await hashNodeAssets(assets)
   const input = { version: Script.version, channel: Script.channel, models: modelsData, assetHash, target }
-  await copyNodeAssets(assets)
   await build(mainConfig(input))
+  await copyNodeAssets(assets)
 
   const host = target.platform === process.platform && target.arch === process.arch
   if (host) {

packages/cli/script/node-assets.ts

@@ -3,7 +3,7 @@ import { copyFile, mkdir, readdir, readFile, stat } from "node:fs/promises"
 import path from "node:path"
 import { fileURLToPath } from "node:url"
 import { getNodeAssets } from "@opentui/core/node-assets"
-import { attentionSoundAssets, type NodeTarget, photonWasmAsset, shellParserWasmAssets } from "../src/node/target"
+import { attentionSoundAssets, type NodeTarget, photonWasmAsset } from "../src/node/target"
 
 const dir = path.resolve(import.meta.dirname, "..")
 
@@ -43,10 +43,6 @@ export async function collectNodeAssets(target: NodeTarget) {
       key: photonWasmAsset,
       source: fileURLToPath(import.meta.resolve(photonWasmAsset)),
     },
-    ...Object.values(shellParserWasmAssets).map((key) => ({
-      key,
-      source: fileURLToPath(import.meta.resolve(key)),
-    })),
     ...attentionSoundAssets.map((key) => ({
       key,
       source: path.resolve(dir, "../ui/src/assets/audio", path.basename(key)),
@@ -58,9 +54,8 @@ export async function collectNodeAssets(target: NodeTarget) {
         source: path.join(ptyRoot, relative),
       })),
   ]
-  const unique = [...new Map(assets.map((asset) => [asset.key, asset])).values()]
+  await Promise.all(assets.map((asset) => stat(asset.source)))
-  await Promise.all(unique.map((asset) => stat(asset.source)))
+  return assets
-  return unique
 }
 
 export async function hashNodeAssets(assets: readonly NodeAsset[]) {

packages/cli/src/acp/service.ts

@@ -8,7 +8,6 @@ import {
   type SessionMessageInfo,
   type SkillInfo,
 } from "@opencode-ai/client/promise"
-import { withTimestampedFallback } from "@opencode-ai/util/session-title-fallback"
 import type {
   AgentSideConnection,
   AuthenticateRequest,
@@ -214,7 +213,7 @@ export function make(input: { readonly client: OpenCodeClient; readonly connecti
         sessions: page.data.map((session) => ({
           sessionId: session.id,
           cwd: session.location.directory,
-          title: withTimestampedFallback(session),
+          title: session.title,
           updatedAt: new Date(session.time.updated).toISOString(),
         })),
         ...(page.cursor.next ? { nextCursor: page.cursor.next } : {}),
@@ -245,10 +244,7 @@ export function make(input: { readonly client: OpenCodeClient; readonly connecti
       return {}
     },
     forkSession: async (params) => {
-      const forked = await input.client.session.fork({
+      const forked = await input.client.session.fork({ sessionID: params.sessionId })
-        sessionID: params.sessionId,
-        boundary: { type: "through" },
-      })
       const state = await attach(forked, forked.location.directory, params.mcpServers ?? [])
       await replay(state)
       return { sessionId: state.id, configOptions: configOptions(state) }