fix(app): update remaining v2 e2e fixtures

# Conflicts:
#	packages/app/src/utils/server-compat.test.ts
#	packages/app/src/utils/server-compat.ts

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

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

bun.lock

@@ -64,7 +64,6 @@
         "@opencode-ai/client": "workspace:*",
         "@opencode-ai/core": "workspace:*",
         "@opencode-ai/schema": "workspace:*",
-        "@opencode-ai/sdk": "file:vendor/opencode-ai-sdk-1.18.8-dev.tgz",
         "@opencode-ai/session-ui": "workspace:*",
         "@opencode-ai/ui": "workspace:*",
         "@pierre/trees": "1.0.0-beta.4",
@@ -402,8 +401,11 @@
         "immer": "11.1.4",
         "jsonc-parser": "3.3.1",
         "semver": "^7.6.3",
+        "tree-sitter-bash": "0.25.0",
+        "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:",
       },
@@ -600,6 +602,7 @@
         "zod": "catalog:",
       },
       "devDependencies": {
+        "@opencode-ai/theme": "workspace:*",
         "@opentui/core": "catalog:",
         "@opentui/keymap": "catalog:",
         "@opentui/solid": "catalog:",
@@ -611,12 +614,14 @@
         "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",
@@ -1070,9 +1075,11 @@
   },
   "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",
@@ -4069,10 +4076,10 @@
 
     "fetch-blob": ["fetch-blob@3.2.0", "", { "dependencies": { "node-domexception": "^1.0.0", "web-streams-polyfill": "^3.0.3" } }, "sha512-7yAQpD2UMJzLi1Dqv7qFYnPbaPx7ZfFK6PiIxQ4PfkGPyNyl2Ugx+a/umUonmKqjhM4DnfbMvdX6otXq83soQQ=="],
 
-    "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=="],
 
+    "file-uri-to-path": ["file-uri-to-path@1.0.0", "", {}, "sha512-0Zt+s3L7Vf1biwWZ29aARiVYLx7iMGnEUl9x33fbB/j3jR81u/O2LbqK+Bm1CDSNDKVtJ/YjwY7TUd5SkeLQLw=="],
+
     "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=="],
@@ -5631,6 +5638,10 @@
 
     "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=="],
@@ -6327,8 +6338,6 @@
 
     "@openauthjs/openauth/jose": ["jose@5.9.6", "", {}, "sha512-AMlnetc9+CV9asI19zHmrgS/WYsWUwCn2R7RzlbJWD7F9eWYUTGyBmU9o6PxngtLGOiDGPRu+Uc4fhKzbpteZQ=="],
 
-    "@opencode-ai/app/@opencode-ai/sdk": ["@opencode-ai/sdk@vendor/opencode-ai-sdk-1.18.8-dev.tgz", { "dependencies": { "cross-spawn": "7.0.6" } }, "sha512-C2nfk4x0sPINwE5V6DPkFSuH3PkUmKPWHPzxpXC1j+3Ui5hslLCWJbkk8WcOG1Lyt3C0+yp4ea64v/kmtYCO4w=="],
-
     "@opencode-ai/cli/semver": ["semver@7.7.4", "", { "bin": { "semver": "bin/semver.js" } }, "sha512-vFKC2IEtQnVhpT78h1Yp8wzwrf8CM+MzKMHGJZfBtzhZNycRFnXsHk6E5TxIkkMsgNS7mdX3AGB7x2QM2di4lA=="],
 
     "@opencode-ai/console-app/@smithy/eventstream-codec": ["@smithy/eventstream-codec@4.2.7", "", { "dependencies": { "@aws-crypto/crc32": "5.2.0", "@smithy/types": "^4.11.0", "@smithy/util-hex-encoding": "^4.2.0", "tslib": "^2.6.2" } }, "sha512-DrpkEoM3j9cBBWhufqBwnbbn+3nf1N9FP6xuVJ+e220jbactKuQgaZwjwP5CP1t+O94brm2JgVMD2atMGX3xIQ=="],
@@ -6853,6 +6862,8 @@
 
     "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

@@ -0,0 +1,118 @@
+# 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

@@ -46,7 +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.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.
+Use `LLMClient.stream(request)` when callers want incremental `LLMEvent`s. Use `LLMClient.generate(request)` when callers want those same events collected into an `LLMResponse`.
 
 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.
 
@@ -138,13 +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, PreparedRequest, LLMResponse
+    events.ts               Usage, individual events, LLMEvent, 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.prepare/stream/generate
+    client.ts               Route.make + LLMClient.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

@@ -196,7 +196,6 @@ 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.prepare` / `stream` / `generate` to read
+- [x] Rework `LLMClient.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

@@ -50,18 +50,6 @@ 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* () {
@@ -222,33 +210,15 @@ 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, prepare, stream,
-// or tool-loop behavior without spending tokens on every example.
+// enabled at a time so the tutorial can demonstrate generate, stream, 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/llm.ts

@@ -9,25 +9,28 @@ 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 = Omit<
+export type RequestInput<SelectedModel extends Model = Model> = Omit<
   ConstructorParameters<typeof LLMRequest>[0],
-  "system" | "messages" | "tools" | "toolChoice" | "generation" | "http" | "providerOptions"
+  "model" | "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?: ConstructorParameters<typeof LLMRequest>[0]["providerOptions"]
+  readonly providerOptions?: NoInfer<ModelProviderOptions<SelectedModel>>
   readonly http?: HttpOptions.Input
 }
 
@@ -35,7 +38,7 @@ export const generate = LLMClient.generate
 
 export const stream = LLMClient.stream
 
-export const request = (input: RequestInput) => {
+export const request = <const SelectedModel extends Model>(input: RequestInput<SelectedModel>) => {
   const {
     system: requestSystem,
     prompt,
@@ -63,7 +66,7 @@ const GENERATE_OBJECT_TOOL_NAME = "generate_object"
 
 const GENERATE_OBJECT_TOOL_DESCRIPTION = "Return the structured result by calling this tool."
 
-type GenerateObjectBase = Omit<RequestInput, "tools" | "toolChoice">
+type GenerateObjectBase<SelectedModel extends Model = Model> = Omit<RequestInput<SelectedModel>, "tools" | "toolChoice">
 
 export class GenerateObjectResponse<T> {
   constructor(
@@ -80,11 +83,13 @@ export class GenerateObjectResponse<T> {
   }
 }
 
-export interface GenerateObjectOptions<S extends ToolSchema<any>> extends GenerateObjectBase {
+export interface GenerateObjectOptions<S extends ToolSchema<any>, SelectedModel extends Model = Model>
+  extends GenerateObjectBase<SelectedModel> {
   readonly schema: S
 }
 
-export interface GenerateObjectDynamicOptions extends GenerateObjectBase {
+export interface GenerateObjectDynamicOptions<SelectedModel extends Model = Model>
+  extends GenerateObjectBase<SelectedModel> {
   /** Raw JSON Schema object describing the expected output shape. */
   readonly jsonSchema: JsonSchema.JsonSchema
 }
@@ -137,11 +142,11 @@ 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<S extends ToolSchema<any>>(
-  options: GenerateObjectOptions<S>,
+export function generateObject<const SelectedModel extends Model, S extends ToolSchema<any>>(
+  options: GenerateObjectOptions<S, SelectedModel>,
 ): Effect.Effect<GenerateObjectResponse<Schema.Schema.Type<S>>, LLMError>
-export function generateObject(
-  options: GenerateObjectDynamicOptions,
+export function generateObject<const SelectedModel extends Model>(
+  options: GenerateObjectDynamicOptions<SelectedModel>,
 ): Effect.Effect<GenerateObjectResponse<unknown>, LLMError>
 export function generateObject(options: GenerateObjectOptions<ToolSchema<any>> | GenerateObjectDynamicOptions) {
   if ("schema" in options) {

packages/ai/src/provider-package.ts

@@ -1,4 +1,4 @@
-import type { Model } from "./schema"
+import type { Model, ProviderOptions } from "./schema"
 
 export interface Settings extends Readonly<Record<string, unknown>> {
   readonly headers?: Readonly<Record<string, string>>
@@ -9,8 +9,11 @@ export interface Settings extends Readonly<Record<string, unknown>> {
   }
 }
 
-export interface Definition<ProviderSettings extends Settings = Settings> {
-  readonly model: (modelID: string, settings: ProviderSettings) => Model
+export interface Definition<
+  ProviderSettings extends Settings = Settings,
+  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({ id: modelID }),
+    model: (modelID: string | ModelID) => route.model<AnthropicMessages.ProviderOptionsInput>({ id: modelID }),
     configure,
   }
 }
@@ -57,7 +57,10 @@ export const provider = {
   configure,
 }
 
-export const model: ProviderPackage.Definition<Settings>["model"] = (modelID, settings) => {
+export const model: ProviderPackage.Definition<Settings, AnthropicMessages.ProviderOptionsInput>["model"] = (
+  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

@@ -52,7 +52,10 @@ export const configure = (input: Config = {}) => {
 }
 
 export const provider = configure()
-export const model: ProviderPackage.Definition<Settings>["model"] = (modelID, settings) => {
+export const model: ProviderPackage.Definition<Settings, AnthropicMessages.ProviderOptionsInput>["model"] = (
+  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

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

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

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

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

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

@@ -1,8 +1,9 @@
 import type { ProviderPackage } from "../provider-package"
 import { OpenAICompatibleChat } from "../protocols/openai-compatible-chat"
 import type { RouteDefaultsInput } from "../route/client"
-import { ProviderID, type ModelID, type ProviderOptions } from "../schema"
+import { ProviderID, type ModelID } from "../schema"
 import { GoogleVertexShared } from "./google-vertex-shared"
+import type { OpenAIProviderOptionsInput } from "./openai-options"
 
 export const id = ProviderID.make("google-vertex")
 
@@ -11,6 +12,7 @@ export type Config = RouteDefaultsInput &
     readonly baseURL?: string
     readonly location?: string
     readonly project?: string
+    readonly providerOptions?: OpenAIProviderOptionsInput
   }
 
 export interface Settings extends ProviderPackage.Settings {
@@ -19,7 +21,7 @@ export interface Settings extends ProviderPackage.Settings {
   readonly baseURL?: string
   readonly location?: string
   readonly project?: string
-  readonly providerOptions?: ProviderOptions
+  readonly providerOptions?: OpenAIProviderOptionsInput
 }
 
 const route = OpenAICompatibleChat.route.with({
@@ -56,7 +58,7 @@ export const configure = (input: Config = {}) => {
   const route = configuredRoute(input)
   return {
     id,
-    model: (modelID: string | ModelID) => route.model({ id: modelID }),
+    model: (modelID: string | ModelID) => route.model<OpenAIProviderOptionsInput>({ id: modelID }),
     configure,
   }
 }
@@ -66,7 +68,7 @@ export const provider = {
   configure,
 }
 
-export const model: ProviderPackage.Definition<Settings>["model"] = (modelID, settings) => {
+export const model: ProviderPackage.Definition<Settings, OpenAIProviderOptionsInput>["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({ id: modelID }),
+    model: (modelID: string | ModelID) => route.model<AnthropicMessages.ProviderOptionsInput>({ id: modelID }),
     configure,
   }
 }
@@ -101,7 +101,10 @@ export const provider = {
   configure,
 }
 
-export const model: ProviderPackage.Definition<Settings>["model"] = (modelID, settings) => {
+export const model: ProviderPackage.Definition<Settings, AnthropicMessages.ProviderOptionsInput>["model"] = (
+  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

@@ -1,8 +1,9 @@
 import type { ProviderPackage } from "../provider-package"
 import { OpenAICompatibleResponses } from "../protocols/openai-compatible-responses"
 import type { RouteDefaultsInput } from "../route/client"
-import { ProviderID, type ModelID, type ProviderOptions } from "../schema"
+import { ProviderID, type ModelID } from "../schema"
 import { GoogleVertexShared } from "./google-vertex-shared"
+import type { OpenResponsesProviderOptionsInput } from "./open-responses-options"
 
 export const id = ProviderID.make("google-vertex")
 
@@ -11,6 +12,7 @@ export type Config = RouteDefaultsInput &
     readonly baseURL?: string
     readonly location?: string
     readonly project?: string
+    readonly providerOptions?: OpenResponsesProviderOptionsInput
   }
 
 export interface Settings extends ProviderPackage.Settings {
@@ -19,7 +21,7 @@ export interface Settings extends ProviderPackage.Settings {
   readonly baseURL?: string
   readonly location?: string
   readonly project?: string
-  readonly providerOptions?: ProviderOptions
+  readonly providerOptions?: OpenResponsesProviderOptionsInput
 }
 
 const route = OpenAICompatibleResponses.route.with({
@@ -58,7 +60,7 @@ export const configure = (input: Config = {}) => {
   const route = configuredRoute(input)
   return {
     id,
-    model: (modelID: string | ModelID) => route.model({ id: modelID }),
+    model: (modelID: string | ModelID) => route.model<OpenResponsesProviderOptionsInput>({ id: modelID }),
     configure,
   }
 }
@@ -68,7 +70,10 @@ export const provider = {
   configure,
 }
 
-export const model: ProviderPackage.Definition<Settings>["model"] = (modelID, settings) => {
+export const model: ProviderPackage.Definition<Settings, OpenResponsesProviderOptionsInput>["model"] = (
+  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

@@ -77,7 +77,8 @@ const configuredRoute = (input: Config, modelID: string | ModelID) => {
 export const configure = (input: Config = {}) => {
   return {
     id,
-    model: (modelID: string | ModelID) => configuredRoute(input, modelID).model({ id: modelID }),
+    model: (modelID: string | ModelID) =>
+      configuredRoute(input, modelID).model<Gemini.ProviderOptionsInput>({ id: modelID }),
     configure,
   }
 }
@@ -86,7 +87,10 @@ export const provider = {
   id,
   configure,
 }
-export const model: ProviderPackage.Definition<Settings>["model"] = (modelID, settings) => {
+export const model: ProviderPackage.Definition<Settings, Gemini.ProviderOptionsInput>["model"] = (
+  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({ id: modelID }),
+    model: (modelID: string | ModelID) => route.model<Gemini.ProviderOptionsInput>({ id: modelID }),
     image,
     configure,
   }
 }
 
 export const provider = configure()
-export const model: ProviderPackage.Definition<Settings>["model"] = (modelID, settings) =>
+export const model: ProviderPackage.Definition<Settings, Gemini.ProviderOptionsInput>["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({ id: modelID }),
+    model: (modelID: string | ModelID) => route.model<OpenResponsesProviderOptionsInput>({ id: modelID }),
     configure,
   }
 }
@@ -46,7 +46,10 @@ export const provider = {
   configure,
 }
 
-export const model: ProviderPackage.Definition<Settings>["model"] = (modelID, settings) =>
+export const model: ProviderPackage.Definition<Settings, OpenResponsesProviderOptionsInput>["model"] = (
+  modelID,
+  settings,
+) =>
   configure({
     apiKey: settings.apiKey,
     baseURL: settings.baseURL,

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

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

packages/ai/src/providers/openai.ts

@@ -86,10 +86,15 @@ export const configure = (input: Config = {}) => {
   const chatRoute = configuredRoute(OpenAIChat.route, input)
   const modelDefaults = defaults(input)
   const responses = (id: string | ModelID) =>
-    responsesRoute.with(withOpenAIOptions(id, modelDefaults, { textVerbosity: true })).model({ id })
+    responsesRoute
+      .with(withOpenAIOptions(id, modelDefaults, { textVerbosity: true }))
+      .model<OpenAIProviderOptionsInput>({ id })
   const responsesWebSocket = (id: string | ModelID) =>
-    responsesWebSocketRoute.with(withOpenAIOptions(id, modelDefaults, { textVerbosity: true })).model({ id })
-  const chat = (id: string | ModelID) => chatRoute.with(withOpenAIOptions(id, modelDefaults)).model({ id })
+    responsesWebSocketRoute
+      .with(withOpenAIOptions(id, modelDefaults, { textVerbosity: true }))
+      .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,
@@ -132,15 +137,17 @@ const config = (settings: Settings): Config => {
   }
 }
 
-export const model: ProviderPackage.Definition<Settings>["model"] = (modelID, settings) => {
+export const model: ProviderPackage.Definition<Settings, OpenAIProviderOptionsInput>["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>["model"] = (modelID, settings) =>
-  configure(config(settings)).chat(modelID)
+export const chatModel: ProviderPackage.Definition<Settings, OpenAIProviderOptionsInput>["model"] = (
+  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

@@ -5,6 +5,7 @@ import { Framing } from "../route/framing"
 import { Protocol } from "../route/protocol"
 import { AuthOptions, type ProviderAuthOption } from "../route/auth-options"
 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 { isRecord } from "../protocols/shared"
@@ -30,6 +31,12 @@ 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),
 ])
@@ -107,10 +114,21 @@ export const configure = (input: ModelOptions = {}) => {
   const route = configuredRoute(input)
   return {
     id,
-    model: (modelID: string | ModelID) => route.model({ id: modelID }),
+    model: (modelID: string | ModelID) => route.model<OpenRouterProviderOptionsInput>({ id: modelID }),
     configure,
   }
 }
 
 export const provider = configure()
-export const model = provider.model
+export const model: ProviderPackage.Definition<Settings, OpenRouterProviderOptionsInput>["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

@@ -5,14 +5,23 @@ import * as OpenAICompatibleProfiles from "./openai-compatible-profile"
 import * as OpenAICompatibleChat from "../protocols/openai-compatible-chat"
 import * as OpenAIResponses from "../protocols/openai-responses"
 import { XAIImages } from "../protocols/xai-images"
+import type { OpenAIProviderOptionsInput } from "./openai-options"
+import type { ProviderPackage } from "../provider-package"
 
 export const id = ProviderID.make("xai")
 
-export type ModelOptions = RouteDefaultsInput &
+export type ModelOptions = Omit<RouteDefaultsInput, "providerOptions"> &
   ProviderAuthOption<"optional"> & {
     readonly baseURL?: string
+    readonly providerOptions?: OpenAIProviderOptionsInput
   }
 
+export interface Settings extends ProviderPackage.Settings {
+  readonly apiKey?: string
+  readonly baseURL?: string
+  readonly providerOptions?: OpenAIProviderOptionsInput
+}
+
 export type { XAIImageOptions } from "../protocols/xai-images"
 
 export const routes = [OpenAIResponses.route, OpenAICompatibleChat.route]
@@ -42,8 +51,8 @@ const configuredChatRoute = (input: ModelOptions) => {
 export const configure = (input: ModelOptions = {}) => {
   const responsesRoute = configuredResponsesRoute(input)
   const chatRoute = configuredChatRoute(input)
-  const responses = (modelID: string | ModelID) => responsesRoute.model({ id: modelID })
-  const chat = (modelID: string | ModelID) => chatRoute.model({ id: modelID })
+  const responses = (modelID: string | ModelID) => responsesRoute.model<OpenAIProviderOptionsInput>({ id: modelID })
+  const chat = (modelID: string | ModelID) => chatRoute.model<OpenAIProviderOptionsInput>({ id: modelID })
   const image = (modelID: string | ModelID) =>
     XAIImages.model({
       id: modelID,
@@ -63,7 +72,15 @@ export const configure = (input: ModelOptions = {}) => {
 }
 
 export const provider = configure()
-export const model = provider.model
+export const model: ProviderPackage.Definition<Settings, OpenAIProviderOptionsInput>["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)
 export const responses = provider.responses
 export const chat = provider.chat
 export const image = provider.image

packages/ai/src/route/client.ts

@@ -10,7 +10,7 @@ import { WebSocketExecutor } from "./transport"
 import type { Protocol } from "./protocol"
 import { applyCachePolicy } from "../cache-policy"
 import * as ProviderShared from "../protocols/shared"
-import type { LLMError, PreparedRequestOf, ProtocolID, ProviderOptions } from "../schema"
+import type { LLMError, ProtocolID, ProviderOptions } from "../schema"
 import {
   GenerationOptions,
   HttpOptions,
@@ -20,7 +20,6 @@ import {
   ModelLimits,
   LLMError as LLMErrorClass,
   LLMEvent,
-  PreparedRequest,
   ProviderID,
   mergeGenerationOptions,
   mergeHttpOptions,
@@ -46,7 +45,7 @@ export interface Route<Body, Prepared = unknown> {
   readonly defaults: RouteDefaults
   readonly body: RouteBody<Body>
   readonly with: (patch: RoutePatch<Body, Prepared>) => Route<Body, Prepared>
-  readonly model: (input: RouteMappedModelInput) => Model
+  readonly model: <Options extends ProviderOptions = ProviderOptions>(input: RouteMappedModelInput) => Model<Options>
   readonly prepareTransport: (body: Body, request: LLMRequest) => Effect.Effect<Prepared, LLMError>
   readonly streamPrepared: (
     prepared: Prepared,
@@ -93,12 +92,12 @@ export interface RoutePatch<Body, Prepared> extends RouteDefaultsInput {
 
 type RouteMappedModelInput = RouteModelInput | RouteRoutedModelInput
 
-const makeRouteModel = (route: AnyRoute, mapped: RouteMappedModelInput) => {
+const makeRouteModel = <Options extends ProviderOptions = ProviderOptions>(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({
+  return Model.make<Options>({
     ...mapped,
     provider,
     route,
@@ -142,17 +141,6 @@ 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
 }
@@ -296,7 +284,8 @@ function makeFromTransport<Body, Prepared, Frame, Event, State>(
           defaults: mergeRouteDefaults(route.defaults, defaults),
         })
       },
-      model: (input) => makeRouteModel(route, input),
+      model: <Options extends ProviderOptions = ProviderOptions>(input: RouteMappedModelInput) =>
+        makeRouteModel<Options>(route, input),
       prepareTransport: (body, request) =>
         routeInput.transport.prepare({
           body,
@@ -370,9 +359,6 @@ export function make<Body, Prepared, Frame, Event, State>(
   })
 }
 
-// `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
@@ -390,17 +376,17 @@ const compile = Effect.fn("LLM.compile")(function* (request: LLMRequest) {
   }
 })
 
-const prepareWith = Effect.fn("LLMClient.prepare")(function* (request: LLMRequest) {
+/** @internal Test-only projection of the execution compiler; not exported from package barrels. */
+export const compileRequest = Effect.fn("LLM.compileRequest")(function* (request: LLMRequest) {
   const compiled = yield* compile(request)
-
-  return new PreparedRequest({
+  return {
     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) =>
@@ -422,9 +408,6 @@ const generateWith = (stream: Interface["stream"]) =>
     )
   })
 
-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* () {
@@ -453,7 +436,7 @@ 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({ prepare: prepareWith as Interface["prepare"], stream, generate: generateWith(stream) })
+    return Service.of({ stream, generate: generateWith(stream) })
   }),
 )
 
@@ -462,7 +445,6 @@ export const Route = { make } as const
 export const LLMClient = {
   Service,
   layer,
-  prepare,
   stream,
   generate,
 } as const

packages/ai/src/schema/events.ts

@@ -1,6 +1,5 @@
 import { Schema } from "effect"
-import { ContentBlockID, FinishReason, ProtocolID, ProviderMetadata, RouteID, ToolCallID } from "./ids"
-import { ModelSchema } from "./options"
+import { ContentBlockID, FinishReason, ProviderMetadata, ToolCallID } from "./ids"
 import { Message, ToolCallPart, ToolOutput, ToolResultPart, ToolResultValue, type ContentPart } from "./messages"
 import { ProviderFailureClassification } from "./errors"
 
@@ -314,29 +313,6 @@ 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/options.ts

@@ -178,7 +178,8 @@ export namespace ModelCompatibility {
   export const make = (input: Input) => (input instanceof ModelCompatibility ? input : new ModelCompatibility(input))
 }
 
-export class Model {
+export class Model<Options extends ProviderOptions = ProviderOptions> {
+  declare protected readonly _ProviderOptions: Options
   readonly id: ModelID
   readonly provider: ProviderID
   readonly route: AnyRoute
@@ -193,8 +194,8 @@ export class Model {
     this.compatibility = input.compatibility
   }
 
-  static make(input: Model.Input) {
-    return new Model({
+  static make<Options extends ProviderOptions = ProviderOptions>(input: Model.Input) {
+    return new Model<Options>({
       id: ModelID.make(input.id),
       provider: ProviderID.make(input.provider),
       route: input.route,
@@ -203,7 +204,7 @@ export class Model {
     })
   }
 
-  static input(model: Model): Model.ConstructorInput {
+  static input<Options extends ProviderOptions>(model: Model<Options>): Model.ConstructorInput {
     return {
       id: model.id,
       provider: model.provider,
@@ -213,9 +214,9 @@ export class Model {
     }
   }
 
-  static update(model: Model, patch: Partial<Model.Input>) {
+  static update<Options extends ProviderOptions>(model: Model<Options>, patch: Partial<Model.Input>) {
     if (Object.keys(patch).length === 0) return model
-    return Model.make({
+    return Model.make<Options>({
       ...Model.input(model),
       ...patch,
     })
@@ -241,6 +242,8 @@ 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

@@ -99,7 +99,6 @@ 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

@@ -2,6 +2,7 @@ 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"
@@ -139,8 +140,7 @@ describe("llm route", () => {
 
   it.effect("selects routes by model route value", () =>
     Effect.gen(function* () {
-      const llm = yield* LLMClient.Service
-      const prepared = yield* llm.prepare(
+      const prepared = yield* compileRequest(
         LLMRequest.update(request, { model: updateModel(request.model, { route: configuredGemini }) }),
       )
 
@@ -173,7 +173,7 @@ describe("llm route", () => {
         framing: fakeFraming,
       })
 
-      const prepared = yield* (yield* LLMClient.Service).prepare(
+      const prepared = yield* compileRequest(
         LLMRequest.update(request, { model: updateModel(request.model, { route: duplicate }) }),
       )
 

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

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

packages/ai/test/compile.test.ts

@@ -4,6 +4,7 @@ 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"
@@ -44,7 +45,7 @@ describe("request option precedence", () => {
     })
   })
 
-  it.effect("prepares bodies with route defaults, model defaults, and call options in order", () =>
+  it.effect("compiles 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/" },
@@ -59,7 +60,7 @@ describe("request option precedence", () => {
           providerOptions: { openai: { reasoningEffort: "medium" } },
         },
       })
-      const prepared = yield* LLMClient.prepare<OpenAIChat.OpenAIChatBody>(
+      const prepared = yield* compileRequest(
         LLM.request({
           model,
           prompt: "Say hello.",
@@ -141,7 +142,7 @@ describe("request option precedence", () => {
       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(
+      const error = yield* compileRequest(
         LLM.request({
           model,
           prompt: "Say hello.",
@@ -164,10 +165,8 @@ describe("request option precedence", () => {
         limits: { output: 128 },
       })
       const model = route.model({ id: "claude-sonnet-4-5", defaults: { limits: { output: 64 } } })
-      const withoutMaxTokens = yield* LLMClient.prepare<AnthropicMessages.AnthropicMessagesBody>(
-        LLM.request({ model, prompt: "Say hello.", cache: "none" }),
-      )
-      const withMaxTokens = yield* LLMClient.prepare<AnthropicMessages.AnthropicMessagesBody>(
+      const withoutMaxTokens = yield* compileRequest(LLM.request({ model, prompt: "Say hello.", cache: "none" }))
+      const withMaxTokens = yield* compileRequest(
         LLM.request({ model, prompt: "Say hello.", cache: "none", generation: { maxTokens: 32 } }),
       )
 

packages/ai/test/exports.test.ts

@@ -53,9 +53,7 @@ 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

@@ -0,0 +1,47 @@
+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/provider-options/anthropic-compatible.types.ts

@@ -0,0 +1,13 @@
+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

@@ -0,0 +1,13 @@
+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

@@ -0,0 +1,13 @@
+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

@@ -0,0 +1,13 @@
+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

@@ -0,0 +1,13 @@
+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

@@ -0,0 +1,13 @@
+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

@@ -0,0 +1,13 @@
+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

@@ -0,0 +1,13 @@
+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

@@ -0,0 +1,17 @@
+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

@@ -0,0 +1,17 @@
+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",
+  // @ts-expect-error Gemini thinking budgets must be numeric.
+  providerOptions: { gemini: { thinkingConfig: { thinkingBudget: "large" } } },
+})

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

@@ -0,0 +1,13 @@
+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

@@ -0,0 +1,13 @@
+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

@@ -0,0 +1,13 @@
+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

@@ -0,0 +1,13 @@
+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",
+  // @ts-expect-error OpenRouter usage must be boolean or an option record.
+  providerOptions: { openrouter: { usage: "yes" } },
+})

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

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

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

@@ -21,6 +21,8 @@ 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()
@@ -29,6 +31,35 @@ 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: { openai: { 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).toEqual({ openai: { 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

@@ -3,6 +3,7 @@ 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"
@@ -44,7 +45,7 @@ const expectToolResult = (body: AnthropicMessages.AnthropicMessagesBody): Anthro
 describe("Anthropic Messages route", () => {
   it.effect("prepares Anthropic Messages target", () =>
     Effect.gen(function* () {
-      const prepared = yield* LLMClient.prepare(request)
+      const prepared = yield* compileRequest(request)
 
       expect(prepared.body).toEqual({
         model: "claude-sonnet-4-5",
@@ -59,7 +60,7 @@ describe("Anthropic Messages route", () => {
 
   it.effect("lowers adaptive thinking settings with effort", () =>
     Effect.gen(function* () {
-      const prepared = yield* LLMClient.prepare<AnthropicMessages.AnthropicMessagesBody>(
+      const prepared = yield* compileRequest(
         LLMRequest.update(request, {
           providerOptions: {
             anthropic: { thinking: { type: "adaptive", display: "summarized" }, effort: "low" },
@@ -76,17 +77,17 @@ describe("Anthropic Messages route", () => {
 
   it.effect("normalizes enabled and disabled thinking settings", () =>
     Effect.gen(function* () {
-      const enabled = yield* LLMClient.prepare<AnthropicMessages.AnthropicMessagesBody>(
+      const enabled = yield* compileRequest(
         LLMRequest.update(request, {
           providerOptions: { anthropic: { thinking: { type: "enabled", budgetTokens: 1_024 } } },
         }),
       )
-      const legacy = yield* LLMClient.prepare<AnthropicMessages.AnthropicMessagesBody>(
+      const legacy = yield* compileRequest(
         LLMRequest.update(request, {
           providerOptions: { anthropic: { thinking: { type: "enabled", budget_tokens: 2_048 } } },
         }),
       )
-      const disabled = yield* LLMClient.prepare<AnthropicMessages.AnthropicMessagesBody>(
+      const disabled = yield* compileRequest(
         LLMRequest.update(request, {
           providerOptions: { anthropic: { thinking: { type: "disabled" } } },
         }),
@@ -100,7 +101,7 @@ describe("Anthropic Messages route", () => {
 
   it.effect("rejects enabled thinking without a budget", () =>
     Effect.gen(function* () {
-      const error = yield* LLMClient.prepare(
+      const error = yield* compileRequest(
         LLMRequest.update(request, {
           providerOptions: { anthropic: { thinking: { type: "enabled" } } },
         }),
@@ -112,7 +113,7 @@ 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* LLMClient.prepare<AnthropicMessages.AnthropicMessagesBody>(
+      const prepared = yield* compileRequest(
         LLM.request({
           model: opus48,
           messages: [
@@ -137,7 +138,7 @@ describe("Anthropic Messages route", () => {
 
   it.effect("lowers chronological system updates to wrapped user text for unsupported Anthropic models", () =>
     Effect.gen(function* () {
-      const prepared = yield* LLMClient.prepare<AnthropicMessages.AnthropicMessagesBody>(
+      const prepared = yield* compileRequest(
         LLM.request({
           model,
           messages: [
@@ -164,7 +165,7 @@ describe("Anthropic Messages route", () => {
 
   it.effect("rejects non-text chronological system update content before send", () =>
     Effect.gen(function* () {
-      const error = yield* LLMClient.prepare(
+      const error = yield* compileRequest(
         LLM.request({
           model: opus48,
           messages: [
@@ -181,7 +182,7 @@ describe("Anthropic Messages route", () => {
   it.effect("falls back for unsupported native chronological system update placement", () =>
     Effect.gen(function* () {
       expect(
-        (yield* LLMClient.prepare<AnthropicMessages.AnthropicMessagesBody>(
+        (yield* compileRequest(
           LLM.request({
             model: opus48,
             messages: [Message.assistant("Plain."), Message.system("After plain assistant.")],
@@ -196,12 +197,11 @@ describe("Anthropic Messages route", () => {
         },
       ])
       expect(
-        (yield* LLMClient.prepare<AnthropicMessages.AnthropicMessagesBody>(
-          LLM.request({ model: opus48, messages: [Message.system("First.")], cache: "none" }),
-        )).body.messages,
+        (yield* compileRequest(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* LLMClient.prepare<AnthropicMessages.AnthropicMessagesBody>(
+        (yield* compileRequest(
           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* LLMClient.prepare(
+      const error = yield* compileRequest(
         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* LLMClient.prepare<AnthropicMessages.AnthropicMessagesBody>(
+      const prepared = yield* compileRequest(
         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* LLMClient.prepare<AnthropicMessages.AnthropicMessagesBody>(
+      const prepared = yield* compileRequest(
         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* LLMClient.prepare<AnthropicMessages.AnthropicMessagesBody>(
+      const prepared = yield* compileRequest(
         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* LLMClient.prepare<AnthropicMessages.AnthropicMessagesBody>(
+      const prepared = yield* compileRequest(
         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* LLMClient.prepare(
+      const error = yield* compileRequest(
         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* LLMClient.prepare<AnthropicMessages.AnthropicMessagesBody>(
+      const prepared = yield* compileRequest(
         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* LLMClient.prepare(
+      const prepared = yield* compileRequest(
         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* LLMClient.prepare(
+      const prepared = yield* compileRequest(
         LLM.request({
           model,
           messages: [
@@ -628,9 +628,7 @@ describe("Anthropic Messages route", () => {
         { type: "reasoning", text: "", providerMetadata: { anthropic: { signature: "sig_1" } } },
       ])
 
-      const prepared = yield* LLMClient.prepare<AnthropicMessages.AnthropicMessagesBody>(
-        LLM.request({ model, messages: [response.message], cache: "none" }),
-      )
+      const prepared = yield* compileRequest(LLM.request({ model, messages: [response.message], cache: "none" }))
       expect(prepared.body.messages).toEqual([
         { role: "assistant", content: [{ type: "thinking", thinking: "", signature: "sig_1" }] },
       ])
@@ -773,7 +771,7 @@ describe("Anthropic Messages route", () => {
           ),
         ),
       )
-      const prepared = yield* LLMClient.prepare<AnthropicMessages.AnthropicMessagesBody>(
+      const prepared = yield* compileRequest(
         LLM.request({
           model,
           messages: [
@@ -1133,7 +1131,7 @@ describe("Anthropic Messages route", () => {
 
   it.effect("round-trips provider-executed assistant content into server tool blocks", () =>
     Effect.gen(function* () {
-      const prepared = yield* LLMClient.prepare(
+      const prepared = yield* compileRequest(
         LLM.request({
           id: "req_round_trip",
           model,
@@ -1184,7 +1182,7 @@ describe("Anthropic Messages route", () => {
 
   it.effect("rejects round-trip for unknown server tool names", () =>
     Effect.gen(function* () {
-      const error = yield* LLMClient.prepare(
+      const error = yield* compileRequest(
         LLM.request({
           id: "req_unknown_server_tool",
           model,
@@ -1261,7 +1259,7 @@ describe("Anthropic Messages route", () => {
 
   it.effect("maps ttlSeconds >= 3600 to cache_control ttl: '1h'", () =>
     Effect.gen(function* () {
-      const prepared = yield* LLMClient.prepare(
+      const prepared = yield* compileRequest(
         LLM.request({
           model,
           system: { type: "text", text: "system", cache: new CacheHint({ type: "ephemeral", ttlSeconds: 3600 }) },
@@ -1277,7 +1275,7 @@ describe("Anthropic Messages route", () => {
 
   it.effect("emits cache_control on tool definitions and tool-result blocks", () =>
     Effect.gen(function* () {
-      const prepared = yield* LLMClient.prepare(
+      const prepared = yield* compileRequest(
         LLM.request({
           model,
           tools: [
@@ -1318,7 +1316,7 @@ 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* LLMClient.prepare(
+      const prepared = yield* compileRequest(
         LLM.request({
           model,
           system: [
@@ -1344,7 +1342,7 @@ 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* LLMClient.prepare(
+      const prepared = yield* compileRequest(
         LLM.request({
           model,
           tools: [

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

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

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

@@ -2,6 +2,7 @@ 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"
@@ -26,7 +27,7 @@ const request = LLM.request({
 describe("Gemini route", () => {
   it.effect("prepares Gemini target", () =>
     Effect.gen(function* () {
-      const prepared = yield* LLMClient.prepare(request)
+      const prepared = yield* compileRequest(request)
 
       expect(prepared.body).toEqual({
         contents: [{ role: "user", parts: [{ text: "Say hello." }] }],
@@ -38,12 +39,12 @@ describe("Gemini route", () => {
 
   it.effect("normalizes Gemini thinking options", () =>
     Effect.gen(function* () {
-      const prepared = yield* LLMClient.prepare<Gemini.GeminiBody>(
+      const prepared = yield* compileRequest(
         LLMRequest.update(request, {
           providerOptions: { gemini: { thinkingConfig: { thinkingBudget: 0, includeThoughts: false } } },
         }),
       )
-      const filtered = yield* LLMClient.prepare<Gemini.GeminiBody>(
+      const filtered = yield* compileRequest(
         LLMRequest.update(request, {
           providerOptions: { gemini: { thinkingConfig: { thinkingBudget: "invalid", includeThoughts: false } } },
         }),
@@ -59,7 +60,7 @@ describe("Gemini route", () => {
 
   it.effect("lowers chronological system updates to wrapped user text in order", () =>
     Effect.gen(function* () {
-      const prepared = yield* LLMClient.prepare<Gemini.GeminiBody>(
+      const prepared = yield* compileRequest(
         LLM.request({
           model,
           messages: [Message.user("Before."), Message.system("Update."), Message.assistant("After.")],
@@ -75,7 +76,7 @@ describe("Gemini route", () => {
 
   it.effect("prepares multimodal user input and tool history", () =>
     Effect.gen(function* () {
-      const prepared = yield* LLMClient.prepare(
+      const prepared = yield* compileRequest(
         LLM.request({
           id: "req_tool_result",
           model,
@@ -143,7 +144,7 @@ describe("Gemini route", () => {
 
   it.effect("continues media tool results as inline model input without base64 text", () =>
     Effect.gen(function* () {
-      const prepared = yield* LLMClient.prepare<Gemini.GeminiBody>(
+      const prepared = yield* compileRequest(
         LLM.request({
           model,
           messages: [
@@ -188,7 +189,7 @@ describe("Gemini route", () => {
 
   it.effect("strips matching data URLs to raw base64 inlineData", () =>
     Effect.gen(function* () {
-      const prepared = yield* LLMClient.prepare<Gemini.GeminiBody>(
+      const prepared = yield* compileRequest(
         LLM.request({
           model,
           messages: [
@@ -229,7 +230,7 @@ describe("Gemini route", () => {
   ] as const)
     it.effect(`rejects ${name}`, () =>
       Effect.gen(function* () {
-        const error = yield* LLMClient.prepare(
+        const error = yield* compileRequest(
           LLM.request({ model, messages: [Message.user({ type: "media", ...media })] }),
         ).pipe(Effect.flip)
         expect(error.message).toMatch(/does not support|does not match|valid base64/)
@@ -238,7 +239,7 @@ describe("Gemini route", () => {
 
   it.effect("rejects oversized image input", () =>
     Effect.gen(function* () {
-      const error = yield* LLMClient.prepare(
+      const error = yield* compileRequest(
         LLM.request({
           model,
           messages: [
@@ -256,7 +257,7 @@ describe("Gemini route", () => {
 
   it.effect("keeps tools and sends function calling mode NONE", () =>
     Effect.gen(function* () {
-      const prepared = yield* LLMClient.prepare(
+      const prepared = yield* compileRequest(
         LLM.request({
           id: "req_tool_choice_none",
           model,
@@ -276,7 +277,7 @@ describe("Gemini route", () => {
 
   it.effect("sanitizes integer enums, dangling required, untyped arrays, and scalar object keys", () =>
     Effect.gen(function* () {
-      const prepared = yield* LLMClient.prepare(
+      const prepared = yield* compileRequest(
         LLM.request({
           id: "req_schema_patch",
           model,
@@ -457,7 +458,7 @@ describe("Gemini route", () => {
         response.events.findIndex((event) => event.type === "tool-call"),
       )
 
-      const prepared = yield* LLMClient.prepare<Gemini.GeminiBody>(
+      const prepared = yield* compileRequest(
         LLM.request({
           model,
           messages: [
@@ -691,7 +692,7 @@ describe("Gemini route", () => {
 
   it.effect("rejects unsupported assistant media content", () =>
     Effect.gen(function* () {
-      const error = yield* LLMClient.prepare(
+      const error = yield* compileRequest(
         LLM.request({
           id: "req_media",
           model,

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

@@ -4,6 +4,7 @@ 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"
@@ -182,7 +183,7 @@ describe("Google Vertex providers", () => {
 
   it.effect("protects the Vertex Messages API version from body overlays", () =>
     Effect.gen(function* () {
-      const error = yield* LLMClient.prepare(
+      const error = yield* compileRequest(
         LLM.request({
           model: GoogleVertexMessages.configure({
             accessToken: "vertex-token",

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

@@ -5,6 +5,7 @@ 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"
 
@@ -84,9 +85,7 @@ for (const item of cases) {
             ),
           ).toBe(true)
 
-          const replay = yield* LLMClient.prepare<OpenAIChat.OpenAIChatBody>(
-            LLM.request({ model: item.model, messages: [response.message] }),
-          )
+          const replay = yield* compileRequest(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

@@ -18,6 +18,7 @@ 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"
@@ -42,11 +43,7 @@ const request = LLM.request({
 describe("OpenAI Chat route", () => {
   it.effect("prepares OpenAI Chat payload", () =>
     Effect.gen(function* () {
-      // 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
+      const prepared = yield* compileRequest(request)
 
       expect(prepared.body).toEqual({
         model: "gpt-4o-mini",
@@ -64,7 +61,7 @@ describe("OpenAI Chat route", () => {
 
   it.effect("lowers chronological system updates to escaped user wrappers in order", () =>
     Effect.gen(function* () {
-      const prepared = yield* LLMClient.prepare<OpenAIChat.OpenAIChatBody>(
+      const prepared = yield* compileRequest(
         LLM.request({
           model,
           messages: [
@@ -87,7 +84,7 @@ describe("OpenAI Chat route", () => {
 
   it.effect("replays canonical reasoning as OpenAI-compatible reasoning_content", () =>
     Effect.gen(function* () {
-      const prepared = yield* LLMClient.prepare<OpenAIChat.OpenAIChatBody>(
+      const prepared = yield* compileRequest(
         LLM.request({
           model,
           messages: [
@@ -105,7 +102,7 @@ describe("OpenAI Chat route", () => {
 
   it.effect("writes reasoning to a configured custom field on every assistant message", () =>
     Effect.gen(function* () {
-      const prepared = yield* LLMClient.prepare<OpenAIChat.OpenAIChatBody>(
+      const prepared = yield* compileRequest(
         LLM.request({
           model: Model.update(model, { compatibility: { reasoningField: "vendor_reasoning" } }),
           messages: [
@@ -131,7 +128,7 @@ describe("OpenAI Chat route", () => {
 
   it.effect("rejects reasoning fields that conflict with assistant message fields", () =>
     Effect.gen(function* () {
-      const error = yield* LLMClient.prepare(
+      const error = yield* compileRequest(
         LLM.request({
           model: Model.update(model, { compatibility: { reasoningField: "content" } }),
           messages: [Message.assistant([{ type: "reasoning", text: "thinking" }])],
@@ -144,7 +141,7 @@ describe("OpenAI Chat route", () => {
 
   it.effect("maps OpenAI provider options to Chat options", () =>
     Effect.gen(function* () {
-      const prepared = yield* LLMClient.prepare<OpenAIChat.OpenAIChatBody>(
+      const prepared = yield* compileRequest(
         LLM.request({
           model: OpenAI.configure({ baseURL: "https://api.openai.test/v1/", apiKey: "test" }).chat("gpt-4o-mini"),
           prompt: "think",
@@ -159,7 +156,7 @@ describe("OpenAI Chat route", () => {
 
   it.effect("passes through custom OpenAI-compatible reasoning effort strings", () =>
     Effect.gen(function* () {
-      const prepared = yield* LLMClient.prepare<OpenAIChat.OpenAIChatBody>(
+      const prepared = yield* compileRequest(
         LLM.request({
           model,
           prompt: "think",
@@ -253,7 +250,7 @@ describe("OpenAI Chat route", () => {
 
   it.effect("prepares assistant tool-call and tool-result messages", () =>
     Effect.gen(function* () {
-      const prepared = yield* LLMClient.prepare(
+      const prepared = yield* compileRequest(
         LLM.request({
           id: "req_tool_result",
           model,
@@ -291,7 +288,7 @@ 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* LLMClient.prepare<OpenAIChat.OpenAIChatBody>(
+      const prepared = yield* compileRequest(
         LLM.request({
           model,
           messages: [
@@ -311,7 +308,7 @@ describe("OpenAI Chat route", () => {
 
   it.effect("continues image tool results as vision input without base64 text", () =>
     Effect.gen(function* () {
-      const prepared = yield* LLMClient.prepare<OpenAIChat.OpenAIChatBody>(
+      const prepared = yield* compileRequest(
         LLM.request({
           model,
           messages: [
@@ -355,7 +352,7 @@ describe("OpenAI Chat route", () => {
 
   it.effect("orders parallel tool responses before one aggregated vision message", () =>
     Effect.gen(function* () {
-      const prepared = yield* LLMClient.prepare<OpenAIChat.OpenAIChatBody>(
+      const prepared = yield* compileRequest(
         LLM.request({
           model,
           messages: [
@@ -405,7 +402,7 @@ describe("OpenAI Chat route", () => {
 
   it.effect("aggregates consecutive tool images with a following system update", () =>
     Effect.gen(function* () {
-      const prepared = yield* LLMClient.prepare<OpenAIChat.OpenAIChatBody>(
+      const prepared = yield* compileRequest(
         LLM.request({
           model,
           messages: [
@@ -446,7 +443,7 @@ describe("OpenAI Chat route", () => {
 
   it.effect("appends system updates without replacing multipart user content", () =>
     Effect.gen(function* () {
-      const prepared = yield* LLMClient.prepare<OpenAIChat.OpenAIChatBody>(
+      const prepared = yield* compileRequest(
         LLM.request({
           model,
           messages: [
@@ -474,7 +471,7 @@ describe("OpenAI Chat route", () => {
   ] as const)
     it.effect(`rejects ${name}`, () =>
       Effect.gen(function* () {
-        const error = yield* LLMClient.prepare(
+        const error = yield* compileRequest(
           LLM.request({ model, messages: [Message.user({ type: "media", ...media })] }),
         ).pipe(Effect.flip)
         expect(error.message).toMatch(/does not support|does not match|valid base64/)
@@ -483,7 +480,7 @@ describe("OpenAI Chat route", () => {
 
   it.effect("rejects oversized image input", () =>
     Effect.gen(function* () {
-      const error = yield* LLMClient.prepare(
+      const error = yield* compileRequest(
         LLM.request({
           model,
           messages: [
@@ -501,7 +498,7 @@ describe("OpenAI Chat route", () => {
 
   it.effect("prepares raw and data URL image media as vision input", () =>
     Effect.gen(function* () {
-      const prepared = yield* LLMClient.prepare<OpenAIChat.OpenAIChatBody>(
+      const prepared = yield* compileRequest(
         LLM.request({
           id: "req_media",
           model,
@@ -528,7 +525,7 @@ describe("OpenAI Chat route", () => {
 
   it.effect("lowers reasoning-only assistant history", () =>
     Effect.gen(function* () {
-      const prepared = yield* LLMClient.prepare<OpenAIChat.OpenAIChatBody>(
+      const prepared = yield* compileRequest(
         LLM.request({
           id: "req_reasoning",
           model,
@@ -619,9 +616,7 @@ describe("OpenAI Chat route", () => {
           openai: { reasoningField: field },
         })
 
-        const replay = yield* LLMClient.prepare<OpenAIChat.OpenAIChatBody>(
-          LLM.request({ model, messages: [response.message] }),
-        )
+        const replay = yield* compileRequest(LLM.request({ model, messages: [response.message] }))
         expect(replay.body.messages).toEqual([{ role: "assistant", content: "Hello", [field]: "thinking" }])
       }
     }),
@@ -647,9 +642,7 @@ describe("OpenAI Chat route", () => {
         openai: { reasoningField: "vendor_reasoning" },
       })
 
-      const replay = yield* LLMClient.prepare<OpenAIChat.OpenAIChatBody>(
-        LLM.request({ model: custom, messages: [response.message] }),
-      )
+      const replay = yield* compileRequest(LLM.request({ model: custom, messages: [response.message] }))
       expect(replay.body.messages).toEqual([{ role: "assistant", content: "Hello", vendor_reasoning: "thinking" }])
     }),
   )
@@ -692,9 +685,7 @@ describe("OpenAI Chat route", () => {
         openai: { reasoningField: "reasoning", reasoningDetails: details },
       })
 
-      const replay = yield* LLMClient.prepare<OpenAIChat.OpenAIChatBody>(
-        LLM.request({ model, messages: [response.message] }),
-      )
+      const replay = yield* compileRequest(LLM.request({ model, messages: [response.message] }))
       expect(replay.body.messages).toEqual([
         {
           role: "assistant",
@@ -737,9 +728,7 @@ describe("OpenAI Chat route", () => {
         openai: { reasoningDetails: details },
       })
 
-      const replay = yield* LLMClient.prepare<OpenAIChat.OpenAIChatBody>(
-        LLM.request({ model, messages: [response.message] }),
-      )
+      const replay = yield* compileRequest(LLM.request({ model, messages: [response.message] }))
       expect(replay.body.messages).toEqual([{ role: "assistant", content: "Hello", reasoning_details: details }])
     }),
   )
@@ -764,9 +753,7 @@ describe("OpenAI Chat route", () => {
         openai: { reasoningField: "reasoning", reasoningDetails: details },
       })
 
-      const replay = yield* LLMClient.prepare<OpenAIChat.OpenAIChatBody>(
-        LLM.request({ model, messages: [response.message] }),
-      )
+      const replay = yield* compileRequest(LLM.request({ model, messages: [response.message] }))
       expect(replay.body.messages).toEqual([
         { role: "assistant", content: "Hello", reasoning: "thinking", reasoning_details: details },
       ])
@@ -839,9 +826,7 @@ describe("OpenAI Chat route", () => {
         openai: { reasoningDetails: [] },
       })
 
-      const replay = yield* LLMClient.prepare<OpenAIChat.OpenAIChatBody>(
-        LLM.request({ model, messages: [response.message] }),
-      )
+      const replay = yield* compileRequest(LLM.request({ model, messages: [response.message] }))
       expect(replay.body.messages).toEqual([{ role: "assistant", content: "Hello", reasoning_details: [] }])
     }),
   )
@@ -889,9 +874,7 @@ describe("OpenAI Chat route", () => {
         response.events.findIndex(LLMEvent.is.textStart),
       )
 
-      const replay = yield* LLMClient.prepare<OpenAIChat.OpenAIChatBody>(
-        LLM.request({ model, messages: [response.message] }),
-      )
+      const replay = yield* compileRequest(LLM.request({ model, messages: [response.message] }))
       expect(replay.body.messages).toEqual([
         { role: "assistant", content: "Hello", reasoning: "thinking", reasoning_details: merged },
       ])
@@ -918,9 +901,7 @@ 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* LLMClient.prepare<OpenAIChat.OpenAIChatBody>(
-        LLM.request({ model, messages: [response.message] }),
-      )
+      const replay = yield* compileRequest(LLM.request({ model, messages: [response.message] }))
       expect(replay.body.messages).toEqual([{ role: "assistant", content: null, reasoning_details: details }])
     }),
   )
@@ -950,7 +931,7 @@ 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* LLMClient.prepare<OpenAIChat.OpenAIChatBody>(
+      const replay = yield* compileRequest(
         LLM.request({
           model,
           messages: [
@@ -979,7 +960,7 @@ 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* LLMClient.prepare<OpenAIChat.OpenAIChatBody>(
+      const replay = yield* compileRequest(
         LLM.request({
           model,
           messages: [
@@ -1004,7 +985,7 @@ 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* LLMClient.prepare<OpenAIChat.OpenAIChatBody>(
+      const replay = yield* compileRequest(
         LLM.request({
           model,
           messages: [

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

@@ -3,6 +3,7 @@ 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"
@@ -52,7 +53,7 @@ const providerFamilies = [
 describe("OpenAI-compatible Chat route", () => {
   it.effect("prepares generic Chat target", () =>
     Effect.gen(function* () {
-      const prepared = yield* LLMClient.prepare(
+      const prepared = yield* compileRequest(
         LLMRequest.update(request, {
           tools: [ToolDefinition.make({ name: "lookup", description: "Lookup data", inputSchema: { type: "object" } })],
           toolChoice: ToolChoice.make({ type: "required" }),
@@ -127,7 +128,7 @@ describe("OpenAI-compatible Chat route", () => {
 
   it.effect("matches AI SDK compatible basic request body fixture", () =>
     Effect.gen(function* () {
-      const prepared = yield* LLMClient.prepare(request)
+      const prepared = yield* compileRequest(request)
 
       expect(prepared.body).toEqual({
         model: "deepseek-chat",
@@ -145,7 +146,7 @@ describe("OpenAI-compatible Chat route", () => {
 
   it.effect("matches AI SDK compatible tool request body fixture", () =>
     Effect.gen(function* () {
-      const prepared = yield* LLMClient.prepare(
+      const prepared = yield* compileRequest(
         LLM.request({
           id: "req_tool_parity",
           model,

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

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

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

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

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

@@ -2,6 +2,7 @@ import { describe, expect } from "bun:test"
 import { Effect } from "effect"
 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"
@@ -19,7 +20,7 @@ describe("OpenRouter", () => {
       })
       expect(model.route.endpoint.baseURL).toBe("https://openrouter.ai/api/v1")
 
-      const prepared = yield* LLMClient.prepare(LLM.request({ model, prompt: "Say hello." }))
+      const prepared = yield* compileRequest(LLM.request({ model, prompt: "Say hello." }))
 
       expect(prepared.route).toBe("openrouter")
       expect(prepared.body).toMatchObject({
@@ -32,7 +33,7 @@ describe("OpenRouter", () => {
 
   it.effect("applies OpenRouter payload options from the model helper", () =>
     Effect.gen(function* () {
-      const prepared = yield* LLMClient.prepare(
+      const prepared = yield* compileRequest(
         LLM.request({
           model: OpenRouter.configure({
             apiKey: "test-key",
@@ -100,7 +101,7 @@ 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* LLMClient.prepare<OpenRouter.OpenRouterBody>(
+      const prepared = yield* compileRequest(
         LLM.request({
           model: OpenRouter.configure({ apiKey: "test-key" }).model("anthropic/claude-sonnet-4.6"),
           messages: [
@@ -133,7 +134,7 @@ describe("OpenRouter", () => {
         { type: "reasoning.encrypted", id: "state", data: "opaque" },
         { type: "reasoning.encrypted", id: "state", data: "opaque" },
       ]
-      const prepared = yield* LLMClient.prepare<OpenRouter.OpenRouterBody>(
+      const prepared = yield* compileRequest(
         LLM.request({
           model: OpenRouter.configure({ apiKey: "test-key" }).model("anthropic/claude-sonnet-4.6"),
           messages: [
@@ -158,7 +159,7 @@ 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* LLMClient.prepare<OpenRouter.OpenRouterBody>(
+      const prepared = yield* compileRequest(
         LLM.request({
           model: OpenRouter.configure({ apiKey: "test-key" }).model("anthropic/claude-sonnet-4.6"),
           messages: [
@@ -179,7 +180,7 @@ describe("OpenRouter", () => {
 
   it.effect("omits scalar reasoning without continuation details", () =>
     Effect.gen(function* () {
-      const prepared = yield* LLMClient.prepare<OpenRouter.OpenRouterBody>(
+      const prepared = yield* compileRequest(
         LLM.request({
           model: OpenRouter.configure({ apiKey: "test-key" }).model("anthropic/claude-sonnet-4.6"),
           messages: [Message.assistant({ type: "reasoning", text: "Thinking" })],

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

@@ -3,7 +3,8 @@ import { Effect } from "effect"
 import { LLM } from "../src"
 import { OpenAIChat } from "../src/protocols"
 import { ToolSchemaProjection } from "../src/protocols/utils/tool-schema"
-import { Auth, LLMClient } from "../src/route"
+import { Auth } from "../src/route"
+import { compileRequest } from "../src/route/client"
 import { it } from "./lib/effect"
 
 describe("tool schema projections", () => {
@@ -79,7 +80,7 @@ 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* LLMClient.prepare<OpenAIChat.OpenAIChatBody>(
+      const prepared = yield* compileRequest(
         LLM.request({
           model,
           prompt: "Use the tool.",

packages/app/V1_API_MIGRATION.md

@@ -1,220 +0,0 @@
-# V1 API Migration Checklist
-
-The app is currently hybrid. In this document, V1 refers to the legacy unprefixed server APIs used by `@opencode-ai/sdk/v2`, despite the SDK package name.
-
-## Events
-
-- [x] Replace `GET /global/event` with `GET /api/event`.
-  - `src/context/server-sdk.tsx`
-- [x] Reduce current granular session and message events into the existing app projections.
-  - `src/context/server-session-v2-reducer.ts`
-  - `src/context/server-session.ts`
-- [ ] Remove transitional session event dependencies: `session.created`, `session.updated`, `session.diff`, `session.status`, `session.idle`, and `session.error`.
-  - `src/context/global-sync/event-reducer.ts`
-  - `src/context/server-session.ts`
-  - `src/context/notification.tsx`
-  - `src/pages/session/usage-exceeded-dialogs.tsx`
-- [ ] Remove legacy message event compatibility: `message.updated`, `message.removed`, `message.part.updated`, `message.part.removed`, and `message.part.delta`.
-  - `src/context/global-sync/event-reducer.ts`
-  - `src/context/server-session.ts`
-- [x] Adapt current permission and question events to the existing request model.
-  - `src/context/global-sync/event-reducer.ts`
-  - `src/context/permission.tsx`
-- [x] Consume current file watcher events.
-  - `src/context/file.tsx`
-- [x] Consume current VCS events.
-  - `src/context/global-sync/event-reducer.ts`
-  - `src/pages/session.tsx`
-- [x] Consume current `pty.exited` events.
-  - `src/context/terminal.tsx`
-- [ ] Migrate LSP and reference events.
-  - `src/context/global-sync/event-reducer.ts`
-
-## Sessions
-
-- [x] Replace `GET /session/status` with one server-scoped `GET /api/session/active` snapshot plus V2 execution events.
-  - `src/context/server-sync.tsx`
-- [x] Migrate session listing from `GET /session`.
-  - `src/context/server-sync.tsx`
-  - `src/context/directory-sync.ts`
-  - `src/pages/layout.tsx`
-- [x] Migrate the remaining direct session read from `GET /session/:sessionID`.
-  - `src/components/titlebar.tsx`
-- [x] Migrate session updates from `PATCH /session/:sessionID`.
-  - `src/context/directory-sync.ts`
-  - `src/context/layout.tsx`
-  - `src/pages/home.tsx`
-  - `src/pages/layout.tsx`
-  - `src/pages/session/timeline/message-timeline.tsx`
-  - `src/components/titlebar-tab-nav.tsx`
-  - Renames use `POST /api/session/:sessionID/rename`; archival uses `POST /api/session/:sessionID/archive`.
-- [x] Migrate session deletion from `DELETE /session/:sessionID`.
-  - `src/pages/session/timeline/message-timeline.tsx`
-- [x] Remove session diff loading from `GET /session/:sessionID/diff`.
-  - Historical Session diffs remain unavailable until the current API defines their snapshot semantics.
-- [x] Migrate abort from `POST /session/:sessionID/abort`.
-  - `src/components/prompt-input/submit.ts`
-  - `src/pages/session/use-session-commands.tsx`
-  - `src/pages/session.tsx`
-- [x] Migrate revert and unrevert from `POST /session/:sessionID/revert` and `POST /session/:sessionID/unrevert`.
-  - `src/pages/session/use-session-commands.tsx`
-  - `src/pages/session.tsx`
-- [x] Replace `POST /session/:sessionID/summarize` with the current compact API.
-  - `src/pages/session/use-session-commands.tsx`
-- [x] Migrate slash commands from `POST /session/:sessionID/command`.
-  - `src/components/prompt-input/submit.ts`
-- [x] Migrate shell execution from `POST /session/:sessionID/shell`.
-  - `src/components/prompt-input/submit.ts`
-- [x] Migrate session fork from `POST /session/:sessionID/fork`.
-  - `src/components/dialog-fork.tsx`
-- [ ] Migrate sharing from `POST /session/:sessionID/share` and `DELETE /session/:sessionID/share`.
-  - `src/pages/session/use-session-commands.tsx`
-  - `src/pages/session/timeline/message-timeline.tsx`
-  - Blocked: the current API has no sharing contract or implementation.
-
-## Session Compatibility Fallbacks
-
-These calls are retained as fallback adapters. The current production path supplies the current session and message APIs.
-
-- [ ] Remove fallback `GET /session/:sessionID` after compatibility support is unnecessary.
-  - `src/context/server-session.ts`
-- [ ] Remove fallback `GET /session/:sessionID/message` after compatibility support is unnecessary.
-  - `src/context/server-session.ts`
-- [ ] Remove fallback `GET /session/:sessionID/message/:messageID` after compatibility support is unnecessary.
-  - `src/context/server-session.ts`
-
-## Filesystem
-
-- [ ] Migrate file listing from `GET /file`.
-  - `src/context/file.tsx`
-- [ ] Migrate file reads from `GET /file/content`.
-  - `src/context/file.tsx`
-  - `src/pages/session/review-tab.tsx`
-  - `src/pages/session/v2/review-panel-v2.tsx`
-- [x] Migrate path discovery from `GET /path` to `GET /api/path`.
-  - `src/context/global-sync/bootstrap.ts`
-  - `src/components/dialog-select-directory.tsx`
-  - `src/components/dialog-select-directory-v2.tsx`
-
-## Projects And Worktrees
-
-- [x] Migrate project listing from `GET /project` to `GET /api/project`.
-  - `src/context/global-sync/bootstrap.ts`
-- [x] Migrate the current project lookup from `GET /project/current` to `GET /api/project/current`.
-  - `src/context/global-sync/bootstrap.ts`
-- [ ] Migrate Git initialization from `POST /project/git/init`.
-  - `src/pages/session.tsx`
-- [x] Migrate project updates from `PATCH /project/:projectID` to `PATCH /api/project/:projectID`.
-  - `src/context/layout.tsx`
-  - `src/components/edit-project.ts`
-  - `src/pages/layout.tsx`
-- [ ] Migrate experimental worktree listing, creation, removal, and reset from `/experimental/worktree`.
-  - `src/pages/layout.tsx`
-  - `src/components/prompt-input/submit.ts`
-  - Listing now uses `GET /api/project/:projectID/directories`; create, removal, and reset remain.
-- [ ] Migrate instance disposal from `POST /instance/dispose`.
-  - `src/pages/layout.tsx`
-
-## VCS
-
-- [x] Migrate repository information from `GET /vcs` to `GET /api/vcs`.
-  - `src/context/global-sync/bootstrap.ts`
-- [x] Migrate diffs from `GET /vcs/diff` to `GET /api/vcs/diff`.
-  - `src/pages/session.tsx`
-- [x] Migrate status from `GET /vcs/status` to `GET /api/vcs/status`.
-  - `src/pages/layout.tsx`
-
-## Configuration And Authentication
-
-- [ ] Migrate global configuration reads from `GET /global/config`.
-  - `src/context/global-sync/bootstrap.ts`
-- [ ] Migrate directory configuration reads from `GET /config`.
-  - `src/context/global-sync/bootstrap.ts`
-- [ ] Migrate global configuration updates from `PATCH /global/config`.
-  - `src/context/server-sync.tsx`
-- [x] Migrate provider authentication method discovery from `GET /provider/auth` to `GET /api/integration/:integrationID`.
-  - `src/components/dialog-connect-provider.tsx`
-- [x] Migrate built-in provider OAuth authorization and callbacks to `/api/integration/:integrationID/connect/oauth/*`.
-  - `src/components/dialog-connect-provider.tsx`
-- [ ] Migrate remaining credentials from `PUT /auth/:providerID` and `DELETE /auth/:providerID`.
-  - Built-in provider key connections now use `POST /api/integration/:integrationID/connect/key`.
-  - `src/components/dialog-connect-provider.tsx`
-  - `src/components/dialog-custom-provider.tsx`
-  - `src/components/settings-providers.tsx`
-  - `src/components/settings-v2/providers.tsx`
-- [ ] Migrate global disposal from `POST /global/dispose`.
-  - `src/components/dialog-connect-provider.tsx`
-  - `src/components/settings-providers.tsx`
-  - `src/components/settings-v2/providers.tsx`
-
-## Permissions And Questions
-
-- [x] Migrate permission listing from `GET /permission` to `GET /api/permission/request`.
-  - `src/context/global-sync/bootstrap.ts`
-  - `src/context/permission.tsx`
-- [x] Migrate permission responses from `/session/:sessionID/permissions/:permissionID`.
-  - `src/context/permission.tsx`
-  - `src/pages/session/composer/session-composer-state.ts`
-- [x] Migrate question listing from `GET /question` to `GET /api/question/request`.
-  - `src/context/global-sync/bootstrap.ts`
-- [x] Migrate question replies and rejections from `/question/:requestID/*` to `/api/session/:sessionID/question/:requestID/*`.
-  - `src/pages/session/composer/session-question-dock.tsx`
-
-## Commands, MCP, LSP, And References
-
-- [x] Migrate command listing from `GET /command` to `GET /api/command`.
-  - `src/context/global-sync/bootstrap.ts`
-  - `src/context/server-sync.tsx`
-- [x] Migrate MCP listing, connection, and disconnection from `/mcp` to `/api/mcp`.
-  - `src/context/server-sync.tsx`
-- [ ] Replace legacy MCP authentication with the Integration OAuth workflow.
-  - `src/context/server-sync.tsx`
-- [x] Migrate experimental resource listing from `GET /experimental/resource` to `GET /api/mcp/resource`.
-  - `src/context/server-sync.tsx`
-- [ ] Migrate LSP status from `GET /lsp`.
-  - `src/context/server-sync.tsx`
-- [x] Move `GET /api/reference` off the legacy generated SDK transport.
-  - `src/context/global-sync/bootstrap.ts`
-
-## Search
-
-- [x] Migrate global session search from `GET /experimental/session` to `GET /api/session`.
-  - `src/components/command-palette.ts`
-  - `src/components/dialog-command-palette-v2.tsx`
-
-## PTY And Terminal
-
-- [x] Migrate PTY creation, reads, updates, and deletion from `/pty` to `/api/pty`.
-  - `src/context/terminal.tsx`
-  - `src/components/terminal.tsx`
-- [x] Migrate shell listing from `GET /pty/shells` to `GET /api/pty/shells`.
-  - `src/components/settings-general.tsx`
-  - `src/components/settings-v2/general.tsx`
-- [x] Migrate connection tokens from `POST /pty/:ptyID/connect-token` to `POST /api/pty/:ptyID/connect-token`.
-  - `src/components/terminal.tsx`
-- [x] Migrate the direct WebSocket connection from `/pty/:ptyID/connect` to `/api/pty/:ptyID/connect`.
-  - `src/components/terminal.tsx`
-
-## Legacy Types And Adapters
-
-These are not V1 network requests, but they keep the UI coupled to V1 data contracts.
-
-- [ ] Replace the current-session-to-legacy-session adapter.
-  - `src/utils/session.ts`
-- [ ] Replace the current-message-to-legacy-message-and-part adapter.
-  - `src/utils/session-message.ts`
-- [ ] Replace current agent, provider, and model adapters to legacy SDK structures.
-  - `src/context/global-sync/utils.ts`
-- [ ] Replace legacy `Session`, `Message`, `Part`, `PermissionRequest`, `QuestionRequest`, `Project`, `FileNode`, `FileDiffInfo`, and `Event` types throughout app state and rendering.
-- [ ] Remove the `@opencode-ai/sdk` runtime dependency after all legacy calls and types are gone.
-  - `package.json`
-
-## Test Infrastructure
-
-- [ ] Replace V1 endpoint mocks with current API mocks.
-  - `e2e/utils/mock-server.ts`
-- [x] Replace `/global/event` and `/event` interception with current event transport handling.
-  - `e2e/utils/sse-transport.ts`
-- [ ] Replace `SessionV1` and legacy SDK fixtures in timeline performance tests.
-  - `e2e/performance/timeline-stability/fixture.ts`
-- [ ] Remove remaining legacy SDK type fixtures from unit and browser tests.

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

@@ -2,17 +2,8 @@ import { base64Encode } from "@opencode-ai/core/util/encode"
 import { Event } from "@opencode-ai/schema/event"
 import { SessionStatusEvent } from "@opencode-ai/schema/session-status-event"
 import { SessionV1 } from "@opencode-ai/schema/session-v1"
-import type {
-  AssistantMessage,
-  GlobalEvent,
-  Message,
-  Part,
-  Session,
-  SessionStatus,
-  ToolPart,
-  ToolState,
-  UserMessage,
-} from "@opencode-ai/sdk/v2/client"
+import type { SessionV1Info, SessionStatus } from "@opencode-ai/client/promise"
+import type { AssistantMessage, Message, Part, ToolPart, ToolState, UserMessage } from "../../../src/types"
 import { expect, type Page } from "@playwright/test"
 import { Schema } from "effect"
 import { mockOpenCodeServer } from "../../utils/mock-server"
@@ -27,18 +18,29 @@ export const assistantID = "msg_1001_timeline_assistant"
 export const title = "Timeline visual stability"
 export const model = { providerID: "opencode", modelID: "claude-opus-4-6", variant: "max" }
 
-type TimelinePayload = Extract<
-  GlobalEvent["payload"],
-  {
-    type:
-      | "message.updated"
-      | "message.removed"
-      | "message.part.updated"
-      | "message.part.removed"
-      | "message.part.delta"
-      | "session.status"
+type Session = SessionV1Info
+type GlobalEvent = {
+  directory: string
+  project?: string
+  workspace?: string
+  payload: {
+    id: string
+    type: string
+    properties: Record<string, unknown>
   }
->
+}
+
+type TimelineProperties = {
+  "message.updated": { sessionID: string; info: Message }
+  "message.removed": { sessionID: string; messageID: string }
+  "message.part.updated": { sessionID: string; part: Part; time: number }
+  "message.part.removed": { sessionID: string; messageID: string; partID: string }
+  "message.part.delta": { sessionID: string; messageID: string; partID: string; field: string; delta: string }
+  "session.status": { sessionID: string; status: SessionStatus }
+}
+type TimelinePayload = {
+  [Type in keyof TimelineProperties]: { id: string; type: Type; properties: TimelineProperties[Type] }
+}[keyof TimelineProperties]
 
 type DeepReadonly<Value> = Value extends readonly unknown[]
   ? { readonly [Key in keyof Value]: DeepReadonly<Value[Key]> }
@@ -97,7 +99,6 @@ export async function setupTimeline(
     locale?: string
     deviceScaleFactor?: number
     seedHistory?: boolean
-    protocol?: "v1" | "v2"
   } = {},
 ) {
   const sessions = input.sessions ?? [session()]
@@ -115,7 +116,6 @@ export async function setupTimeline(
     retry: input.eventRetry ?? 20,
   })
   await mockOpenCodeServer(page, {
-    protocol: input.protocol,
     directory,
     project: project(),
     provider: provider(),
@@ -235,7 +235,7 @@ export function event(type: TimelinePayload["type"], properties: TimelinePayload
 }
 
 export function validateTimelineEvent(input: unknown): TimelineEvent {
-  return decodeEvent(input, decodeOptions)
+  return decodeEvent(input, decodeOptions) as TimelineEvent
 }
 
 export function validateTimelineMessages(input: readonly TimelineMessage[]): TimelineMessage[] {
@@ -460,7 +460,7 @@ export function toolPart(
   input: Record<string, unknown>,
   options: ToolOptions<ToolStatus> = {},
 ): Omit<ToolPart, "sessionID" | "messageID"> {
-  const base = { id, type: "tool" as const, callID: `call_${id}`, tool }
+  const base = { id, type: "tool" as const, callID: id, tool }
   if (state === "pending") return { ...base, state: { status: state, input, raw: "" } }
   if (state === "running")
     return {

packages/app/e2e/performance/unit/mock-server.test.ts

@@ -1,6 +1,121 @@
 import { expect, test } from "bun:test"
+import type { SessionMessageInfo } from "@opencode-ai/client/promise"
 import type { Page, Route } from "@playwright/test"
-import { mockOpenCodeServer } from "../../utils/mock-server"
+import { currentMessage, mockOpenCodeServer } from "../../utils/mock-server"
+
+test("preserves current messages", () => {
+  const message = {
+    id: "msg_current",
+    type: "user",
+    time: { created: 1 },
+    text: "current",
+    files: [{ data: "e30=", mime: "application/json", source: { type: "inline" } }],
+  } satisfies SessionMessageInfo
+
+  expect(currentMessage(message)).toBe(message)
+})
+
+test("converts rich legacy messages to current message types", () => {
+  expect(
+    currentMessage({
+      info: { id: "msg_user", role: "user", time: { created: 1 } },
+      parts: [
+        { type: "text", text: "Use @src/a.ts with @explore" },
+        {
+          type: "file",
+          mime: "application/json",
+          filename: "data.json",
+          url: "data:application/json;base64,e30=",
+        },
+        {
+          type: "file",
+          mime: "text/plain",
+          filename: "a.ts",
+          url: "src/a.ts",
+          source: { type: "file", text: { value: "@src/a.ts", start: 4, end: 13 } },
+        },
+        { type: "agent", name: "explore", source: { value: "@explore", start: 19, end: 27 } },
+      ],
+    }),
+  ).toEqual({
+    id: "msg_user",
+    type: "user",
+    time: { created: 1 },
+    text: "Use @src/a.ts with @explore",
+    files: [
+      { data: "e30=", mime: "application/json", name: "data.json", source: { type: "inline" } },
+      {
+        data: "",
+        mime: "text/plain",
+        name: "a.ts",
+        source: { type: "uri", uri: "src/a.ts" },
+        mention: { text: "@src/a.ts", start: 4, end: 13 },
+      },
+    ],
+    agents: [{ name: "explore", mention: { text: "@explore", start: 19, end: 27 } }],
+  })
+
+  expect(
+    currentMessage({
+      info: {
+        id: "msg_assistant",
+        role: "assistant",
+        time: { created: 2, completed: 5 },
+        agent: "explore",
+        modelID: "model",
+        providerID: "provider",
+        variant: "high",
+        cost: 0.5,
+        tokens: { input: 1, output: 2, reasoning: 3, cache: { read: 4, write: 5 } },
+        finish: "tool-calls",
+        error: { name: "MessageAbortedError", data: { message: "Stopped" } },
+      },
+      parts: [
+        { type: "text", text: "Answer" },
+        { type: "reasoning", text: "Thinking", time: { start: 2, end: 3 } },
+        {
+          id: "prt_tool",
+          callID: "call_tool",
+          type: "tool",
+          tool: "read",
+          state: {
+            status: "completed",
+            input: { filePath: "src/a.ts" },
+            output: "contents",
+            metadata: { title: "a.ts" },
+            time: { start: 3, end: 4 },
+          },
+        },
+      ],
+    }),
+  ).toEqual({
+    id: "msg_assistant",
+    type: "assistant",
+    time: { created: 2, completed: 5 },
+    agent: "explore",
+    model: { id: "model", providerID: "provider", variant: "high" },
+    cost: 0.5,
+    tokens: { input: 1, output: 2, reasoning: 3, cache: { read: 4, write: 5 } },
+    finish: "tool-calls",
+    error: { type: "MessageAbortedError", message: "Stopped" },
+    content: [
+      { type: "text", text: "Answer" },
+      { type: "reasoning", text: "Thinking", time: { created: 2, completed: 3 } },
+      {
+        type: "tool",
+        id: "call_tool",
+        name: "read",
+        time: { created: 3, ran: 3, completed: 4 },
+        state: {
+          status: "completed",
+          input: { filePath: "src/a.ts" },
+          content: [{ type: "text", text: "contents" }],
+          metadata: { title: "a.ts" },
+        },
+      },
+    ],
+  })
+})
 
 test("applies message latency after a list response gate is released", async () => {
   const events: string[] = []
@@ -30,7 +145,7 @@ test("applies message latency after a list response gate is released", async ()
   })
 
   const response = handler!({
-    request: () => ({ url: () => "http://127.0.0.1:4096/session/session/message" }),
+    request: () => ({ url: () => "http://127.0.0.1:4096/api/session/session/message" }),
     fulfill: () => {
       events.push("fulfill")
       return Promise.resolve()

packages/app/e2e/regression/cross-server-tab-close.spec.ts

@@ -85,21 +85,17 @@ async function mockServers(page: Page, requests: string[]) {
     const current = url.origin === serverA ? sessionA : sessionB
     const directory = url.searchParams.get("directory")
     if (directory && directory !== current.directory) return json(route, { name: "InvalidDirectory" }, 500)
-    if (url.pathname === "/global/event" || url.pathname === "/event" || url.pathname === "/api/event")
+    if (url.pathname === "/api/event")
       return sse(route)
-    if (url.pathname === "/global/health") return json(route, {}, 404)
     if (url.pathname === "/api/health") return json(route, { pid: 1 })
     if (url.pathname === "/api/session") return json(route, { data: [currentSession(current)], cursor: {} })
     if (url.pathname === "/api/session/active") return json(route, { data: {} })
     if (url.pathname === `/api/session/${current.id}`) return json(route, { data: currentSession(current) })
     if (url.pathname === `/api/session/${current.id}/message`) return json(route, { data: [], cursor: {} })
-    if (url.pathname === `/session/${current.id}`) return json(route, current)
     if (/^\/session\/[^/]+$/.test(url.pathname)) return json(route, { name: "NotFoundError" }, 404)
-    if (url.pathname === `/session/${current.id}/message`) return json(route, [])
     if (/^\/session\/[^/]+\/(children|todo|diff)$/.test(url.pathname)) return json(route, [])
     if (["/skill", "/command", "/lsp", "/formatter", "/permission", "/question", "/vcs/diff"].includes(url.pathname))
       return json(route, [])
-    if (["/global/config", "/config", "/provider/auth", "/mcp"].includes(url.pathname)) return json(route, {})
     if (url.pathname === "/provider")
       return json(route, { all: [], connected: [], default: { providerID: "", modelID: "" } })
     if (url.pathname === "/agent") return json(route, [{ name: "build", mode: "primary" }])

packages/app/e2e/regression/remote-session-settings.spec.ts

@@ -3,7 +3,7 @@ import { expect, test, type Page, type Route } from "@playwright/test"
 import { installSseTransport } from "../utils/sse-transport"
 import { currentSession } from "../utils/mock-server"
 
-const serverA = "http://127.0.0.1:4096"
+const serverA = `http://127.0.0.1:${process.env.PLAYWRIGHT_SERVER_PORT ?? "4096"}`
 const serverB = "http://127.0.0.1:4097"
 const directoryA = "C:/server-a"
 const directoryB = "/home/server-b"
@@ -32,7 +32,7 @@ test("session settings use the remote server context", async ({ page }) => {
     .poll(() =>
       permissionRequests.some((request) => {
         const url = new URL(request)
-        return url.origin === serverB && url.searchParams.get("directory") === directoryB
+        return url.origin === serverB && url.searchParams.get("location[directory]") === directoryB
       }),
     )
     .toBe(true)
@@ -67,7 +67,7 @@ test("auto-accept responds for an unfocused server session", async ({ page }) =>
     .poll(() =>
       permissionRequests.some((request) => {
         const url = new URL(request)
-        return url.origin === serverA && url.searchParams.get("directory") === directoryA
+        return url.origin === serverA && url.searchParams.get("location[directory]") === directoryA
       }),
     )
     .toBe(true)
@@ -99,10 +99,10 @@ test("auto-accept responds for an unfocused server session", async ({ page }) =>
     .toEqual([
       {
         origin: serverA,
-        directory: directoryA,
+        directory: undefined,
         sessionID: sessionA.id,
         permissionID: "permission-background-a",
-        body: { response: "once" },
+        body: { reply: "once" },
       },
     ])
 
@@ -127,17 +127,17 @@ test("auto-accept responds for an unfocused server session", async ({ page }) =>
     .toEqual([
       {
         origin: serverA,
-        directory: directoryA,
+        directory: undefined,
         sessionID: sessionA.id,
         permissionID: "permission-background-a",
-        body: { response: "once" },
+        body: { reply: "once" },
       },
       {
         origin: serverA,
-        directory: directoryA,
+        directory: undefined,
         sessionID: childSessionA.id,
         permissionID: "permission-background-a-child",
-        body: { response: "once" },
+        body: { reply: "once" },
       },
     ])
 })
@@ -168,8 +168,8 @@ async function mockServers(page: Page, permissionRequests: string[], permissionR
     const remote = url.origin === serverB
     const directory = remote ? directoryB : directoryA
     const sessions = remote ? [sessionB] : [sessionA, childSessionA]
-    const requestDirectory = url.searchParams.get("directory")
-    const response = url.pathname.match(/^\/session\/([^/]+)\/permissions\/([^/]+)$/)
+    const requestDirectory = url.searchParams.get("location[directory]")
+    const response = url.pathname.match(/^\/api\/session\/([^/]+)\/permission\/([^/]+)\/reply$/)
     if (route.request().method() === "POST" && response) {
       permissionResponses.push({
         origin: url.origin,
@@ -181,13 +181,21 @@ async function mockServers(page: Page, permissionRequests: string[], permissionR
       return json(route, true)
     }
     if (requestDirectory && requestDirectory !== directory) return json(route, { name: "InvalidDirectory" }, 500)
-    if (url.pathname === "/global/event" || url.pathname === "/event" || url.pathname === "/api/event")
+    if (url.pathname === "/api/event")
       return sse(route)
-    if (url.pathname === "/global/health") return json(route, { healthy: true })
-    if (url.pathname === "/api/provider" || url.pathname === "/api/model" || url.pathname === "/api/agent")
-      return json(route, { data: [] })
-    if (url.pathname === "/api/model/default") return json(route, { data: null })
-    if (["/api/command", "/api/reference", "/api/permission/request", "/api/question/request"].includes(url.pathname))
+    if (url.pathname === "/api/provider")
+      return json(route, {
+        location: { directory },
+        data: [{ id: remote ? "server-b" : "server-a", name: remote ? "Server B Provider" : "Server A Provider", package: "test" }],
+      })
+    if (url.pathname === "/api/model") return json(route, { location: { directory }, data: [model(remote)] })
+    if (url.pathname === "/api/model/default") return json(route, { location: { directory }, data: model(remote) })
+    if (url.pathname === "/api/agent") return json(route, { location: { directory }, data: [] })
+    if (url.pathname === "/api/permission/request") {
+      permissionRequests.push(url.toString())
+      return json(route, { location: { directory }, data: [] })
+    }
+    if (["/api/command", "/api/reference", "/api/question/request"].includes(url.pathname))
       return json(route, { location: { directory }, data: [] })
     if (url.pathname === "/api/mcp") return json(route, { location: { directory }, data: [] })
     if (url.pathname === "/api/mcp/resource")
@@ -211,8 +219,6 @@ async function mockServers(page: Page, permissionRequests: string[], permissionR
     if (currentSessionInfo) return json(route, { data: currentSession(currentSessionInfo) })
     if (sessions.some((session) => url.pathname === `/api/session/${session.id}/message`))
       return json(route, { data: [], cursor: {} })
-    const current = sessions.find((session) => url.pathname === `/session/${session.id}`)
-    if (current) return json(route, current)
     if (/^\/session\/[^/]+$/.test(url.pathname)) return json(route, { name: "NotFoundError" }, 404)
     if (/^\/session\/[^/]+\/message$/.test(url.pathname)) return json(route, [])
     if (/^\/session\/[^/]+\/(children|todo|diff)$/.test(url.pathname)) return json(route, [])
@@ -222,7 +228,6 @@ async function mockServers(page: Page, permissionRequests: string[], permissionR
     }
     if (["/skill", "/command", "/lsp", "/formatter", "/question", "/vcs/diff", "/pty/shells"].includes(url.pathname))
       return json(route, [])
-    if (["/global/config", "/config", "/provider/auth", "/mcp"].includes(url.pathname)) return json(route, {})
     if (url.pathname === "/provider") return json(route, provider(remote ? "server-b" : "server-a"))
     if (url.pathname === "/agent") return json(route, [{ name: "build", mode: "primary" }])
     if (url.pathname === "/project" || url.pathname === "/project/current") {
@@ -288,6 +293,25 @@ function provider(id: string) {
   }
 }
 
+function model(remote: boolean) {
+  const id = remote ? "server-b" : "server-a"
+  const name = remote ? "Server B" : "Server A"
+  return {
+    id,
+    modelID: id,
+    providerID: id,
+    name: `${name} Model`,
+    family: id,
+    capabilities: { tools: true, input: ["text"], output: ["text"] },
+    variants: [],
+    time: { released: Date.now() },
+    cost: [{ input: 0, output: 0, cache: { read: 0, write: 0 } }],
+    status: "active",
+    enabled: true,
+    limit: { context: 200_000, output: 32_000 },
+  }
+}
+
 function json(route: Route, body: unknown, status = 200) {
   return route.fulfill({
     status,

packages/app/e2e/regression/remote-tab-busy.spec.ts

@@ -58,22 +58,18 @@ async function mockServers(page: Page) {
     const current = url.origin === serverA ? sessionA : sessionB
     const directory = url.searchParams.get("directory")
     if (directory && directory !== current.directory) return json(route, { name: "InvalidDirectory" }, 500)
-    if (url.pathname === "/global/event" || url.pathname === "/event" || url.pathname === "/api/event")
+    if (url.pathname === "/api/event")
       return sse(route, url.pathname === "/api/event")
-    if (url.pathname === "/global/health") return json(route, {}, 404)
     if (url.pathname === "/api/health") return json(route, { pid: 1 })
     if (url.pathname === "/api/session/active")
       return json(route, { data: url.origin === serverB ? { [sessionB.id]: { type: "running" } } : {} })
     if (url.pathname === "/api/session") return json(route, { data: [currentSession(current)], cursor: {} })
     if (url.pathname === `/api/session/${current.id}`) return json(route, { data: currentSession(current) })
     if (url.pathname === `/api/session/${current.id}/message`) return json(route, { data: [], cursor: {} })
-    if (url.pathname === `/session/${current.id}`) return json(route, current)
     if (/^\/session\/[^/]+$/.test(url.pathname)) return json(route, { name: "NotFoundError" }, 404)
-    if (url.pathname === `/session/${current.id}/message`) return json(route, [])
     if (/^\/session\/[^/]+\/(children|todo|diff)$/.test(url.pathname)) return json(route, [])
     if (["/skill", "/command", "/lsp", "/formatter", "/permission", "/question", "/vcs/diff"].includes(url.pathname))
       return json(route, [])
-    if (["/global/config", "/config", "/provider/auth", "/mcp"].includes(url.pathname)) return json(route, {})
     if (url.pathname === "/provider")
       return json(route, { all: [], connected: [], default: { providerID: "", modelID: "" } })
     if (url.pathname === "/agent") return json(route, [{ name: "build", mode: "primary" }])

packages/app/e2e/regression/review-open-file.spec.ts

@@ -127,7 +127,7 @@ test("opens and searches project files inline", async ({ page }) => {
   await expect(panel.getByRole("tab", { name: "nested.ts" })).toHaveAttribute("data-selected", "")
   await expect(sidebarToggle).toBeEnabled()
   await expect(panel.getByText("contents:src/nested.ts", { exact: true })).toBeVisible()
-  expect(searches).toContainEqual({ query: "nested", dirs: "false", limit: 200 })
+  expect(searches).toContainEqual({ query: "nested", dirs: "file", limit: 200 })
 
   await panel.getByRole("button", { name: "Open file" }).click()
   await expect(panel.getByRole("tab", { name: "nested.ts" })).toHaveCount(1)

packages/app/e2e/regression/review-state-persistence.spec.ts

@@ -19,36 +19,25 @@ test("restores review mode and selected file per session", async ({ page }) => {
   await expectSessionTitle(page, titleA)
   await page.getByRole("button", { name: "Toggle review" }).click()
 
-  await selectMode(page, "Git changes", "Branch changes")
-  await selectFile(page, "beta.ts")
+  await selectFile(page, "alpha.ts")
 
   await switchSession(page, titleB)
   await expect(page.getByRole("button", { name: "Git changes" })).toBeVisible()
   await selectFile(page, "gamma.ts")
 
   await switchSession(page, titleA)
-  await expect(page.getByRole("button", { name: "Branch changes" })).toBeVisible()
-  await expectSelectedFile(page, "beta.ts")
-  await selectMode(page, "Branch changes", "Git changes")
   await expectSelectedFile(page, "alpha.ts")
-  await selectMode(page, "Git changes", "Branch changes")
-  await expectSelectedFile(page, "beta.ts")
 
   await page.reload()
   await expectSessionTitle(page, titleA)
-  await expect(page.getByRole("button", { name: "Branch changes" })).toBeVisible()
-  await expectSelectedFile(page, "beta.ts")
+  await expect(page.getByRole("button", { name: "Git changes" })).toBeVisible()
+  await expectSelectedFile(page, "alpha.ts")
 
   await switchSession(page, titleB)
   await expect(page.getByRole("button", { name: "Git changes" })).toBeVisible()
   await expectSelectedFile(page, "gamma.ts")
 })
 
-async function selectMode(page: Page, current: string, next: string) {
-  await page.getByRole("button", { name: current }).click()
-  await page.getByRole("option", { name: next }).dispatchEvent("click")
-}
-
 async function selectFile(page: Page, file: string) {
   await page.getByRole("button", { name: file }).click()
   await expectSelectedFile(page, file)
@@ -65,7 +54,6 @@ async function switchSession(page: Page, title: string) {
 
 async function setup(page: Page) {
   await mockOpenCodeServer(page, {
-    protocol: "v1",
     directory,
     project: {
       id: projectID,
@@ -89,22 +77,27 @@ async function setup(page: Page) {
     sessions: [session(sessionA, titleA, 1700000000000), session(sessionB, titleB, 1700000001000)],
     pageMessages: () => ({ items: [] }),
   })
-  await page.route(/\/vcs(?:\?.*)?$/, (route) =>
+  await page.route(/\/api\/vcs(?:\?.*)?$/, (route) =>
     route.fulfill({
       status: 200,
       contentType: "application/json",
-      body: JSON.stringify({ branch: "feature", default_branch: "dev" }),
+      body: JSON.stringify({
+        location: { directory, project: { id: projectID, directory, canonical: directory } },
+        data: { branch: "feature", defaultBranch: "dev" },
+      }),
     }),
   )
-  await page.route("**/vcs/diff**", (route) =>
+  await page.route("**/api/vcs/diff**", (route) =>
     route.fulfill({
       status: 200,
       contentType: "application/json",
-      body: JSON.stringify(
-        new URL(route.request().url()).searchParams.get("mode") === "branch"
-          ? [diff("src/alpha.ts"), diff("src/beta.ts")]
-          : [diff("src/alpha.ts"), diff("src/gamma.ts")],
-      ),
+      body: JSON.stringify({
+        location: { directory, project: { id: projectID, directory, canonical: directory } },
+        data:
+          new URL(route.request().url()).searchParams.get("mode") === "branch"
+            ? [diff("src/alpha.ts"), diff("src/beta.ts")]
+            : [diff("src/alpha.ts"), diff("src/gamma.ts")],
+      }),
     }),
   )
   await page.addInitScript(

packages/app/e2e/regression/review-terminal-stacked.spec.ts

@@ -25,7 +25,6 @@ test("keeps the review tree and terminal sized when both panels are open", async
   let detailFailures = 1
   await page.setViewportSize({ width: 1400, height: 900 })
   await mockOpenCodeServer(page, {
-    protocol: "v1",
     directory,
     project: {
       id: projectID,
@@ -62,33 +61,32 @@ test("keeps the review tree and terminal sized when both panels are open", async
     events: () => events.splice(0, 1),
     eventRetry: 16,
   })
-  await page.route(/\/vcs(?:\?.*)?$/, (route) =>
+  await page.route(/\/api\/vcs(?:\?.*)?$/, (route) =>
     route.fulfill({
       status: 200,
       contentType: "application/json",
       body: JSON.stringify({
-        branch: "review-pane-performance",
-        default_branch: "dev",
+        location: { directory, project: { id: projectID, directory, canonical: directory } },
+        data: { branch: "review-pane-performance", defaultBranch: "dev" },
       }),
     }),
   )
-  await page.route("**/vcs/diff**", (route) => {
+  await page.route("**/api/vcs/diff**", (route) => {
     const url = new URL(route.request().url())
-    const scope = url.searchParams.get("directory")?.replaceAll("\\", "/")
+    const scope = url.searchParams.get("location[directory]")?.replaceAll("\\", "/")
     const detail = scope?.endsWith("/src/branch/d00027")
     if (detail && detailFailures-- > 0) return route.fulfill({ status: 500, body: "retry detail" })
     return route.fulfill({
       status: 200,
       contentType: "application/json",
-      body: JSON.stringify(
-        url.searchParams.get("mode") === "branch"
-          ? detail
-            ? branchDiffs
-                .filter((diff) => diff.file.startsWith("src/branch/d00027/"))
-                .map((diff) => fileDiff(diff.file, diff.additions, true, detailVersion))
-            : branchDiffs
-          : Array.from({ length: 7 }, (_, index) => fileDiff(`src/git-${index}.ts`, 1)),
-      ),
+      body: JSON.stringify({
+        location: { directory, project: { id: projectID, directory, canonical: directory } },
+        data: detail
+          ? branchDiffs
+              .filter((diff) => diff.file.startsWith("src/branch/d00027/"))
+              .map((diff) => fileDiff(diff.file, diff.additions, true, detailVersion))
+          : branchDiffs,
+      }),
     })
   })
   await page.route("**/pty*", (route) =>
@@ -109,7 +107,7 @@ test("keeps the review tree and terminal sized when both panels are open", async
       }),
     }),
   )
-  await page.route("**/pty/pty_review_terminal*", (route) =>
+  await page.route("**/api/pty/pty_review_terminal*", (route) =>
     route.fulfill({
       status: 200,
       contentType: "application/json",
@@ -127,7 +125,7 @@ test("keeps the review tree and terminal sized when both panels are open", async
       }),
     }),
   )
-  await page.route("**/pty/pty_review_terminal/connect-token*", (route) =>
+  await page.route("**/api/pty/pty_review_terminal/connect-token*", (route) =>
     route.fulfill({
       status: 200,
       contentType: "application/json",
@@ -137,7 +135,7 @@ test("keeps the review tree and terminal sized when both panels are open", async
       }),
     }),
   )
-  await page.routeWebSocket("**/pty/pty_review_terminal/connect", () => undefined)
+  await page.routeWebSocket("**/api/pty/pty_review_terminal/connect", () => undefined)
   await page.addInitScript(() => {
     localStorage.setItem("settings.v3", JSON.stringify({ general: { newLayoutDesigns: true } }))
     localStorage.setItem(
@@ -149,9 +147,7 @@ test("keeps the review tree and terminal sized when both panels are open", async
   await page.goto(`/${base64Encode(directory)}/session/${sessionID}`)
   await expectSessionTitle(page, title)
   await expect(page.locator("#review-panel")).toBeVisible()
-  await expectTree(page, 8, "git-0.ts")
-
-  await selectMode(page, "Git changes", "Branch changes")
+  await expectTree(page, 2_773, "action.yml")
   await expect(page.locator("#session-side-panel-review-tab")).toHaveText("Files Changed 2740")
   await page.keyboard.press("Control+Backquote")
   await expect(page.locator("#terminal-panel")).toBeVisible()
@@ -174,9 +170,9 @@ test("keeps the review tree and terminal sized when both panels are open", async
   expect(bottomGap).toBeLessThanOrEqual(16)
   const lazyDiff = page.waitForRequest((request) => {
     const url = new URL(request.url())
-    return (
-      url.pathname === "/vcs/diff" &&
-      url.searchParams.get("directory")?.replaceAll("\\", "/").endsWith("/src/branch/d00027") === true
+      return (
+        url.pathname === "/api/vcs/diff" &&
+      url.searchParams.get("location[directory]")?.replaceAll("\\", "/").endsWith("/src/branch/d00027") === true
     )
   })
   await lastFile.click()
@@ -190,59 +186,46 @@ test("keeps the review tree and terminal sized when both panels are open", async
   const refreshedDiff = page.waitForRequest((request) => {
     const url = new URL(request.url())
     return (
-      url.pathname === "/vcs/diff" &&
-      url.searchParams.get("directory")?.replaceAll("\\", "/").endsWith("/src/branch/d00027") === true
+      url.pathname === "/api/vcs/diff" &&
+      url.searchParams.get("location[directory]")?.replaceAll("\\", "/").endsWith("/src/branch/d00027") === true
     )
   })
   sessionStatus[sessionID] = { type: "idle" }
   events.push(statusEvent("idle"))
   await refreshedDiff
   await expect(preview).toContainText("after-2")
-  await selectMode(page, "Branch changes", "Git changes")
-  await expectTree(page, 8, "git-0.ts")
-  await page.getByRole("button", { name: "git-0.ts" }).click()
-  await selectMode(page, "Git changes", "Branch changes")
-  await expectTree(page, 2_773, "action.yml")
-
   const filter = page.getByRole("searchbox", { name: "Filter files" })
   await filter.fill("generated-2738")
   await expectTree(page, 1, "generated-2738.ts")
   await filter.fill("")
-  await expectTree(page, 2_773, "action.yml")
+  await expectTree(page, 2_773, "generated-2738.ts")
 
   await page.getByRole("button", { name: "Toggle file tree" }).click()
   await expect(page.locator('[data-slot="session-review-v2-sidebar"]')).toHaveCount(0)
   await expect(page.locator('#review-panel [data-component="file-tree-v2"]')).toHaveCount(0)
   await page.getByRole("button", { name: "Toggle file tree" }).click()
-  await expectTree(page, 2_773, "action.yml")
+  await expectTree(page, 2_773, "generated-2738.ts")
 
   await page.keyboard.press("Control+Backquote")
   await expect(page.locator("#terminal-panel")).toHaveCount(0)
-  await expectTree(page, 2_773, "action.yml")
+  await expectTree(page, 2_773, "generated-2738.ts")
   await page.keyboard.press("Control+Backquote")
   await expect(page.locator("#terminal-panel")).toBeVisible()
-  await expectTree(page, 2_773, "action.yml")
+  await expectTree(page, 2_773, "generated-2738.ts")
 
   await page.getByRole("button", { name: "Toggle review" }).click()
   await expect(page.locator("#review-panel")).toHaveCount(0)
   await page.getByRole("button", { name: "Toggle review" }).click()
-  await expectTree(page, 2_773, "action.yml")
+  await expectTree(page, 2_773, "generated-2738.ts")
   await page.setViewportSize({ width: 1_000, height: 700 })
-  await expectTree(page, 2_773, "action.yml")
+  await expectTree(page, 2_773, "generated-2738.ts")
   await expectStackGeometry(page)
   await page.setViewportSize({ width: 1_000, height: 120 })
   await page.setViewportSize({ width: 1_400, height: 900 })
-  await expectTree(page, 2_773, "action.yml")
+  await expectTree(page, 2_773, "generated-2738.ts")
   await expectStackGeometry(page)
 })
 
-async function selectMode(page: Page, current: string, next: string) {
-  await page.getByRole("button", { name: current }).click()
-  const option = page.getByRole("option", { name: next })
-  await expect(option).toBeVisible()
-  await option.click()
-}
-
 async function expectTree(page: Page, total: number, file: string) {
   await expectMountedTree(page, total)
   await expect(page.getByRole("button", { name: file })).toBeVisible()

packages/app/e2e/regression/session-request-docks.spec.ts

@@ -173,7 +173,6 @@ async function mockServer(
   },
 ) {
   await mockOpenCodeServer(page, {
-    protocol: "v2",
     directory,
     project: {
       id: projectID,

packages/app/e2e/regression/session-timeline-collapse-state.spec.ts

@@ -52,7 +52,7 @@ const editPart = {
   sessionID,
   messageID: assistantMessageID,
   type: "tool",
-  callID: "call_edit_regression",
+  callID: editPartID,
   tool: "edit",
   state: {
     status: "completed",

packages/app/e2e/regression/session-timeline-context-resize.spec.ts

@@ -14,8 +14,8 @@ const projectID = "proj_context_resize_regression"
 const sessionID = "ses_context_resize_regression"
 const title = "Context resize regression"
 const model = { providerID: "opencode", modelID: "claude-opus-4-6", variant: "max" }
-const contextIDs = ["prt_0100_read", "prt_0101_glob", "prt_0102_grep", "prt_0103_list"]
-const followingTextID = "prt_0104_text"
+const contextIDs = ["ctx_0100_read", "ctx_0101_glob", "ctx_0102_grep", "ctx_0103_list"]
+const followingTextID = `${id("msg_assistant", 10)}:text:0`
 
 type Message = {
   info: Record<string, unknown> & { id: string; role: "user" | "assistant" }
@@ -263,7 +263,7 @@ function turn(index: number, target: boolean, status: "running" | "completed" =
             ),
             contextTool(contextIDs[3]!, assistantID, "list", { path: "src" }, status),
             {
-              id: followingTextID,
+              id: "prt_0104_text",
               sessionID,
               messageID: assistantID,
               type: "text",
@@ -295,7 +295,7 @@ function contextTool(
     sessionID,
     messageID,
     type: "tool",
-    callID: `call_${partID}`,
+    callID: partID,
     tool,
     state: {
       status,

packages/app/e2e/regression/session-timeline-history-root.spec.ts

@@ -10,7 +10,6 @@ import {
   status,
   textPart,
   title,
-  userID,
   userMessage,
 } from "../performance/timeline-stability/fixture"
 import { mockOpenCodeServer } from "../utils/mock-server"
@@ -19,18 +18,22 @@ import { expectSessionTitle } from "../utils/waits"
 
 const initialPageSize = 20
 const historyPageSize = 200
-const assistants = Array.from({ length: initialPageSize + 1 }, (_, index) =>
-  assistantMessage([textPart(`prt_history_root_${index}`, `Assistant response ${index}`)], {
-    id: `msg_${String(index + 1001).padStart(4, "0")}_history_root_assistant`,
-    parentID: userID,
-    created: 1700000001000 + index * 1_000,
-    completed: index < initialPageSize,
-  }),
-)
-const messages = [userMessage(), ...assistants]
+const messages = Array.from({ length: initialPageSize + 1 }, (_, index) => {
+  const id = `msg_${String(index + 1001).padStart(4, "0")}_history_root_user`
+  return [
+    userMessage(undefined, { id, created: 1700000000000 + index * 2_000 }),
+    assistantMessage([textPart(`prt_history_root_${index}`, `Assistant response ${index}`)], {
+      id: `msg_${String(index + 1001).padStart(4, "0")}_history_root_assistant`,
+      parentID: id,
+      created: 1700000001000 + index * 2_000,
+      completed: index < initialPageSize,
+    }),
+  ]
+}).flat()
+const assistants = messages.filter((message) => message.info.role === "assistant")
 const lastAssistant = assistants.at(-1)!
-const lastPartID = assistants.at(-1)!.parts[0]!.id
-const userPartID = `prt_${userID}_text`
+const lastPartID = `${assistants.at(-1)!.info.id}:text:0`
+const userPartID = `${messages.at(-2)!.info.id}:text:0`
 const completed = {
   ...lastAssistant.info,
   time: { ...lastAssistant.info.time, completed: lastAssistant.info.time.created + 15_000 },
@@ -154,15 +157,23 @@ for (const scenario of scenarios) {
     await expectSessionTitle(page, title)
     await expect(page.locator(`[data-timeline-part-id="${lastPartID}"]`)).toBeVisible()
     await expect(page.locator(`[data-timeline-part-id="${userPartID}"]`)).toBeVisible()
-    await expect.poll(() => requests.filter((request) => request.phase === "start").length).toBe(2)
+    const viewport = page.locator(".scroll-view__viewport", { has: page.locator("[data-timeline-row]") })
+    await viewport.hover()
+    const deadline = Date.now() + 10_000
+    while (requests.filter((request) => request.phase === "start").length < 2) {
+      if (Date.now() >= deadline) throw new Error("Timed out scrolling to the history boundary")
+      await page.mouse.wheel(0, -240)
+      await page.waitForTimeout(20)
+    }
     expect(requests.filter((request) => request.phase === "end")).toHaveLength(1)
-    expect(sequence.slice(0, 4)).toEqual([
+    expect(sequence.slice(0, 3)).toEqual([
       "messages:start:latest",
       "messages:end:latest",
-      `message:${userID}`,
       `messages:start:${messages.at(-initialPageSize)!.info.id}`,
     ])
-    await expect(page.locator('[data-timeline-part-id^="prt_history_root_"]')).toHaveCount(initialPageSize)
+    await expect(page.locator('[data-timeline-part-id*="_history_root_assistant:text:0"]')).toHaveCount(
+      initialPageSize / 2,
+    )
     await page.evaluate(() => {
       ;(
         window as Window & {
@@ -174,7 +185,9 @@ for (const scenario of scenarios) {
     expect(await visibleContentHidden(page)).toBe(false)
     const beforeHistory = await probeSamples(page)
     history.resolve()
-    await expect(page.locator('[data-timeline-part-id^="prt_history_root_"]')).toHaveCount(assistants.length)
+    await expect
+      .poll(() => page.locator('[data-timeline-part-id*="_history_root_assistant:text:0"]').count())
+      .toBeGreaterThan(initialPageSize / 2)
     await expect.poll(() => requests.filter((request) => request.phase === "end").length).toBe(2)
     await expect(page.getByRole("button", { name: "Stop" })).toBeVisible()
     await waitForProbeSamples(page, beforeHistory)
@@ -182,7 +195,7 @@ for (const scenario of scenarios) {
       { before: undefined, limit: initialPageSize },
       { before: messages.at(-initialPageSize)!.info.id, limit: historyPageSize },
     ])
-    expect(roots).toEqual([{ sessionID, messageID: userID }])
+    expect(roots).toEqual([])
 
     const message = messageUpdated(scenario.info)
     const idle = status("idle")

packages/app/e2e/regression/session-timeline-lifecycle-state.spec.ts

@@ -103,7 +103,7 @@ test("moves busy through retry and recovery to final idle content", async ({ pag
   await timeline.send(status("idle"), 350)
   await expect(page.locator('[data-timeline-row="Retry"]')).toHaveCount(0)
   await expect(page.locator('[data-timeline-row="Thinking"]')).toHaveCount(0)
-  await expect(page.locator('[data-timeline-row="DiffSummary"]')).toBeVisible()
+  await expect(page.locator('[data-timeline-part-id="prt_recovered"]')).toContainText("Recovered response")
 })
 
 function lines(count: number) {

packages/app/e2e/regression/session-timeline-projection.spec.ts

@@ -89,7 +89,6 @@ test.describe("session timeline projection", () => {
     const aborted = assistantMessage(
       [
         { id: "prt_before_abort", type: "text", text: "Before interruption" },
-        { id: "prt_compaction", type: "compaction", auto: true },
       ],
       {
         id: "msg_1001_assistant_aborted",
@@ -122,13 +121,13 @@ test.describe("session timeline projection", () => {
     await scroller.evaluate((element) => (element.scrollTop = 0))
 
     await expect(page.locator('[data-timeline-row="TurnDivider"]')).toHaveCount(1)
-    await expect(page.getByText("Session compacted", { exact: true })).toBeVisible()
+    await expect(page.getByText("Before interruption", { exact: true })).toBeVisible()
     await expect(page.getByText("Visible provider failure")).toBeVisible()
     await scroller.evaluate((element) => (element.scrollTop = element.scrollHeight))
     await expect(page.locator('[data-timeline-row="TurnGap"]')).toBeVisible()
   })
 
-  test("renders comment strips and historical diff summary overflow", async ({ page }) => {
+  test("renders legacy synthetic comments as ordinary V2 user text", async ({ page }) => {
     const user = userMessage(
       [
         userText("The user made the following comment regarding lines 4 through 8 of src/a.ts: Keep this stable", {
@@ -159,10 +158,14 @@ test.describe("session timeline projection", () => {
     const scroller = page.locator(".scroll-view__viewport", { has: page.locator("[data-timeline-row]") })
     await scroller.evaluate((element) => (element.scrollTop = 0))
 
-    await expect(page.locator('[data-timeline-row="CommentStrip"]')).toBeVisible()
-    await expect(page.getByText("Keep this stable", { exact: true })).toBeVisible()
-    await expect(page.locator('[data-timeline-row="DiffSummary"]')).toBeVisible()
-    await expect(page.getByText(/show all/i)).toBeVisible()
+    await expect(
+      page.getByText(
+        "The user made the following comment regarding lines 4 through 8 of src/a.ts: Keep this stable Continue after the comment",
+        { exact: true },
+      ),
+    ).toBeVisible()
+    await expect(page.locator('[data-timeline-row="CommentStrip"]')).toHaveCount(0)
+    await expect(page.locator('[data-timeline-row="DiffSummary"]')).toHaveCount(0)
   })
 
   test("renders interruption independently when the turn is not compacted", async ({ page }) => {

packages/app/e2e/regression/session-timeline-reasoning-projection.spec.ts

@@ -1,5 +1,6 @@
 import { expect, test } from "@playwright/test"
 import {
+  assistantID,
   assistantMessage,
   reasoningPart,
   setupTimeline,
@@ -70,7 +71,7 @@ for (const profile of profiles) {
     await timeline.send(status("busy"), 150)
 
     await expect(page.locator('[data-timeline-row="Thinking"]')).toHaveCount(profile.thinking ? 1 : 0)
-    await expect(page.locator(`[data-timeline-part-id="${reasoningID}"]`)).toHaveCount(profile.body ? 1 : 0)
+    await expect(page.locator(`[data-timeline-part-id="${assistantID}:reasoning:0"]`)).toHaveCount(profile.body ? 1 : 0)
     if (!profile.summaries && profile.reasoning.trim()) {
       await expect(page.getByText("Inspecting stability", { exact: true })).toBeVisible()
     }
@@ -89,5 +90,5 @@ test("does not infer reasoning visibility from provider identity", async ({ page
 
   await expect(page.locator('[data-timeline-row="Thinking"]')).toHaveCount(0)
   await expect(page.locator('[data-timeline-part-id*="reasoning"]')).toHaveCount(0)
-  await expect(page.locator('[data-timeline-part-id="prt_provider_text"]')).toBeVisible()
+  await expect(page.locator(`[data-timeline-part-id="${assistantID}:text:0"]`)).toBeVisible()
 })

packages/app/e2e/regression/session-timeline-reducer-projection.spec.ts

@@ -23,10 +23,11 @@ test("groups singleton and separated context operations at correct boundaries",
   ]
   await setupTimeline(page, { messages: [userMessage(), assistantMessage(parts)] })
 
-  await expect(page.locator('[data-timeline-part-ids="prt_boundary_01_read"]')).toBeVisible()
-  await expect(page.locator('[data-timeline-part-ids="prt_boundary_03_glob,prt_boundary_04_grep"]')).toBeVisible()
+  await expect(
+    page.locator('[data-timeline-part-ids="prt_boundary_01_read,prt_boundary_03_glob,prt_boundary_04_grep"]'),
+  ).toBeVisible()
   await expect(page.locator('[data-timeline-part-ids="prt_boundary_06_list"]')).toBeVisible()
-  await expect(page.locator('[data-timeline-row="AssistantPart"]')).toHaveCount(5)
+  await expect(page.locator('[data-timeline-row="AssistantPart"]')).toHaveCount(4)
 })
 
 test("reducer-hardening: converges when idle arrives before final part and message completion", async ({ page }) => {

packages/app/e2e/regression/session-timeline-shell-outline.spec.ts

@@ -145,7 +145,6 @@ test("allows paint rounding for every framed row but not fixed turn gaps", async
       }),
     ],
   })
-  await expect(page.locator('[data-timeline-row="DiffSummary"]')).toBeVisible()
   await expect(page.locator('[data-timeline-row="TurnGap"]')).toBeVisible()
 
   const rows = await page.locator("[data-timeline-key]").evaluateAll((elements) =>

packages/app/e2e/regression/session-timeline-transport.spec.ts

@@ -90,7 +90,7 @@ test("reconnects after a stream error", async ({ page }) => {
 })
 
 test("does not request replay when reconnecting the volatile V2 event stream", async ({ page }) => {
-  const timeline = await setupTimeline(page, { eventRetry: 10, protocol: "v2" })
+  const timeline = await setupTimeline(page, { eventRetry: 10 })
   const first = await timeline.transport.send(partUpdated(textPart("prt_transport_id", "event with id")), {
     id: "timeline-event-7",
   })
@@ -107,10 +107,10 @@ test("passes through non-event fetches", async ({ page }) => {
   const timeline = await setupTimeline(page)
 
   const health = await page.evaluate(async () => {
-    const response = await fetch("/global/health")
+    const response = await fetch("/api/health")
     return response.json()
   })
 
-  expect(health).toEqual({ healthy: true })
+  expect(health).toEqual({ healthy: true, version: "2.0.0", pid: 1 })
   expect(await timeline.transport.connections()).toHaveLength(1)
 })

packages/app/e2e/regression/session-todo-dock-navigation.spec.ts

@@ -23,7 +23,7 @@ type EventPayload = {
 
 test.use({ viewport: { width: 1440, height: 900 }, reducedMotion: "no-preference" })
 
-test("animates todo lifecycle without replaying it across session tabs", async ({ page }) => {
+test("animates todo opening without replaying it across session tabs", async ({ page }) => {
   test.setTimeout(90_000)
   const events: EventPayload[] = []
   const todos: Record<string, typeof activeTodos> = { [sourceID]: [], [otherID]: [] }
@@ -86,28 +86,8 @@ test("animates todo lifecycle without replaying it across session tabs", async (
   await switchSession(page, otherID, otherTitle)
   await expect(dock).toHaveCount(0)
 
-  const returningOpen = sampleDock(page, 700)
-  await switchSession(page, sourceID, sourceTitle)
-  const openSamples = (await returningOpen).filter((sample) => sample.present)
-  expect(openSamples.length).toBeGreaterThan(0)
-  expect(openSamples[0]!.opacity).toBeGreaterThan(0.98)
-  expect(openSamples[0]!.height).toBeGreaterThan(70)
-  await expect(dock.locator('[data-state="in_progress"]')).toHaveCount(1)
-
-  const completedTodos = activeTodos.map((todo) => ({ ...todo, status: "completed" }))
-  const closing = sampleDock(page, 1_000)
-  todos[sourceID] = completedTodos
-  events.push(todoEvent(sourceID, completedTodos))
-  await expect(dock).toHaveCount(0)
-  expect((await closing).some((sample) => sample.opacity > 0.05 && sample.opacity < 0.95)).toBe(true)
-  todos[sourceID] = []
-  events.push(todoEvent(sourceID, []))
-
-  await switchSession(page, otherID, otherTitle)
-  const returningEmpty = sampleDock(page, 700)
   await switchSession(page, sourceID, sourceTitle)
   await expect(dock).toHaveCount(0)
-  expect((await returningEmpty).every((sample) => !sample.present)).toBe(true)
 })
 
 function session(id: string, title: string, created: number) {

packages/app/e2e/regression/tab-navigate-mousedown.spec.ts

@@ -89,23 +89,19 @@ async function mockServer(page: Page) {
     if (url.origin !== server) return route.fallback()
     if ([`/api/session/${unresolvedSessionID}`, `/session/${unresolvedSessionID}`].includes(url.pathname))
       return new Promise(() => {})
-    if (url.pathname === "/global/event" || url.pathname === "/event" || url.pathname === "/api/event")
+    if (url.pathname === "/api/event")
       return sse(route)
-    if (url.pathname === "/global/health") return json(route, { healthy: true })
     if (url.pathname === "/api/session") return json(route, { data: sessions.map(currentSession), cursor: {} })
     if (url.pathname === "/api/session/active") return json(route, { data: {} })
     const currentSessionInfo = sessions.find((item) => url.pathname === `/api/session/${item.id}`)
     if (currentSessionInfo) return json(route, { data: currentSession(currentSessionInfo) })
     if (sessions.some((item) => url.pathname === `/api/session/${item.id}/message`))
       return json(route, { data: [], cursor: {} })
-    const byId = sessions.find((item) => url.pathname === `/session/${item.id}`)
-    if (byId) return json(route, byId)
     if (/^\/session\/[^/]+$/.test(url.pathname)) return json(route, { name: "NotFoundError" }, 404)
     if (/^\/session\/[^/]+\/message$/.test(url.pathname)) return json(route, [])
     if (/^\/session\/[^/]+\/(children|todo|diff)$/.test(url.pathname)) return json(route, [])
     if (["/skill", "/command", "/lsp", "/formatter", "/permission", "/question", "/vcs/diff"].includes(url.pathname))
       return json(route, [])
-    if (["/global/config", "/config", "/provider/auth", "/mcp"].includes(url.pathname)) return json(route, {})
     if (url.pathname === "/provider")
       return json(route, { all: [], connected: [], default: { providerID: "", modelID: "" } })
     if (url.pathname === "/agent") return json(route, [{ name: "build", mode: "primary" }])

packages/app/e2e/regression/terminal-hidden.spec.ts

@@ -10,7 +10,6 @@ const title = "Hidden terminal regression"
 test("unmounts the terminal panel while it is hidden", async ({ page }) => {
   await page.setViewportSize({ width: 1400, height: 900 })
   await mockOpenCodeServer(page, {
-    protocol: "v2",
     directory,
     project: {
       id: projectID,

packages/app/e2e/regression/terminal-tab-switch.spec.ts

@@ -66,7 +66,6 @@ async function readProbe(page: Page) {
 
 async function setup(page: Page) {
   await mockOpenCodeServer(page, {
-    protocol: "v2",
     directory,
     project: {
       id: projectID,

packages/app/e2e/smoke/session-timeline.fixture.ts

@@ -21,7 +21,7 @@ const words = [
   "vector",
 ]
 
-const serverKey = "http://127.0.0.1:4096"
+const serverKey = `http://127.0.0.1:${process.env.PLAYWRIGHT_SERVER_PORT ?? "4096"}`
 const sourceID = "ses_smoke_source"
 const targetID = "ses_smoke_target"
 const directory = "C:/OpenCode/SmokeProject"
@@ -134,7 +134,7 @@ function toolPart(
   return {
     id: id(`prt_tool_${tool}_${partIndex}`, index),
     type: "tool",
-    callID: id("call", index * 10 + partIndex),
+    callID: id("call", index * 100 + partIndex),
     tool,
     state: {
       status: "completed",
@@ -235,8 +235,17 @@ function renderable(part: MessagePart) {
   return part.type !== "step-start" && part.type !== "step-finish" && part.type !== "patch"
 }
 
-function orderedParts(message: Message) {
-  return message.parts.slice().sort((a, b) => a.id.localeCompare(b.id))
+function currentPartIDs(message: Message) {
+  const ordinals = { text: 0, reasoning: 0 }
+  return message.parts
+    .flatMap((part) => {
+      if (!renderable(part)) return []
+      if (part.type === "text") return [`${message.info.id}:text:${ordinals.text++}`]
+      if (part.type === "reasoning") return [`${message.info.id}:reasoning:${ordinals.reasoning++}`]
+      if (part.type === "tool") return [typeof part.callID === "string" ? part.callID : part.id]
+      return []
+    })
+    .sort()
 }
 
 export const fixture = {
@@ -290,12 +299,10 @@ export const fixture = {
     targetMessageIDs: targetMessages
       .filter((message) => message.info.role === "user")
       .map((message) => message.info.id),
-    targetPartIDs: targetMessages.flatMap((message) =>
-      orderedParts(message)
-        .filter(renderable)
-        .map((part) => part.id),
-    ),
-    expandedShellPartID: targetMessages.flatMap((message) => message.parts).find((part) => part.tool === "bash")!.id,
+    targetPartIDs: targetMessages.flatMap(currentPartIDs),
+    expandedShellPartID: targetMessages
+      .flatMap((message) => message.parts)
+      .find((part) => part.tool === "bash")!.callID,
   },
 }
 

packages/app/e2e/smoke/session-timeline.spec.ts

@@ -125,20 +125,19 @@ test.describe("smoke: session timeline", () => {
     })
     await configureSmokePage(page, fixture.directory)
     await page.addInitScript(
-      ({ dirBase64, sourceID, targetID }) => {
+      ({ server, sourceID, targetID }) => {
         localStorage.setItem(
           "opencode.window.browser.dat:tabs",
           JSON.stringify(
             [sourceID, targetID].map((sessionId) => ({
               type: "session",
-              server: "http://127.0.0.1:4096",
-              dirBase64,
+              server,
               sessionId,
             })),
           ),
         )
       },
-      { dirBase64: base64Encode(fixture.directory), sourceID: fixture.sourceID, targetID: fixture.targetID },
+      { server: fixture.serverKey, sourceID: fixture.sourceID, targetID: fixture.targetID },
     )
 
     await page.goto(`/${base64Encode(fixture.directory)}/session/${fixture.targetID}`)
@@ -251,20 +250,19 @@ test.describe("smoke: session timeline", () => {
     })
     await configureSmokePage(page, fixture.directory)
     await page.addInitScript(
-      ({ dirBase64, sourceID, targetID }) => {
+      ({ server, sourceID, targetID }) => {
         localStorage.setItem(
           "opencode.window.browser.dat:tabs",
           JSON.stringify(
             [sourceID, targetID].map((sessionId) => ({
               type: "session",
-              server: "http://127.0.0.1:4096",
-              dirBase64,
+              server,
               sessionId,
             })),
           ),
         )
       },
-      { dirBase64: base64Encode(fixture.directory), sourceID: fixture.sourceID, targetID: fixture.targetID },
+      { server: fixture.serverKey, sourceID: fixture.sourceID, targetID: fixture.targetID },
     )
     await page.goto(`/${base64Encode(fixture.directory)}/session/${fixture.sourceID}`)
     await expectSessionTitle(page, fixture.expected.sourceTitle)

packages/app/e2e/tsconfig.json

@@ -12,6 +12,7 @@
     "./performance/unit/visual-stability.test.ts",
     "./reproduction/timeline-suspense/**/*.ts",
     "./reproduction/timeline-suspense/**/*.tsx",
+    "../src/types.ts",
     "../src/pages/session/timeline/observe-element-offset.ts",
     "./regression/new-session-panel-corner.spec.ts",
     "./regression/session-timeline-context-resize.spec.ts",

packages/app/e2e/user-story/model-selection-flow.spec.ts

@@ -4,11 +4,7 @@ import { expectAppVisible } from "../utils/waits"
 
 const directory = "C:/OpenCode/NewProject"
 
-test("creates a session in a new project, connects OpenCode Go, and selects its model", async ({ page }) => {
-  let connectedGo = false
-  let pendingGo = false
-  const connections: Array<{ integrationID: string; body: unknown }> = []
-
+test("creates a session in a new project and selects its model", async ({ page }) => {
   await mockOpenCodeServer(page, {
     directory,
     project: {
@@ -46,17 +42,9 @@ test("creates a session in a new project, connects OpenCode Go, and selects its
           },
         },
       ],
-      connected: connectedGo ? ["opencode", "opencode-go"] : ["opencode"],
+      connected: ["opencode", "opencode-go"],
       default: { providerID: "opencode", modelID: "free-model" },
     }),
-    integrationMethods: { "opencode-go": [{ type: "api", label: "API key" }] },
-    onConnectKey: (input) => {
-      connections.push(input)
-      if (input.integrationID === "opencode-go") pendingGo = true
-    },
-    onInstanceDispose: () => {
-      if (pendingGo) connectedGo = true
-    },
     sessions: [],
     pageMessages: () => ({ items: [] }),
     fileList: (path) =>
@@ -66,6 +54,17 @@ test("creates a session in a new project, connects OpenCode Go, and selects its
   await page.addInitScript(() => {
     localStorage.setItem("settings.v3", JSON.stringify({ general: { newLayoutDesigns: true } }))
     localStorage.setItem("opencode.global.dat:server", JSON.stringify({ projects: { local: [] } }))
+    localStorage.setItem(
+      "opencode.global.dat:model",
+      JSON.stringify({
+        user: [
+          { providerID: "opencode", modelID: "free-model", visibility: "show" },
+          { providerID: "opencode-go", modelID: "go-model-1", visibility: "show" },
+        ],
+        recent: [],
+        variant: {},
+      }),
+    )
   })
 
   await page.goto("/")
@@ -79,16 +78,7 @@ test("creates a session in a new project, connects OpenCode Go, and selects its
 
   const modelControl = page.locator('[data-action="prompt-model"]')
   await modelControl.click()
-  await expect(page.locator('[data-section="free-models"]')).toContainText("Free models provided by OpenCode")
-
-  await page.locator('[data-provider-id="opencode-go"]').click()
-  await page.locator('[data-input="provider-api-key"]').fill("mock-go-api-key")
-  await page.locator('[data-action="provider-connect-submit"]').click()
-  await expect(page.locator('[data-component="dialog-v2"]')).toHaveCount(0)
-  expect(connections).toEqual([{ integrationID: "opencode-go", body: { type: "api", key: "mock-go-api-key" } }])
-
-  await expect(modelControl).toHaveAttribute("data-control-type", "popover")
-  await modelControl.click()
+  await expect(page.locator('[data-option-key="opencode:free-model"]')).toBeVisible()
   const goModel = page.locator('[data-option-key="opencode-go:go-model-1"]')
   await expect(goModel).toBeVisible()
   await goModel.click()

packages/app/e2e/utils/mock-server.ts

@@ -1,14 +1,17 @@
 import type { Page, Route } from "@playwright/test"
-
-const emptyList = new Set(["/skill", "/command", "/lsp", "/formatter", "/vcs/status", "/vcs/diff"])
-const emptyObject = new Set(["/global/config", "/config", "/provider/auth", "/mcp", "/experimental/resource"])
+import type {
+  JsonValue,
+  PromptAgentAttachment,
+  PromptFileAttachment,
+  SessionMessageAssistant,
+  SessionMessageInfo,
+  SessionStructuredError,
+} from "@opencode-ai/client/promise"
 
 export interface MockServerConfig {
-  protocol?: "v1" | "v2"
   provider: unknown | (() => unknown)
   integrationMethods?: Record<string, unknown[]>
   onConnectKey?: (input: { integrationID: string; body: unknown }) => void
-  onInstanceDispose?: () => void
   directory: string
   project: unknown
   sessions: ({ id: string } & Record<string, unknown>)[]
@@ -33,21 +36,6 @@ export interface MockServerConfig {
 export async function mockOpenCodeServer(page: Page, config: MockServerConfig) {
   const cursors = new Map<string, string>()
   let nextCursor = 0
-  const staticRoutes: Record<string, unknown> = {
-    "/path": {
-      state: config.directory,
-      config: config.directory,
-      worktree: config.directory,
-      directory: config.directory,
-      home: "C:/OpenCode",
-    },
-    "/project": [config.project],
-    "/project/current": config.project,
-    "/agent": [{ name: "build", mode: "primary" }],
-    "/vcs": { branch: "main", default_branch: "main" },
-    "/session": config.sessions,
-  }
-
   await page.route("**/*", async (route) => {
     const url = new URL(route.request().url())
     const targetPort = process.env.PLAYWRIGHT_SERVER_PORT ?? "4096"
@@ -57,61 +45,15 @@ export async function mockOpenCodeServer(page: Page, config: MockServerConfig) {
     if (url.port !== targetPort && url.port !== appPort) return route.fallback()
 
     const path = url.pathname
-    if (path === "/global/event" || path === "/event" || path === "/api/event") {
+    if (path === "/api/event") {
       const events = config.events?.()
       return sse(
         route,
-        path === "/api/event"
-          ? [{ id: "evt_mock_connected", type: "server.connected", data: {} }, ...(events?.map(currentEvent) ?? [])]
-          : [
-              ...(path === "/global/event"
-                ? [{ payload: { id: "evt_mock_connected", type: "server.connected", properties: {} } }]
-                : []),
-              ...(events ?? []),
-            ],
+        [{ id: "evt_mock_connected", type: "server.connected", data: {} }, ...(events?.map(currentEvent) ?? [])],
         config.eventRetry,
       )
     }
-    if (path === "/global/health")
-      return config.protocol === "v2" ? json(route, {}, undefined, 404) : json(route, { healthy: true })
-    if (path === "/api/health" && config.protocol === "v2")
-      return json(route, { healthy: true, version: "2.0.0", pid: 1 })
-    if (path === "/experimental/capabilities") return json(route, { backgroundSubagents: true })
-    if (path === "/provider")
-      return json(route, typeof config.provider === "function" ? config.provider() : config.provider)
-    if (path === "/provider/auth") return json(route, config.integrationMethods ?? {})
-    const legacyAuth = path.match(/^\/auth\/([^/]+)$/)?.[1]
-    if (legacyAuth && route.request().method() === "PUT") {
-      config.onConnectKey?.({ integrationID: legacyAuth, body: route.request().postDataJSON() })
-      return json(route, true)
-    }
-    if (path === "/instance/dispose" && route.request().method() === "POST") {
-      config.onInstanceDispose?.()
-      return json(route, true)
-    }
-    if (path === "/permission")
-      return json(route, typeof config.permissions === "function" ? config.permissions() : (config.permissions ?? []))
-    if (path === "/question")
-      return json(route, typeof config.questions === "function" ? config.questions() : (config.questions ?? []))
-    if (path === "/session/status")
-      return json(
-        route,
-        typeof config.sessionStatus === "function" ? config.sessionStatus() : (config.sessionStatus ?? {}),
-      )
-    if (path === "/vcs/diff" && config.vcsDiff) return json(route, config.vcsDiff)
-    if (path === "/file" && config.fileList)
-      return json(route, await config.fileList(url.searchParams.get("path") ?? ""))
-    if (path === "/file/content" && config.fileContent)
-      return json(route, await config.fileContent(url.searchParams.get("path") ?? ""))
-    if (path === "/find/file" && config.findFiles)
-      return json(
-        route,
-        await config.findFiles({
-          query: url.searchParams.get("query") ?? "",
-          dirs: url.searchParams.get("dirs") ?? undefined,
-          limit: url.searchParams.has("limit") ? Number(url.searchParams.get("limit")) : undefined,
-        }),
-      )
+    if (path === "/api/health") return json(route, { healthy: true, version: "2.0.0", pid: 1 })
     if (path === "/api/reference")
       return json(route, {
         location: {
@@ -134,7 +76,17 @@ export async function mockOpenCodeServer(page: Page, config: MockServerConfig) {
           },
         ],
       })
+    if (path === "/api/provider")
+      return json(route, {
+        location: location(config),
+        data: currentProviders(providerConfig(config)),
+      })
+    if (path === "/api/model") return json(route, { location: location(config), data: currentModels(providerConfig(config)) })
+    if (path === "/api/model/default")
+      return json(route, { location: location(config), data: currentDefaultModel(providerConfig(config)) })
+    if (path === "/api/integration") return json(route, { location: location(config), data: [] })
     if (path === "/api/command") return json(route, { location: location(config), data: [] })
+    if (path === "/api/plugin") return json(route, { location: location(config), data: [] })
     if (path === "/api/mcp") return json(route, { location: location(config), data: [] })
     if (path === "/api/mcp/resource")
       return json(route, { location: location(config), data: { resources: [], templates: [] } })
@@ -142,25 +94,31 @@ export async function mockOpenCodeServer(page: Page, config: MockServerConfig) {
     if (integration && route.request().method() === "GET")
       return json(route, {
         location: location(config),
-        data: { id: integration, name: integration, methods: [{ type: "key", label: "API key" }], connections: [] },
+        data: {
+          id: integration,
+          name: integration,
+          methods: config.integrationMethods?.[integration] ?? [{ type: "key", label: "API key" }],
+          connections: [],
+        },
       })
     const integrationConnect = path.match(/^\/api\/integration\/([^/]+)\/connect\/key$/)?.[1]
     if (integrationConnect && route.request().method() === "POST") {
       config.onConnectKey?.({ integrationID: integrationConnect, body: route.request().postDataJSON() })
       return route.fulfill({ status: 204, headers: { "access-control-allow-origin": "*" } })
     }
+    if (/^\/api\/credential\/[^/]+$/.test(path) && route.request().method() === "DELETE")
+      return route.fulfill({ status: 204, headers: { "access-control-allow-origin": "*" } })
     if (path === "/api/project") return json(route, [config.project])
     if (path === "/api/project/current")
       return json(route, { id: (config.project as { id?: string }).id, directory: config.directory })
-    if (path.startsWith("/api/project/") && route.request().method() === "PATCH") return json(route, config.project)
-    if (path === "/api/path")
-      return json(route, {
-        state: config.directory,
-        config: config.directory,
-        worktree: config.directory,
-        directory: config.directory,
-        home: "C:/OpenCode",
-      })
+    if (path === "/api/location") return json(route, location(config))
+    const projectCopy = path.match(/^\/experimental\/project\/([^/]+)\/copy$/)?.[1]
+    if (projectCopy && route.request().method() === "POST") {
+      const input = route.request().postDataJSON() as { directory: string; name?: string }
+      return json(route, { directory: `${input.directory}/${input.name ?? "copy"}` })
+    }
+    if (projectCopy && route.request().method() === "DELETE")
+      return route.fulfill({ status: 204, headers: { "access-control-allow-origin": "*" } })
     if (path === "/api/permission/request")
       return json(route, {
         location: location(config),
@@ -177,11 +135,43 @@ export async function mockOpenCodeServer(page: Page, config: MockServerConfig) {
       return json(route, { location: location(config), data: { branch: "main", defaultBranch: "main" } })
     if (path === "/api/vcs/status") return json(route, { location: location(config), data: [] })
     if (path === "/api/vcs/diff") return json(route, { location: location(config), data: config.vcsDiff ?? [] })
+    if (path === "/api/fs/list" && config.fileList)
+      return json(route, {
+        location: location(config),
+        data: await config.fileList(url.searchParams.get("path") ?? ""),
+      })
+    const fileRead = path.match(/^\/api\/fs\/read\/(.+)$/)?.[1]
+    if (fileRead && config.fileContent) {
+      const value = await config.fileContent(decodeURIComponent(fileRead))
+      const content = value && typeof value === "object" && "content" in value ? String(value.content) : String(value ?? "")
+      return route.fulfill({ status: 200, body: content, headers: { "content-type": "application/octet-stream" } })
+    }
+    if (path === "/api/fs/find" && config.findFiles) {
+      const entries = await config.findFiles({
+        query: url.searchParams.get("query") ?? "",
+        dirs: url.searchParams.get("type") ?? undefined,
+        limit: url.searchParams.has("limit") ? Number(url.searchParams.get("limit")) : undefined,
+      })
+      return json(route, {
+        location: location(config),
+        data: Array.isArray(entries)
+          ? entries.map((entry) =>
+              typeof entry === "string"
+                ? {
+                    name: entry.split(/[\\/]/).at(-1) ?? entry,
+                    path: entry,
+                    absolute: `${config.directory}/${entry}`,
+                    type: "directory",
+                    ignored: false,
+                  }
+                : entry,
+            )
+          : entries,
+      })
+    }
     if (path === "/api/pty/shells") return json(route, { location: location(config), data: [] })
     if (/^\/api\/pty\/[^/]+\/connect-token$/.test(path))
       return json(route, { location: location(config), data: { ticket: "e2e-ticket", expires_in: 60 } })
-    if (emptyObject.has(path)) return json(route, {})
-    if (emptyList.has(path)) return json(route, [])
     if (path === "/api/session") {
       const directory = url.searchParams.get("directory")
       const parentID = url.searchParams.get("parentID")
@@ -208,7 +198,9 @@ export async function mockOpenCodeServer(page: Page, config: MockServerConfig) {
       })
     }
     if (path === "/api/session/active") {
-      const statuses = (config.sessionStatus ?? {}) as Record<string, { type?: string }>
+      const statuses = (
+        typeof config.sessionStatus === "function" ? config.sessionStatus() : (config.sessionStatus ?? {})
+      ) as Record<string, { type?: string }>
       return json(route, {
         data: Object.fromEntries(
           Object.entries(statuses).flatMap(([id, status]) =>
@@ -226,12 +218,6 @@ export async function mockOpenCodeServer(page: Page, config: MockServerConfig) {
     if (/^\/api\/session\/[^/]+\/permission\/[^/]+\/reply$/.test(path) && route.request().method() === "POST") {
       return route.fulfill({ status: 204, headers: { "access-control-allow-origin": "*" } })
     }
-    if (/^\/question\/[^/]+\/(reply|reject)$/.test(path) && route.request().method() === "POST") {
-      return json(route, true)
-    }
-    if (/^\/session\/[^/]+\/permissions\/[^/]+$/.test(path) && route.request().method() === "POST") {
-      return json(route, true)
-    }
     if (
       /^\/api\/session\/[^/]+\/(archive|rename|interrupt|revert\/clear|revert\/commit)$/.test(path) &&
       route.request().method() === "POST"
@@ -241,7 +227,6 @@ export async function mockOpenCodeServer(page: Page, config: MockServerConfig) {
     if (/^\/api\/session\/[^/]+$/.test(path) && route.request().method() === "DELETE") {
       return route.fulfill({ status: 204, headers: { "access-control-allow-origin": "*" } })
     }
-    if (path in staticRoutes) return json(route, staticRoutes[path])
 
     const currentSessionMatch = path.match(/^\/api\/session\/([^/]+)$/)
     if (currentSessionMatch) {
@@ -252,28 +237,15 @@ export async function mockOpenCodeServer(page: Page, config: MockServerConfig) {
       })
     }
 
-    const sessionMatch = path.match(/^\/session\/([^/]+)$/)
-    if (sessionMatch) {
-      const session = config.sessions.find((s) => s.id === sessionMatch[1])
-      return json(route, session ?? {})
-    }
-
-    const projectMatch = path.match(/^\/project\/([^/]+)$/)
-    if (projectMatch) return json(route, config.project)
-
-    const messageMatch = path.match(/^\/session\/([^/]+)\/message\/([^/]+)$/)
-    if (messageMatch) {
-      config.onMessage?.({ sessionID: messageMatch[1]!, messageID: messageMatch[2]! })
+    const currentMessageMatch = path.match(/^\/api\/session\/([^/]+)\/message\/([^/]+)$/)
+    if (currentMessageMatch) {
+      config.onMessage?.({ sessionID: currentMessageMatch[1]!, messageID: currentMessageMatch[2]! })
       if (config.messageDelay !== undefined) await new Promise((resolve) => setTimeout(resolve, config.messageDelay))
-      const message = config.message?.(messageMatch[1]!, messageMatch[2]!)
+      const message = config.message?.(currentMessageMatch[1]!, currentMessageMatch[2]!)
       if (message === undefined) return json(route, { error: "Message not found" }, undefined, 404)
-      return json(route, message)
+      return json(route, { data: currentMessage(message) })
     }
 
-    const todoMatch = path.match(/^\/session\/([^/]+)\/todo$/)
-    if (todoMatch) return json(route, config.todos?.(todoMatch[1]!) ?? [])
-    if (/^\/session\/[^/]+\/(children|diff)$/.test(path)) return json(route, [])
-
     const currentMessagesMatch = path.match(/^\/api\/session\/([^/]+)\/message$/)
     if (currentMessagesMatch) {
       const token = url.searchParams.get("cursor") ?? undefined
@@ -292,23 +264,6 @@ export async function mockOpenCodeServer(page: Page, config: MockServerConfig) {
       })
     }
 
-    const messagesMatch = path.match(/^\/session\/([^/]+)\/message$/)
-    if (messagesMatch) {
-      const token = url.searchParams.get("before") ?? undefined
-      const before = token ? cursors.get(token) : undefined
-      if (token && !before) return json(route, { error: "Invalid cursor" }, undefined, 400)
-      config.onMessages?.({ sessionID: messagesMatch[1], before, phase: "start" })
-      await config.beforeMessagesResponse?.({ sessionID: messagesMatch[1]!, before })
-      if (config.messageDelay !== undefined) await new Promise((resolve) => setTimeout(resolve, config.messageDelay))
-      const limit = Number(url.searchParams.get("limit") ?? 80)
-      const pageData = config.pageMessages(messagesMatch[1], limit, before)
-      config.onMessages?.({ sessionID: messagesMatch[1], before, phase: "end" })
-      if (!pageData.cursor) return json(route, pageData.items)
-      const cursor = `cursor_${++nextCursor}`
-      cursors.set(cursor, pageData.cursor)
-      return json(route, pageData.items, { "x-next-cursor": cursor })
-    }
-
     if (url.port === targetPort && targetPort !== appPort) return json(route, {})
     return route.fallback()
   })
@@ -317,10 +272,75 @@ export async function mockOpenCodeServer(page: Page, config: MockServerConfig) {
 function location(config: MockServerConfig) {
   return {
     directory: config.directory,
-    project: { id: (config.project as { id?: string }).id, directory: config.directory },
+    project: { id: (config.project as { id?: string }).id, directory: config.directory, canonical: config.directory },
   }
 }
 
+function providerConfig(config: MockServerConfig) {
+  return typeof config.provider === "function" ? config.provider() : config.provider
+}
+
+function currentProviders(value: unknown) {
+  if (!record(value) || !Array.isArray(value.all)) return Array.isArray(value) ? value : []
+  return value.all.filter(record).flatMap((provider) =>
+    typeof provider.id === "string" && typeof provider.name === "string"
+      ? [{ id: provider.id, name: provider.name, package: provider.id }]
+      : [],
+  )
+}
+
+function currentModels(value: unknown) {
+  if (!record(value) || !Array.isArray(value.all)) return []
+  return value.all.filter(record).flatMap((provider) => {
+    if (typeof provider.id !== "string" || !record(provider.models)) return []
+    return Object.values(provider.models)
+      .filter(record)
+      .flatMap((model) => {
+        if (typeof model.id !== "string" || typeof model.name !== "string") return []
+        const limit = record(model.limit) ? model.limit : {}
+        const cost = record(model.cost) ? model.cost : {}
+        return [
+          {
+            id: model.id,
+            modelID: model.id,
+            providerID: provider.id,
+            name: model.name,
+            capabilities: { tools: true, input: ["text"], output: ["text"] },
+            variants: record(model.variants)
+              ? Object.entries(model.variants).map(([id, settings]) => ({
+                  id,
+                  ...(jsonRecord(settings) ? { settings: jsonRecord(settings) } : {}),
+                }))
+              : [],
+            time: { released: Date.now() },
+            cost: [
+              {
+                input: typeof cost.input === "number" ? cost.input : 0,
+                output: typeof cost.output === "number" ? cost.output : 0,
+                cache: { read: 0, write: 0 },
+              },
+            ],
+            status: "active",
+            enabled: true,
+            limit: {
+              context: typeof limit.context === "number" ? limit.context : 200_000,
+              output: typeof limit.output === "number" ? limit.output : 32_000,
+            },
+          },
+        ]
+      })
+  })
+}
+
+function currentDefaultModel(value: unknown) {
+  if (!record(value) || !record(value.default)) return null
+  const selected = value.default
+  const models = currentModels(value)
+  return models.find(
+    (model) => model.providerID === selected.providerID && model.id === selected.modelID,
+  ) ?? null
+}
+
 function currentPermission(value: unknown) {
   const permission = value as Record<string, unknown>
   if (permission.action) return permission
@@ -364,65 +384,224 @@ export function currentSession(session: { id: string } & Record<string, unknown>
   }
 }
 
-function currentMessage(value: unknown) {
-  const item = value as {
-    info: Record<string, unknown> & { id: string; role: "user" | "assistant"; time: { created: number } }
-    parts: Array<Record<string, unknown> & { type: string }>
+export function currentMessage(value: unknown): SessionMessageInfo {
+  if (isCurrentMessage(value)) return value
+  if (!record(value) || !record(value.info) || !Array.isArray(value.parts)) throw new Error("Invalid message fixture")
+
+  const info = value.info
+  const parts = value.parts.filter(record)
+  if (typeof info.id !== "string" || !record(info.time) || typeof info.time.created !== "number")
+    throw new Error("Invalid legacy message fixture")
+
+  const time = {
+    created: info.time.created,
+    ...(typeof info.time.completed === "number" ? { completed: info.time.completed } : {}),
   }
-  if (item.info.role === "user") {
+  if (info.role === "user") {
     return {
-      id: item.info.id,
+      id: info.id,
       type: "user",
-      time: item.info.time,
-      text: item.parts
+      time: { created: time.created },
+      text: parts
         .flatMap((part) => (part.type === "text" && typeof part.text === "string" ? [part.text] : []))
         .join("\n"),
+      files: parts.flatMap((part) => (part.type === "file" ? legacyFile(part) : [])),
+      agents: parts.flatMap((part) => (part.type === "agent" ? legacyAgent(part) : [])),
     }
   }
+  if (info.role !== "assistant") throw new Error("Invalid legacy message role")
+
   return {
-    id: item.info.id,
+    id: info.id,
     type: "assistant",
-    time: item.info.time,
-    agent: item.info.agent ?? "build",
-    model: { id: item.info.modelID ?? "model", providerID: item.info.providerID ?? "provider" },
-    cost: item.info.cost,
-    tokens: item.info.tokens,
-    error: item.info.error,
-    content: item.parts.flatMap<unknown>((part) => {
-      if (part.type === "text" || part.type === "reasoning") return [{ type: part.type, text: part.text ?? "" }]
-      if (part.type !== "tool") return []
-      const state = part.state as Record<string, unknown>
-      return [
-        {
-          type: "tool",
-          id: part.id,
-          name: part.tool,
-          time: state.time ?? { created: item.info.time.created },
-          state:
-            state.status === "pending"
-              ? { status: "streaming", input: state.raw ?? JSON.stringify(state.input ?? {}) }
-              : state.status === "completed"
-                ? {
-                    status: "completed",
-                    input: state.input ?? {},
-                    structured: state.metadata ?? {},
-                    content: [{ type: "text", text: state.output ?? "" }],
-                  }
-                : state.status === "error"
-                  ? {
-                      status: "error",
-                      input: state.input ?? {},
-                      structured: state.metadata ?? {},
-                      content: [],
-                      error: { type: "ToolError", message: state.error ?? "Tool failed" },
-                    }
-                  : { status: "running", input: state.input ?? {}, structured: state.metadata ?? {}, content: [] },
-        },
-      ]
-    }),
+    time,
+    agent: typeof info.agent === "string" ? info.agent : typeof info.mode === "string" ? info.mode : "build",
+    model: {
+      id: typeof info.modelID === "string" ? info.modelID : "model",
+      providerID: typeof info.providerID === "string" ? info.providerID : "provider",
+      ...(typeof info.variant === "string" ? { variant: info.variant } : {}),
+    },
+    content: parts.flatMap((part) => legacyAssistantContent(part, time.created)),
+    ...(typeof info.cost === "number" ? { cost: info.cost } : {}),
+    ...(tokens(info.tokens) ? { tokens: tokens(info.tokens) } : {}),
+    ...(structuredError(info.error) ? { error: structuredError(info.error) } : {}),
+    ...(finish(info.finish) ? { finish: finish(info.finish) } : {}),
   }
 }
 
+function isCurrentMessage(value: unknown): value is SessionMessageInfo {
+  return record(value) && typeof value.id === "string" && typeof value.type === "string" && !record(value.info)
+}
+
+function legacyFile(part: Record<string, unknown>): PromptFileAttachment[] {
+  if (typeof part.mime !== "string" || typeof part.url !== "string") return []
+  const data = part.url.match(/^data:[^,]*;base64,(.*)$/)?.[1] ?? ""
+  const source = record(part.source) ? part.source : undefined
+  const sourceText = source && record(source.text) ? source.text : undefined
+  const mention = mentionFrom(sourceText)
+  const uri = source?.type === "resource" && typeof source.uri === "string" ? source.uri : part.url
+  return [
+    {
+      data,
+      mime: part.mime,
+      source: part.url.startsWith("data:") ? { type: "inline" } : { type: "uri", uri },
+      ...(typeof part.filename === "string" ? { name: part.filename } : {}),
+      ...(mention ? { mention } : {}),
+    },
+  ]
+}
+
+function legacyAgent(part: Record<string, unknown>): PromptAgentAttachment[] {
+  if (typeof part.name !== "string") return []
+  const mention = mentionFrom(record(part.source) ? part.source : undefined)
+  return [{ name: part.name, ...(mention ? { mention } : {}) }]
+}
+
+function mentionFrom(value: Record<string, unknown> | undefined) {
+  if (
+    !value ||
+    typeof value.value !== "string" ||
+    typeof value.start !== "number" ||
+    typeof value.end !== "number"
+  )
+    return
+  return { text: value.value, start: value.start, end: value.end }
+}
+
+function legacyAssistantContent(
+  part: Record<string, unknown>,
+  created: number,
+): SessionMessageAssistant["content"] {
+  if (part.type === "text" && typeof part.text === "string")
+    return [{ type: "text", text: part.text, ...(jsonRecord(part.metadata) ? { state: jsonRecord(part.metadata) } : {}) }]
+  if (part.type === "reasoning" && typeof part.text === "string") {
+    const time = record(part.time) ? part.time : undefined
+    return [
+      {
+        type: "reasoning",
+        text: part.text,
+        ...(jsonRecord(part.metadata) ? { state: jsonRecord(part.metadata) } : {}),
+        ...(time && typeof time.start === "number"
+          ? {
+              time: {
+                created: time.start,
+                ...(typeof time.end === "number" ? { completed: time.end } : {}),
+              },
+            }
+          : {}),
+      },
+    ]
+  }
+  if (part.type !== "tool" || typeof part.id !== "string" || typeof part.tool !== "string" || !record(part.state))
+    return []
+
+  const state = part.state
+  const time = record(state.time) ? state.time : undefined
+  const toolTime = {
+    created: time && typeof time.start === "number" ? time.start : created,
+    ...(time && typeof time.start === "number" ? { ran: time.start } : {}),
+    ...(time && typeof time.end === "number" ? { completed: time.end } : {}),
+  }
+  const input = jsonRecord(state.input) ?? {}
+  const metadata = jsonRecord(state.metadata)
+  const base = {
+    type: "tool" as const,
+    id: typeof part.callID === "string" ? part.callID : part.id,
+    name: part.tool,
+    time: toolTime,
+    ...(typeof part.executed === "boolean" ? { executed: part.executed } : {}),
+    ...(jsonRecord(part.providerState) ? { providerState: jsonRecord(part.providerState) } : {}),
+    ...(jsonRecord(part.providerResultState) ? { providerResultState: jsonRecord(part.providerResultState) } : {}),
+  }
+  if (state.status === "pending")
+    return [{ ...base, state: { status: "streaming", input: typeof state.raw === "string" ? state.raw : JSON.stringify(input) } }]
+  if (state.status === "completed")
+    return [
+      {
+        ...base,
+        state: {
+          status: "completed",
+          input,
+          content: [{ type: "text", text: typeof state.output === "string" ? state.output : "" }],
+          ...(metadata ? { metadata } : {}),
+        },
+      },
+    ]
+  if (state.status === "error")
+    return [
+      {
+        ...base,
+        state: {
+          status: "error",
+          input,
+          error: structuredError(state.error) ?? { type: "ToolError", message: "Tool failed" },
+          ...(metadata ? { metadata } : {}),
+        },
+      },
+    ]
+  return [{ ...base, state: { status: "running", input, metadata: metadata ?? {} } }]
+}
+
+function structuredError(value: unknown): SessionStructuredError | undefined {
+  if (typeof value === "string") return { type: "Error", message: value }
+  if (!record(value)) return
+  if (typeof value.type === "string" && typeof value.message === "string")
+    return { type: value.type, message: value.message }
+  if (typeof value.name !== "string" || !record(value.data) || typeof value.data.message !== "string") return
+  return { type: value.name, message: value.data.message }
+}
+
+function tokens(value: unknown): SessionMessageAssistant["tokens"] | undefined {
+  if (!record(value) || !record(value.cache)) return
+  if (
+    typeof value.input !== "number" ||
+    typeof value.output !== "number" ||
+    typeof value.reasoning !== "number" ||
+    typeof value.cache.read !== "number" ||
+    typeof value.cache.write !== "number"
+  )
+    return
+  return {
+    input: value.input,
+    output: value.output,
+    reasoning: value.reasoning,
+    cache: { read: value.cache.read, write: value.cache.write },
+  }
+}
+
+function finish(value: unknown): SessionMessageAssistant["finish"] | undefined {
+  if (
+    value === "stop" ||
+    value === "length" ||
+    value === "tool-calls" ||
+    value === "content-filter" ||
+    value === "error" ||
+    value === "unknown"
+  )
+    return value
+}
+
+function jsonRecord(value: unknown): Record<string, JsonValue> | undefined {
+  if (!record(value)) return
+  return Object.fromEntries(
+    Object.entries(value).flatMap(([key, item]) => {
+      const next = jsonValue(item)
+      return next === undefined ? [] : [[key, next]]
+    }),
+  )
+}
+
+function jsonValue(value: unknown): JsonValue | undefined {
+  if (value === null || typeof value === "string" || typeof value === "boolean") return value
+  if (typeof value === "number") return Number.isFinite(value) ? value : null
+  if (Array.isArray(value)) return value.map((item) => jsonValue(item) ?? null)
+  return jsonRecord(value)
+}
+
+function record(value: unknown): value is Record<string, unknown> {
+  return !!value && typeof value === "object" && !Array.isArray(value)
+}
+
 function json(route: Route, body: unknown, headers?: Record<string, string>, status = 200) {
   return route.fulfill({
     status,

packages/app/e2e/utils/sse-transport.ts

@@ -3,7 +3,7 @@ import type { Page } from "@playwright/test"
 export type SseConnectionRecord = {
   id: number
   url: string
-  path: "/global/event" | "/event" | "/api/event"
+  path: "/api/event"
   headers: Record<string, string>
   openedAt: number
   endedAt?: number
@@ -174,10 +174,7 @@ export async function installSseTransport<T>(
       const fetch = (input: RequestInfo | URL, init?: RequestInit) => {
         const request = new Request(input, init)
         const url = new URL(request.url)
-        if (
-          url.origin !== server ||
-          (url.pathname !== "/global/event" && url.pathname !== "/event" && url.pathname !== "/api/event")
-        )
+        if (url.origin !== server || url.pathname !== "/api/event")
           return originalFetch(request)
 
         const id = ++nextConnectionID
@@ -193,18 +190,9 @@ export async function installSseTransport<T>(
             record.controller = controller
             connections.push(record)
             if (retry !== undefined) controller.enqueue(encoder.encode(`retry: ${retry}\n\n`))
-            if (url.pathname === "/api/event")
-              controller.enqueue(
-                encoder.encode(frame({ id: `evt_mock_connected_${id}`, type: "server.connected", data: {} })),
-              )
-            if (url.pathname === "/global/event")
-              controller.enqueue(
-                encoder.encode(
-                  frame({
-                    payload: { id: `evt_mock_connected_${id}`, type: "server.connected", properties: {} },
-                  }),
-                ),
-              )
+            controller.enqueue(
+              encoder.encode(frame({ id: `evt_mock_connected_${id}`, type: "server.connected", data: {} })),
+            )
             request.signal.addEventListener(
               "abort",
               () => {

packages/app/package.json

@@ -56,7 +56,6 @@
     "@opencode-ai/client": "workspace:*",
     "@opencode-ai/core": "workspace:*",
     "@opencode-ai/schema": "workspace:*",
-    "@opencode-ai/sdk": "file:vendor/opencode-ai-sdk-1.18.8-dev.tgz",
     "@opencode-ai/session-ui": "workspace:*",
     "@opencode-ai/ui": "workspace:*",
     "@pierre/trees": "1.0.0-beta.4",

packages/app/src/app.tsx

@@ -238,26 +238,6 @@ function UiI18nBridge(props: ParentProps) {
 }
 
 function LayoutCompatibility(props: ParentProps) {
-  const global = useGlobal()
-  const navigate = useNavigate()
-  const server = useServer()
-  const settings = useSettings()
-
-  createEffect(() => {
-    if (settings.general.newLayoutDesigns()) return
-    const current = server.current
-    if (!current) return
-    const protocol = global.ensureServerCtx(current).sdk.protocolKind()
-    if (protocol !== "v2") return
-    const next = global.servers.list().find((s) => {
-      if (ServerConnection.key(s) === ServerConnection.key(current)) return false
-      return global.ensureServerCtx(s).sdk.protocolKind() !== "v2"
-    })
-    if (!next) return
-    navigate("/")
-    queueMicrotask(() => server.setActive(ServerConnection.key(next)))
-  })
-
   return <>{props.children}</>
 }
 

packages/app/src/components/command-palette.ts

@@ -1,5 +1,5 @@
 import { getFilename } from "@opencode-ai/core/util/path"
-import type { Project } from "@opencode-ai/sdk/v2/client"
+import type { Project } from "@/types"
 import type { SessionInfo } from "@opencode-ai/client/promise"
 import { useDialog } from "@opencode-ai/ui/context/dialog"
 import { createMemo, onCleanup } from "solid-js"

packages/app/src/components/dialog-custom-provider.tsx

@@ -130,26 +130,9 @@ export function CustomProviderForm(props: { autofocus?: boolean } = {}) {
   }
 
   const saveMutation = useMutation(() => ({
-    mutationFn: async (result: NonNullable<ReturnType<typeof validate>>) => {
-      if ((await serverSDK().protocol) !== "v1") throw new Error("Custom providers are unavailable on this server")
-      const disabledProviders = serverSync().data.config.disabled_providers ?? []
-      const nextDisabled = disabledProviders.filter((id) => id !== result.providerID)
-
-      if (result.key) {
-        await serverSDK().client.auth.set({
-          providerID: result.providerID,
-          auth: {
-            type: "api",
-            key: result.key,
-          },
-        })
-      }
-
-      await serverSync().updateConfig({
-        provider: { [result.providerID]: result.config },
-        disabled_providers: nextDisabled,
-      })
-      return result
+    mutationFn: async (result: NonNullable<ReturnType<typeof validate>>): Promise<typeof result> => {
+      // TODO: Restore custom providers when V2 exposes config and arbitrary credential APIs.
+      throw new Error(`Custom provider ${result.providerID} is unavailable`)
     },
     onSuccess: (result) => {
       dialog.close()

packages/app/src/components/dialog-edit-project-v2.tsx

@@ -146,7 +146,7 @@ export function DialogEditProjectV2(props: { project: LocalProject; server: Serv
           <ButtonV2 type="button" variant="neutral" disabled={model.save.isPending} onClick={model.close}>
             {language.t("common.cancel")}
           </ButtonV2>
-          <ButtonV2 type="submit" variant="contrast" disabled={model.save.isPending}>
+          <ButtonV2 type="submit" variant="contrast" disabled={!model.supported || model.save.isPending}>
             {model.save.isPending ? language.t("common.saving") : language.t("common.save")}
           </ButtonV2>
         </DialogFooter>

packages/app/src/components/dialog-edit-project.tsx

@@ -160,7 +160,7 @@ export function DialogEditProject(props: { project: LocalProject; server: Server
           <Button type="button" variant="ghost" size="large" onClick={model.close}>
             {language.t("common.cancel")}
           </Button>
-          <Button type="submit" variant="primary" size="large" disabled={model.save.isPending}>
+          <Button type="submit" variant="primary" size="large" disabled={!model.supported || model.save.isPending}>
             {model.save.isPending ? language.t("common.saving") : language.t("common.save")}
           </Button>
         </div>

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

@@ -8,7 +8,7 @@ import { Dialog } from "@opencode-ai/ui/dialog"
 import { List } from "@opencode-ai/ui/list"
 import { showToast } from "@/utils/toast"
 import { extractPromptFromParts } from "@/utils/prompt"
-import type { TextPart as SDKTextPart } from "@opencode-ai/sdk/v2/client"
+import type { TextPart as SDKTextPart } from "@/types"
 import { base64Encode } from "@opencode-ai/core/util/encode"
 import { useLanguage } from "@/context/language"
 
@@ -69,7 +69,7 @@ export const DialogFork: Component = () => {
     const dir = base64Encode(sdk().directory)
 
     sdk()
-      .api.session.fork({ sessionID, messageID: item.id })
+      .api.session.fork({ sessionID, boundary: { type: "before", messageID: item.id } })
       .then((forked) => {
         dialog.close()
         prompt.set(restored, undefined, { dir, id: forked.id })

packages/app/src/components/dialog-select-directory-v2.tsx

@@ -8,7 +8,7 @@ import { createEffect, createMemo, createResource, createSignal, For, onCleanup,
 import { useGlobal } from "@/context/global"
 import { useLanguage } from "@/context/language"
 import { ServerConnection } from "@/context/server"
-import type { Path } from "@opencode-ai/sdk/v2/client"
+import type { Path } from "@/types"
 import {
   absoluteTreePath,
   activeTreeNavigation,
@@ -66,16 +66,21 @@ export function DialogSelectDirectoryV2(props: DialogSelectDirectoryV2Props) {
   let pathArea: HTMLDivElement | undefined
   let navigation = 0
 
-  const missingBase = createMemo(() => !(sync.data.path.home || sync.data.path.directory))
   const [fallbackPath] = createResource(
-    () => (missingBase() ? true : undefined),
-    async (): Promise<Path | undefined> => {
-      if ((await sdk.protocol) !== "v1") return
-      return sdk.client.path
+    () => (!(sync.data.path.home || sync.data.path.directory) ? true : undefined),
+    () =>
+      sdk.api.location
         .get()
-        .then((result) => result.data)
-        .catch(() => undefined)
-    },
+        .then(
+          (location): Path => ({
+            state: "",
+            config: "",
+            worktree: location.project.directory,
+            directory: location.directory,
+            home: "",
+          }),
+        )
+        .catch(() => undefined),
     { initialValue: undefined },
   )
   const home = createMemo(() => sync.data.path.home || fallbackPath()?.home || "")

packages/app/src/components/dialog-select-directory.tsx

@@ -9,7 +9,7 @@ import { useLanguage } from "@/context/language"
 import { ServerConnection } from "@/context/server"
 import { useGlobal } from "@/context/global"
 import { cleanPickerInput, createDirectorySearch, displayPickerPath } from "./directory-picker-domain"
-import type { Path } from "@opencode-ai/sdk/v2/client"
+import type { Path } from "@/types"
 
 interface DialogSelectDirectoryProps {
   title?: string
@@ -57,16 +57,21 @@ export function DialogSelectDirectory(props: DialogSelectDirectoryProps) {
   const [filter, setFilter] = createSignal("")
   let list: ListRef | undefined
 
-  const missingBase = createMemo(() => !(sync.data.path.home || sync.data.path.directory))
   const [fallbackPath] = createResource(
-    () => (missingBase() ? true : undefined),
-    async (): Promise<Path | undefined> => {
-      if ((await sdk.protocol) !== "v1") return
-      return sdk.client.path
+    () => (!(sync.data.path.home || sync.data.path.directory) ? true : undefined),
+    () =>
+      sdk.api.location
         .get()
-        .then((result) => result.data)
-        .catch(() => undefined)
-    },
+        .then(
+          (location): Path => ({
+            state: "",
+            config: "",
+            worktree: location.project.directory,
+            directory: location.directory,
+            home: "",
+          }),
+        )
+        .catch(() => undefined),
     { initialValue: undefined },
   )
 

packages/app/src/components/dialog-select-server.tsx

@@ -16,7 +16,6 @@ import { useGlobal } from "@/context/global"
 import { useLanguage } from "@/context/language"
 import { usePlatform } from "@/context/platform"
 import { normalizeServerUrl, ServerConnection, useServer } from "@/context/server"
-import { detectServerProtocol } from "@/utils/server-protocol"
 import { type ServerHealth, useCheckServerHealth } from "@/utils/server-health"
 import { useSettings } from "@/context/settings"
 import { useTabs } from "@/context/tabs"
@@ -264,13 +263,6 @@ export function useServerManagementController(options: { onSelect?: () => void;
         setStore("addServer", { error: language.t("dialog.server.add.error") })
         return
       }
-      if (
-        !settings.general.newLayoutDesigns() &&
-        (await detectServerProtocol(conn.http, platform.fetch ?? globalThis.fetch)) === "v2"
-      ) {
-        setStore("addServer", { error: language.t("dialog.server.add.error") })
-        return
-      }
 
       resetAdd()
       if (options.navigateOnAdd === false) {
@@ -315,13 +307,6 @@ export function useServerManagementController(options: { onSelect?: () => void;
         setStore("editServer", { error: language.t("dialog.server.add.error") })
         return
       }
-      if (
-        !settings.general.newLayoutDesigns() &&
-        (await detectServerProtocol(conn.http, platform.fetch ?? globalThis.fetch)) === "v2"
-      ) {
-        setStore("editServer", { error: language.t("dialog.server.add.error") })
-        return
-      }
       if (normalized === input.original.http.url) {
         server.add(conn)
       } else {
@@ -360,9 +345,7 @@ export function useServerManagementController(options: { onSelect?: () => void;
 
   const sortedItems = createMemo(() => {
     const raw = items()
-    const list = settings.general.newLayoutDesigns()
-      ? raw
-      : raw.filter((x) => global.ensureServerCtx(x).sdk.protocolKind() !== "v2")
+    const list = raw
     if (!list.length) return list
     const active = current()
     const order = new Map(list.map((url, index) => [url, index] as const))