refactor(app): establish v2 session controller

refactor(app): extract session timeline controller
2026-08-01 16:03:19 -04:00 · 2026-07-29 14:25:07 +08:00 · 2026-07-29 14:24:28 +08:00
462 changed files with 8027 additions and 12274 deletions
@@ -2,12 +2,15 @@ import type { Context } from "../../../packages/plugin/src/tui/context"

 export default {
  id: "test.tui-discovery-smoke",
-  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,
-    // })
+  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)
  },
 }
@@ -67,7 +67,6 @@
        "@opencode-ai/sdk": "file:vendor/opencode-ai-sdk-1.18.8-dev.tgz",
        "@opencode-ai/session-ui": "workspace:*",
        "@opencode-ai/ui": "workspace:*",
-        "@opencode-ai/util": "workspace:*",
        "@pierre/trees": "1.0.0-beta.4",
        "@sentry/solid": "catalog:",
        "@shikijs/transformers": "3.9.2",
@@ -88,6 +87,7 @@
        "@tanstack/solid-query": "5.91.4",
        "@tanstack/solid-virtual": "catalog:",
        "@thisbeyond/solid-dnd": "0.7.5",
+        "diff": "catalog:",
        "effect": "catalog:",
        "fuzzysort": "catalog:",
        "ghostty-web": "github:anomalyco/ghostty-web#83c0a07b8628b748aed073b232cb4b52a6ca11c1",
@@ -143,10 +143,7 @@
        "open": "10.1.2",
        "semver": "catalog:",
        "solid-js": "catalog:",
-        "tree-sitter-bash": "0.25.0",
-        "tree-sitter-powershell": "0.25.10",
        "uqr": "0.1.3",
-        "web-tree-sitter": "0.25.10",
        "ws": "8.21.0",
      },
      "devDependencies": {
@@ -383,6 +380,7 @@
        "@opencode-ai/ai": "workspace:*",
        "@opencode-ai/codemode": "workspace:*",
        "@opencode-ai/effect-drizzle-sqlite": "workspace:*",
+        "@opencode-ai/effect-sqlite-node": "workspace:*",
        "@opencode-ai/plugin": "workspace:*",
        "@opencode-ai/schema": "workspace:*",
        "@opencode-ai/util": "workspace:*",
@@ -403,11 +401,9 @@
        "ignore": "7.0.5",
        "immer": "11.1.4",
        "jsonc-parser": "3.3.1",
-        "tree-sitter-bash": "0.25.0",
-        "tree-sitter-powershell": "0.25.10",
+        "semver": "^7.6.3",
        "turndown": "7.2.0",
        "venice-ai-sdk-provider": "2.1.1",
-        "web-tree-sitter": "0.25.10",
        "which": "6.0.1",
        "zod": "catalog:",
      },
@@ -424,6 +420,7 @@
        "@tsconfig/bun": "catalog:",
        "@types/bun": "catalog:",
        "@types/node": "catalog:",
+        "@types/semver": "catalog:",
        "@types/turndown": "5.0.5",
        "@types/which": "3.0.4",
        "drizzle-kit": "catalog:",
@@ -517,7 +514,6 @@
        "@opencode-ai/core": "workspace:*",
        "@opencode-ai/session-ui": "workspace:*",
        "@opencode-ai/ui": "workspace:*",
-        "@opencode-ai/util": "workspace:*",
        "@pierre/diffs": "catalog:",
        "@solidjs/meta": "catalog:",
        "@solidjs/router": "catalog:",
@@ -604,7 +600,6 @@
        "zod": "catalog:",
      },
      "devDependencies": {
-        "@opencode-ai/theme": "workspace:*",
        "@opentui/core": "catalog:",
        "@opentui/keymap": "catalog:",
        "@opentui/solid": "catalog:",
@@ -616,14 +611,12 @@
        "typescript": "catalog:",
      },
      "peerDependencies": {
-        "@opencode-ai/theme": "workspace:*",
        "@opentui/core": ">=0.4.5",
        "@opentui/keymap": ">=0.4.5",
        "@opentui/solid": ">=0.4.5",
        "solid-js": ">=1.9.0",
      },
      "optionalPeers": [
-        "@opencode-ai/theme",
        "@opentui/core",
        "@opentui/keymap",
        "@opentui/solid",
@@ -1077,11 +1070,9 @@
  },
  "trustedDependencies": [
    "esbuild",
-    "tree-sitter-powershell",
    "protobufjs",
    "electron",
    "web-tree-sitter",
-    "tree-sitter-bash",
  ],
  "patchedDependencies": {
    "@pierre/trees@1.0.0-beta.4": "patches/@pierre%2Ftrees@1.0.0-beta.4.patch",
@@ -4078,10 +4069,10 @@

    "fetch-blob": ["fetch-blob@3.2.0", "", { "dependencies": { "node-domexception": "^1.0.0", "web-streams-polyfill": "^3.0.3" } }, "sha512-7yAQpD2UMJzLi1Dqv7qFYnPbaPx7ZfFK6PiIxQ4PfkGPyNyl2Ugx+a/umUonmKqjhM4DnfbMvdX6otXq83soQQ=="],

-    "ffi-rs": ["ffi-rs@1.3.2", "", { "optionalDependencies": { "@yuuang/ffi-rs-android-arm64": "1.3.2", "@yuuang/ffi-rs-darwin-arm64": "1.3.2", "@yuuang/ffi-rs-darwin-x64": "1.3.2", "@yuuang/ffi-rs-linux-arm-gnueabihf": "1.3.2", "@yuuang/ffi-rs-linux-arm64-gnu": "1.3.2", "@yuuang/ffi-rs-linux-arm64-musl": "1.3.2", "@yuuang/ffi-rs-linux-x64-gnu": "1.3.2", "@yuuang/ffi-rs-linux-x64-musl": "1.3.2", "@yuuang/ffi-rs-win32-arm64-msvc": "1.3.2", "@yuuang/ffi-rs-win32-ia32-msvc": "1.3.2", "@yuuang/ffi-rs-win32-x64-msvc": "1.3.2" } }, "sha512-4s8dX9VbBw/jd5NOuE3EJRqXaIVdjMyiumeeDzrOhtjQRwp6Bz2za7iksWXTnvTQKV/tTdm1s1w7mObe92zPjQ=="],
-
    "file-uri-to-path": ["file-uri-to-path@1.0.0", "", {}, "sha512-0Zt+s3L7Vf1biwWZ29aARiVYLx7iMGnEUl9x33fbB/j3jR81u/O2LbqK+Bm1CDSNDKVtJ/YjwY7TUd5SkeLQLw=="],

+    "ffi-rs": ["ffi-rs@1.3.2", "", { "optionalDependencies": { "@yuuang/ffi-rs-android-arm64": "1.3.2", "@yuuang/ffi-rs-darwin-arm64": "1.3.2", "@yuuang/ffi-rs-darwin-x64": "1.3.2", "@yuuang/ffi-rs-linux-arm-gnueabihf": "1.3.2", "@yuuang/ffi-rs-linux-arm64-gnu": "1.3.2", "@yuuang/ffi-rs-linux-arm64-musl": "1.3.2", "@yuuang/ffi-rs-linux-x64-gnu": "1.3.2", "@yuuang/ffi-rs-linux-x64-musl": "1.3.2", "@yuuang/ffi-rs-win32-arm64-msvc": "1.3.2", "@yuuang/ffi-rs-win32-ia32-msvc": "1.3.2", "@yuuang/ffi-rs-win32-x64-msvc": "1.3.2" } }, "sha512-4s8dX9VbBw/jd5NOuE3EJRqXaIVdjMyiumeeDzrOhtjQRwp6Bz2za7iksWXTnvTQKV/tTdm1s1w7mObe92zPjQ=="],
+
    "filelist": ["filelist@1.0.6", "", { "dependencies": { "minimatch": "^5.0.1" } }, "sha512-5giy2PkLYY1cP39p17Ech+2xlpTRL9HLspOfEgm0L6CwBXBTgsK5ou0JtzYuepxkaQ/tvhCFIJ5uXo0OrM2DxA=="],

    "fill-range": ["fill-range@7.1.1", "", { "dependencies": { "to-regex-range": "^5.0.1" } }, "sha512-YsGpe3WHLK8ZYi4tWDg2Jy3ebRz2rXowDxnld4bkQB00cc/1Zw9AWnC0i9ztDJitivtQvaI9KaLyKrc+hBW0yg=="],
@@ -5640,10 +5631,6 @@

    "traverse": ["traverse@0.3.9", "", {}, "sha512-iawgk0hLP3SxGKDfnDJf8wTz4p2qImnyihM5Hh/sGvQ3K37dPi/w8sRhdNIxYA1TwFwc5mDhIJq+O0RsvXBKdQ=="],

-    "tree-sitter-bash": ["tree-sitter-bash@0.25.0", "", { "dependencies": { "node-addon-api": "^8.2.1", "node-gyp-build": "^4.8.2" }, "peerDependencies": { "tree-sitter": "^0.25.0" }, "optionalPeers": ["tree-sitter"] }, "sha512-gZtlj9+qFS81qKxpLfD6H0UssQ3QBc/F0nKkPsiFDyfQF2YBqYvglFJUzchrPpVhZe9kLZTrJ9n2J6lmka69Vg=="],
-
-    "tree-sitter-powershell": ["tree-sitter-powershell@0.25.10", "", { "dependencies": { "node-addon-api": "^7.1.0", "node-gyp-build": "^4.8.0" }, "peerDependencies": { "tree-sitter": "^0.25.0" }, "optionalPeers": ["tree-sitter"] }, "sha512-bEt8QoySpGFnU3aa8WedQyNMaN6aTwy/WUbvIVt0JSKF+BbJoSHNHu+wCbhj7xLMsfB0AuffmiJm+B8gzva8Lg=="],
-
    "treeverse": ["treeverse@3.0.0", "", {}, "sha512-gcANaAnd2QDZFmHFEOF4k7uc1J/6a6z3DJMd/QwEyxLoKGiptJRwid582r7QIsFlFMIZ3SnxfS52S4hm2DHkuQ=="],

    "trim-lines": ["trim-lines@3.0.1", "", {}, "sha512-kRj8B+YHZCc9kQYdWfJB2/oUl9rA99qbowYYBtr4ui4mZyAQ2JpvVBd/6U2YloATfqBhBTSMhTpgBHtU0Mf3Rg=="],
@@ -6866,8 +6853,6 @@

    "topojson-client/commander": ["commander@2.20.3", "", {}, "sha512-GpVkmM8vF2vQUkj2LvZmD35JxeJOLCwJ9cUkugyk2nuhbv3+mJvpLYYt+0+USMxE+oj+ey/lJEnhZw75x/OMcQ=="],

-    "tree-sitter-bash/node-addon-api": ["node-addon-api@8.9.0", "", {}, "sha512-ekZMeaaIzSQTSpr7X2X3iJM7lTzgnx8ahAG9pJfT/7+14mlEM8ZYQ9cgCDvSSRbReFK0oHli3WrZdCiRsgAT9Q=="],
-
    "tw-to-css/postcss": ["postcss@8.4.31", "", { "dependencies": { "nanoid": "^3.3.6", "picocolors": "^1.0.0", "source-map-js": "^1.0.2" } }, "sha512-PS08Iboia9mts/2ygV3eLpY5ghnUcfLV/EXTOW1E2qYxJKGGBUtNjN76FYHnMs36RmARn41bC0AZmn+rR0OVpQ=="],

    "tw-to-css/tailwindcss": ["tailwindcss@3.3.2", "", { "dependencies": { "@alloc/quick-lru": "^5.2.0", "arg": "^5.0.2", "chokidar": "^3.5.3", "didyoumean": "^1.2.2", "dlv": "^1.1.3", "fast-glob": "^3.2.12", "glob-parent": "^6.0.2", "is-glob": "^4.0.3", "jiti": "^1.18.2", "lilconfig": "^2.1.0", "micromatch": "^4.0.5", "normalize-path": "^3.0.0", "object-hash": "^3.0.0", "picocolors": "^1.0.0", "postcss": "^8.4.23", "postcss-import": "^15.1.0", "postcss-js": "^4.0.1", "postcss-load-config": "^4.0.1", "postcss-nested": "^6.0.1", "postcss-selector-parser": "^6.0.11", "postcss-value-parser": "^4.2.0", "resolve": "^1.22.2", "sucrase": "^3.32.0" }, "bin": { "tailwind": "lib/cli.js", "tailwindcss": "lib/cli.js" } }, "sha512-9jPkMiIBXvPc2KywkraqsUfbfj+dHDb+JPWtSJa9MLFdrPyazI7q6WX2sUrm7R9eVR7qqv3Pas7EvQFzxKnI6w=="],
@@ -1,118 +0,0 @@
-# V1 to V2 Database Migration
-
-## Approach
-
- Use the `dev` branch database schema and migration registry as the V1 baseline.
- Remove migrations that exist only on the V2 branch.
- Generate one canonical migration from the `dev` schema to the final V2 schema.
- Add explicit data operations to that migration where generated DDL is insufficient.
- Test the migration against a populated database at the exact `dev` schema.
-
-## Preserve
-
-The canonical V1 data remains in its existing tables. In particular, preserve `session`, `message`, and `part` rows.
-
-Preserve `workspace` rows and existing `session.workspace_id` values unchanged. The migration must not clear or rebuild
-workspace relationships.
-
-Keep the `todo` table and its data unchanged. V2 does not currently migrate todos into another representation, and the
-generated migration must not drop the table.
-
-## Truncate
-
-Truncate these pre-launch V2 tables before applying schema changes:
-
- `event`
- `event_sequence`
- `session_message`
-
-These rows are not canonical V1 data. Truncating `event` before adding the required `event.created` column means the
-column needs neither a backfill nor a default. After truncation, rebuild `session_message` from canonical V1 `message`
-and `part` rows rather than retaining its pre-launch V2 contents.
-
-## Message Backfill
-
-Backfill canonical V1 history from `message` and `part` into `session_message`. This is the main data transformation in
-the migration. Preserving the V1 tables alone keeps the data safe but does not make existing history visible through the
-V2 session APIs, which read `session_message`.
-
-Reuse each V1 `message.id` as the corresponding `session_message.id`. Stable IDs keep the migration deterministic and
-avoid rewriting other persisted state that may refer to a message.
-
-Within each session, order V1 messages by `time_created` and then `id`, matching the existing V1 message index. Assign
-contiguous `session_message.seq` values starting at `0`.
-
-Map ordinary V1 messages one-to-one by role. Each ordinary V1 user message becomes one V2 `user` row, and each ordinary
-V1 assistant message becomes one V2 `assistant` row. Fold the source message's ordered V1 parts into that row's V2
-payload.
-
-Handle semantic marker parts before applying the ordinary mapping. In particular, a V1 user message containing a
-`compaction` part and its paired assistant summary represent one compaction operation, not two ordinary messages. Special
-part mappings must be decided explicitly before implementing the backfill.
-
-V1 synthetic content is represented by user text parts with `synthetic: true`, not by a separate message role. A V1 user
-message whose visible text parts are all synthetic should become a V2 `synthetic` message. If a V1 user message mixes
-ordinary and synthetic content, preserve the ordinary content in the V2 `user` row and emit the synthetic content as an
-adjacent V2 `synthetic` row. Ignore text parts marked `ignored`, matching V1 model-history behavior.
-
-Use the V1 compaction user message ID as the ID of the collapsed V2 compaction message. This matches V2's use of the
-admitted compaction input ID and preserves references to the initiating message.
-
-For a completed compaction, create one V2 `compaction` row with `status: "completed"`. Set `reason` from the V1
-compaction part's `auto` flag, join the paired summary assistant's nonempty text parts with blank lines for `summary`, and
-serialize the retained V1 tail beginning at `tail_start_id` for `recent`. Use an empty `recent` value when no tail was
-retained, and use the compaction user message creation time. Do not emit the paired summary assistant as a separate V2
-assistant row.
-
-After rebuilding `session_message`, seed `event_sequence` with one row per migrated session. Set its watermark to that
-session's maximum backfilled `session_message.seq`. This prevents new V2 events from reusing sequence numbers or sorting
-before migrated history. The `event` table remains empty.
-
-## Drop
-
-Drop these pre-launch V2 tables without preserving or transforming their rows:
-
- `session_input`
- `session_context_epoch`
-
-Do not transfer `session_input` rows into `session_pending`.
-
-## Create Empty
-
-Let the generated migration create these tables empty:
-
- `instruction_blob`
- `instruction_entry`
- `instruction_state`
- `session_pending`
- `kv`
-
-V1 has no canonical data to backfill into these tables. V2 initializes their state as it runs.
-
-## Fork Storage
-
-V1 has no fork-boundary state to backfill. New V2 forks use a required message boundary and persist it in
-`session.fork_boundary`. The durable fork event contains no parent sequence. Its resolved boundary is one of:
-
- `before`: copy messages before the identified message.
- `through`: copy messages through the identified message.
-
-Forking an empty session is not supported. `session.fork_seq` and `session.fork_message_id` are not part of the final V2
-schema.
-
-New nullable session columns, including `fork_session_id`, `fork_boundary`, and `time_suspended`, require no explicit
-backfill. Existing rows naturally receive `NULL` when the generated migration adds the columns.
-
-## Verification
-
-The canonical migration test should seed representative V1 sessions, messages, parts, todos, projects, accounts,
-credentials, permissions, shares, and workspaces. After migration, it should verify:
-
- Preserved rows and encoded values remain unchanged.
- Todo rows remain available in the unchanged `todo` table.
- `event` is empty, and stale pre-launch rows are absent from the rebuilt projections.
- Backfilled `session_message` rows represent the canonical V1 `message` and `part` history.
- Each migrated session's `event_sequence` watermark matches its maximum backfilled message sequence.
- Dropped tables no longer exist.
- New tables exist and are empty.
- The final schema has no ungenerated changes.
@@ -10,9 +10,7 @@

 ## Conventions

-Per-type constructors live on the type, not as top-level re-exports. Use `Message.system(...)`, `Message.user(...)`, `Message.assistant(...)`, `Message.tool(...)`, `LanguageModel.make(...)`, `ToolDefinition.make(...)`, `ToolCallPart.make(...)`, `ToolResultPart.make(...)`, `ToolChoice.make(...)`, `ToolChoice.named(...)`, `SystemPart.make(...)`, and `GenerationOptions.make(...)` directly. The top-level `LLM` namespace is reserved for request-shaped call APIs: `LLM.request`, `LLM.generate`, `LLM.stream`, `LLM.updateRequest`, and `LLM.generateObject`. Two ways to construct the same thing is one too many.
-
- Keep provider-defined string enums forward-compatible. Expose known values for autocomplete while accepting future values with `Known | (string & {})`; use `Schema.String` at runtime unless rejecting unknown values is required for correctness.
+Per-type constructors live on the type, not as top-level re-exports. Use `Message.system(...)`, `Message.user(...)`, `Message.assistant(...)`, `Message.tool(...)`, `Model.make(...)`, `ToolDefinition.make(...)`, `ToolCallPart.make(...)`, `ToolResultPart.make(...)`, `ToolChoice.make(...)`, `ToolChoice.named(...)`, `SystemPart.make(...)`, and `GenerationOptions.make(...)` directly. The top-level `LLM` namespace is reserved for request-shaped call APIs: `LLM.request`, `LLM.generate`, `LLM.stream`, `LLM.updateRequest`, and `LLM.generateObject`. Two ways to construct the same thing is one too many.

 ## Tests

@@ -48,7 +46,7 @@ const response = yield * LLMClient.generate(request)

 `LLM.request(...)` builds an `LLMRequest`. `LLMClient.generate(...)` reads the executable route carried by `request.model.route`, builds the provider-native body, asks the route's transport for a real `HttpClientRequest.HttpClientRequest`, sends it through `RequestExecutor.Service`, parses the provider stream into common `LLMEvent`s, and finally returns an `LLMResponse`.

-Use `LLMClient.stream(request)` when callers want incremental `LLMEvent`s. Use `LLMClient.generate(request)` when callers want those same events collected into an `LLMResponse`.
+Use `LLMClient.stream(request)` when callers want incremental `LLMEvent`s. Use `LLMClient.generate(request)` when callers want those same events collected into an `LLMResponse`. Use `LLMClient.prepare<Body>(request)` to compile a request through the route pipeline without sending it — the optional `Body` type argument narrows `.body` to the route's native shape (e.g. `prepare<OpenAIChatBody>(...)` returns a `PreparedRequestOf<OpenAIChatBody>`). The runtime body is identical; the generic is a type-level assertion.

 Filter or narrow `LLMEvent` streams with `LLMEvent.is.*` (camelCase guards, e.g. `events.filter(LLMEvent.is.toolCall)`). The kebab-case `LLMEvent.guards["tool-call"]` form also works but prefer `is.*` in new code.

@@ -76,7 +74,7 @@ export const route = Route.make({
 })
 ```

-Route defaults are request-shaping defaults such as `headers`, `limits`, `generation`, `providerOptions`, and `http`. Endpoint host/query belongs on the route endpoint. Selected `LanguageModel` values carry only model id, provider id, and the configured route value. Model capability/catalog metadata lives outside this package; protocol support is enforced by request lowering and typed `AIError`s.
+Route defaults are request-shaping defaults such as `headers`, `limits`, `generation`, `providerOptions`, and `http`. Endpoint host/query belongs on the route endpoint. Selected `Model` values carry only model id, provider id, and the configured route value. Model capability/catalog metadata lives outside this package; protocol support is enforced by request lowering and typed `LLMError`s.

 The four-axis decomposition is the reason DeepSeek, TogetherAI, Cerebras, Baseten, Fireworks, and DeepInfra all reuse `OpenAIChat.protocol` verbatim — each provider deployment is a 5-15 line `Route.make(...)` call instead of a 300-400 line route clone. Bug fixes in one protocol propagate to every consumer of that protocol in a single commit.

@@ -128,7 +126,7 @@ const selected = model("gpt-5", {
 })
 ```

-Keep semantic APIs as separate entrypoints, such as OpenAI `chat` and `responses`. Keep transport choices inside the semantic entrypoint settings, so OpenAI Responses HTTP and WebSocket share one entrypoint. Provider facades may still expose named selectors such as `responsesWebSocket` for direct typed call sites; the package-like contract maps its settings to those selectors before returning an executable `LanguageModel`.
+Keep semantic APIs as separate entrypoints, such as OpenAI `chat` and `responses`. Keep transport choices inside the semantic entrypoint settings, so OpenAI Responses HTTP and WebSocket share one entrypoint. Provider facades may still expose named selectors such as `responsesWebSocket` for direct typed call sites; the package-like contract maps its settings to those selectors before returning an executable `Model`.

 Do not expose `Route` in provider package settings. Route composition stays an implementation detail behind `model(...)`.

@@ -138,15 +136,15 @@ Do not expose `Route` in provider package settings. Route composition stays an i
 packages/ai/src/
  schema/                   canonical Schema model, split by concern
    ids.ts                  branded IDs, literal types, ProviderMetadata
-    options.ts              Generation/Provider/Http options, Limits, LanguageModel, cache policy
+    options.ts              Generation/Provider/Http options, Limits, Model, cache policy
    messages.ts             content parts, Message, ToolDefinition, LLMRequest
-    events.ts               Usage, individual events, LLMEvent, LLMResponse
-    errors.ts               error reasons, AIError, ToolFailure
+    events.ts               Usage, individual events, LLMEvent, PreparedRequest, LLMResponse
+    errors.ts               error reasons, LLMError, ToolFailure
    index.ts                barrel
  llm.ts                    request constructors and convenience helpers
  route/
    index.ts                @opencode-ai/ai/route advanced barrel
-    client.ts               Route.make + LLMClient.stream/generate
+    client.ts               Route.make + LLMClient.prepare/stream/generate
    executor.ts             RequestExecutor service + transport error mapping
    protocol.ts             Protocol type + Protocol.make
    endpoint.ts             Endpoint type + Endpoint.path
@@ -96,7 +96,7 @@ contains identity, capabilities, pricing metadata, provider-specific option
 types, reusable request-behavior defaults, and hidden execution behavior.

 Normal users do not need to learn the current `Route` composite. Protocol,
-endpoint, auth, transport, and hooks are bound behind `LanguageModel`.
+endpoint, auth, transport, and hooks are bound behind `Model`.

 ### Request

@@ -539,7 +539,7 @@ Hosted tools do not pretend to have local handlers, and callers do not inspect a

 ### Run stream

-`LLM.stream` returns an Effect `Stream<RunEvent, AIError, Requirements>`.
+`LLM.stream` returns an Effect `Stream<RunEvent, LLMError, Requirements>`.
 Run events explicitly expose orchestration boundaries:

 ```ts
@@ -828,11 +828,11 @@ portable semantic guarantee.

 ## Error Model

-The Effect error channel is a tagged domain union rather than one `AIError`
+The Effect error channel is a tagged domain union rather than one `LLMError`
 wrapper with nested reasons. Illustrative categories:

 ```ts
-type AIError =
+type LLMError =
  | AuthenticationError
  | InvalidRequestError
  | UnsupportedCapabilityError
@@ -1079,7 +1079,7 @@ The redesign intentionally removes or changes these current concepts:
 | `LLM.generate` means one turn           | `LLM.generate` means complete run                           |
 | `LLMClient.generate/stream`             | `LLM.generateTurn/streamTurn` for one turn                  |
 | `LLMClient.layer` requirement           | Standard Effect requirements exposed directly               |
-| Public `Route` mental model             | Hidden behind executable `LanguageModel`                    |
+| Public `Route` mental model             | Hidden behind executable `Model`                            |
 | `Provider.make` structural helper       | Experimental declarative `Provider.define`                  |
 | Schema classes as canonical values      | Plain immutable values plus schema subpath                  |
 | `LLM.updateRequest`                     | Object spread                                               |
@@ -1089,7 +1089,7 @@ The redesign intentionally removes or changes these current concepts:
 | `generateObject`                        | Typed `output` option on `generate`                         |
 | One event union for provider output     | Separate `TurnEvent` and `RunEvent` unions                  |
 | `providerExecuted` dispatch check       | Distinct hosted-tool constructors                           |
-| One wrapped `AIError`                   | Tagged domain error union                                   |
+| One wrapped `LLMError`                  | Tagged domain error union                                   |

 OpenCode should migrate to `generateTurn` / `streamTurn`, preserving its durable
 prompt admission, persistence, permission, tool settlement, and continuation
@@ -195,7 +195,8 @@ The hosted result is represented as a provider-executed tool call and tool resul
 - **`LLM.request({...})`** — build a provider-neutral `LLMRequest`. Accepts ergonomic inputs (`system: string`, `prompt: string`) that normalize into the canonical Schema classes.
 - **`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.
- **`LanguageModel.make(...)` / `ToolCallPart.make(...)` / `ToolResultPart.make(...)` / `ToolDefinition.make(...)`** — model and tool-related 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`.
@@ -33,7 +33,7 @@ Keep durable identity separate from runtime capability:

 - Durable identity is small serializable data like `{ providerID, modelID }` for
  config, sessions, logs, and catalogs.
- Runtime capability is a `LanguageModel` with a route value, protocol, transport, auth,
+- Runtime capability is a `Model` with a route value, protocol, transport, auth,
  and defaults. It is allowed to contain functions and schemas.
 - If persisted identity needs to become executable, resolve it through an app
  boundary first. Do not make `LLMRequest` recover behavior from a global route
@@ -137,7 +137,7 @@ starts hiding the real provider-specific config.
   - accepts model id only
   - returns executable models
   - does not accept endpoint/auth/deployment overrides
-4. **Language Model**
+4. **Model**
   - model id
   - route value
   - provider id
@@ -164,7 +164,7 @@ execution mechanism:
 ```ts
 type ProviderFacade<APIs, Config> = {
  readonly id: ProviderID
-  readonly model: (id: string) => LanguageModel
+  readonly model: (id: string) => Model
  readonly configure: (input?: Config) => ProviderFacade<APIs, Config>
 } & APIs
 ```
@@ -181,8 +181,8 @@ export const OpenAI = {
  configure: configureOpenAI,
 } satisfies ProviderFacade<
  {
-    responses: (id: string) => LanguageModel
-    chat: (id: string) => LanguageModel
+    responses: (id: string) => Model
+    chat: (id: string) => Model
  },
  OpenAIConfig
 >
@@ -528,7 +528,7 @@ The chosen split is:
 ```txt
 Route = execution mechanics
 Provider facade = configured route group
-LanguageModel = selected executable model carrying route value
+Model = selected executable model carrying route value
 App boundary = explicit durable-config -> typed-provider call
 ```

@@ -549,13 +549,13 @@ App boundary = explicit durable-config -> typed-provider call
  entrypoint maps its scoped `transport` setting before constructing the model.
 - No separate public `LLMClient.layerWithWebSocket`. The runtime should expose one
  client layer with the available transport capabilities.
- No executable `ModelRef`. The executable handle is `LanguageModel`; durable model
+- No executable `ModelRef`. The executable handle is `Model`; durable model
  identity stays separate and cannot execute on its own.

 ## Implementation Todo

- [x] Replace the current executable `ModelRef` with `LanguageModel`.
- [x] Change `LanguageModel.route` to carry a route value, not a `RouteID` string.
+- [x] Replace the current executable `ModelRef` with `Model`.
+- [x] Change `Model.route` to carry a route value, not a `RouteID` string.
 - [ ] Keep a separate durable model identity type for persisted/session/catalog
      data, likely `{ providerID, modelID }`, and make it clear that it cannot
      execute without resolver context.
@@ -566,9 +566,9 @@ App boundary = explicit durable-config -> typed-provider call
 - [x] Remove endpoint/auth escape hatches from route model selection; callers must
      configure endpoint/auth through `route.with(...)` or provider facades before
      calling `.model(...)`.
- [x] Remove request-shaping defaults from `LanguageModel`; selected models now carry only
+- [x] Remove request-shaping defaults from `Model`; selected models now carry only
      id, provider, and configured route while defaults live on routes or requests.
- [x] Rework `LLMClient.stream` / `generate` to read
+- [x] Rework `LLMClient.prepare` / `stream` / `generate` to read
      `request.model.route` directly instead of calling `registeredRoute(...)`.
 - [x] Remove `Route.make(...)` global registration from the normal execution
      path; keep route ids only as diagnostics/provider API labels.
@@ -50,6 +50,18 @@ const request = LLM.request({
  },
 })

+// `http` is intentionally not needed for normal calls. This shows the shape for
+// newly released provider fields before they deserve a typed provider option.
+const rawOverlayExample = LLM.request({
+  model,
+  prompt: "Show the final HTTP overlay shape.",
+  http: {
+    body: { metadata: { example: "tutorial" } },
+    headers: { "x-opencode-tutorial": "1" },
+    query: { debug: "1" },
+  },
+})
+
 // 3. `generate` sends the request and collects the event stream into one
 // response object. `response.text` is the collected text output.
 const generateOnce = Effect.gen(function* () {
@@ -210,15 +222,33 @@ const FakeEcho = {
  }),
 }

+// `LLMClient.prepare` is the lower-level inspection hook: it compiles through
+// body conversion, validation, endpoint, auth, and HTTP construction without
+// sending anything over the network.
+const inspectFakeProvider = Effect.gen(function* () {
+  const prepared = yield* LLMClient.prepare(
+    LLM.request({
+      model: FakeEcho.configure().model("tiny-echo"),
+      prompt: "Show me the provider pipeline.",
+    }),
+  )
+
+  console.log("\n== fake provider prepare ==")
+  console.log("route:", prepared.route)
+  console.log("body:", Formatter.formatJson(prepared.body, { space: 2 }))
+})
+
 // Provide the LLM runtime and the HTTP request executor once. Keep one path
-// enabled at a time so the tutorial can demonstrate generate, stream, or
-// tool-loop behavior without spending tokens on every example.
+// enabled at a time so the tutorial can demonstrate generate, prepare, 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))))
@@ -38,7 +38,7 @@ const resolve = (policy: CachePolicy | undefined): CachePolicyObject => {
 // Protocols whose wire format ignores inline cache markers (OpenAI's implicit
 // prefix caching, Gemini's implicit + out-of-band CachedContent). Skip the
 // whole policy pass for these — emitting hints would be harmless but pointless.
-const RESPECTS_INLINE_HINTS = new Set(["anthropic-messages", "bedrock-converse", "openrouter"])
+const RESPECTS_INLINE_HINTS = new Set(["anthropic-messages", "bedrock-converse"])

 const makeHint = (ttlSeconds: number | undefined): CacheHint =>
  ttlSeconds !== undefined ? new CacheHint({ type: "ephemeral", ttlSeconds }) : new CacheHint({ type: "ephemeral" })
@@ -133,7 +133,6 @@ const countHints = (request: LLMRequest) =>

 export const applyCachePolicy = (request: LLMRequest): LLMRequest => {
  if (!RESPECTS_INLINE_HINTS.has(request.model.route.id)) return request
-  if (request.model.route.id === "openrouter" && (request.cache === undefined || request.cache === "auto")) return request
  const policy = resolve(request.cache)
  if (!policy.tools && !policy.system && !policy.messages) return request

@@ -1,25 +1,25 @@
 import { Context, Effect, Layer } from "effect"
 import { RequestExecutor } from "./route/executor"
 import type { ImageOptions, ImageRequest, ImageRequestFor, ImageResponse } from "./image"
-import type { AIError } from "./schema"
+import type { LLMError } from "./schema"

 export type Execute = RequestExecutor.Interface["execute"]

 export interface Interface {
  readonly generate: <Options extends ImageOptions>(
    request: ImageRequestFor<Options>,
-  ) => Effect.Effect<ImageResponse, AIError>
+  ) => Effect.Effect<ImageResponse, LLMError>
 }

 export class Service extends Context.Service<Service, Interface>()("@opencode/ImageClient") {}

 export const generate = <Options extends ImageOptions>(
  request: ImageRequestFor<Options>,
-): Effect.Effect<ImageResponse, AIError> =>
+): Effect.Effect<ImageResponse, LLMError> =>
  Effect.gen(function* () {
    const client = yield* Service
    return yield* client.generate(request)
-  }) as Effect.Effect<ImageResponse, AIError>
+  }) as Effect.Effect<ImageResponse, LLMError>

 export const layer: Layer.Layer<Service, never, RequestExecutor.Service> = Layer.effect(
  Service,
@@ -1,10 +1,13 @@
 import { Effect, Schema } from "effect"
-import { HttpOptions, InvalidRequestReason, AIError, ModelID, ProviderID, ProviderMetadata, Usage } from "./schema"
+import { HttpOptions, InvalidRequestReason, LLMError, ModelID, ProviderID, ProviderMetadata, Usage } from "./schema"
 import { ImageClient, Service, type Execute as ImageExecute } from "./image-client"

 export interface ImageRoute<Options extends ImageOptions = ImageOptions> {
  readonly id: string
-  readonly generate: (request: ImageRequestFor<Options>, execute: ImageExecute) => Effect.Effect<ImageResponse, AIError>
+  readonly generate: (
+    request: ImageRequestFor<Options>,
+    execute: ImageExecute,
+  ) => Effect.Effect<ImageResponse, LLMError>
 }

 export type ImageOptions = Record<string, unknown>
@@ -143,13 +146,13 @@ export function request(input: ImageRequest | ImageRequestInput) {

 export function generate<const Model extends object>(
  input: ImageRequestInput<Model>,
-): Effect.Effect<ImageResponse, AIError, Service>
-export function generate(input: ImageRequest): Effect.Effect<ImageResponse, AIError, Service>
+): Effect.Effect<ImageResponse, LLMError, Service>
+export function generate(input: ImageRequest): Effect.Effect<ImageResponse, LLMError, Service>
 export function generate(input: ImageRequest | ImageRequestInput) {
  return Effect.try({
    try: () => (input instanceof ImageRequest ? input : request(input)),
    catch: (error) =>
-      new AIError({
+      new LLMError({
        module: "Image",
        method: "generate",
        reason: new InvalidRequestReason({ message: error instanceof Error ? error.message : String(error) }),
@@ -5,8 +5,8 @@ export { Provider } from "./provider"
 export { ProviderPackage } from "./provider-package"
 export { isContextOverflow, isContextOverflowFailure } from "./provider-error"
 export type {
-  RouteLanguageModelInput,
-  RouteRoutedLanguageModelInput,
+  RouteModelInput,
+  RouteRoutedModelInput,
  Interface as LLMClientShape,
  Service as LLMClientService,
 } from "./route/client"
@@ -33,7 +33,7 @@ export type {
 export * as LLM from "./llm"
 export type {
  Definition as ProviderDefinition,
-  LanguageModelFactory as ProviderLanguageModelFactory,
-  LanguageModelOptions as ProviderLanguageModelOptions,
+  ModelFactory as ProviderModelFactory,
+  ModelOptions as ProviderModelOptions,
 } from "./provider"
 export type { Definition as ProviderPackageDefinition, Settings as ProviderPackageSettings } from "./provider-package"
@@ -4,33 +4,30 @@ import {
  GenerationOptions,
  HttpOptions,
  InvalidProviderOutputReason,
-  AIError,
+  LLMError,
  LLMEvent,
  LLMRequest,
  LLMResponse,
  Message,
-  LanguageModel,
  SystemPart,
  ToolChoice,
  ToolDefinition,
  type ContentPart,
-  type LanguageModelProviderOptions,
 } 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<SelectedLanguageModel extends LanguageModel = LanguageModel> = Omit<
+export type RequestInput = Omit<
  ConstructorParameters<typeof LLMRequest>[0],
-  "model" | "system" | "messages" | "tools" | "toolChoice" | "generation" | "http" | "providerOptions"
+  "system" | "messages" | "tools" | "toolChoice" | "generation" | "http" | "providerOptions"
 > & {
-  readonly model: SelectedLanguageModel
  readonly system?: string | SystemPart | ReadonlyArray<SystemPart>
  readonly prompt?: string | ContentPart | ReadonlyArray<ContentPart>
  readonly messages?: ReadonlyArray<Message | Message.Input>
  readonly tools?: ReadonlyArray<ToolDefinition.Input>
  readonly toolChoice?: ToolChoice.Input
  readonly generation?: GenerationOptions.Input
-  readonly providerOptions?: NoInfer<LanguageModelProviderOptions<SelectedLanguageModel>>
+  readonly providerOptions?: ConstructorParameters<typeof LLMRequest>[0]["providerOptions"]
  readonly http?: HttpOptions.Input
 }

@@ -38,9 +35,7 @@ export const generate = LLMClient.generate

 export const stream = LLMClient.stream

-export const request = <const SelectedLanguageModel extends LanguageModel>(
-  input: RequestInput<SelectedLanguageModel>,
-) => {
+export const request = (input: RequestInput) => {
  const {
    system: requestSystem,
    prompt,
@@ -68,10 +63,7 @@ const GENERATE_OBJECT_TOOL_NAME = "generate_object"

 const GENERATE_OBJECT_TOOL_DESCRIPTION = "Return the structured result by calling this tool."

-type GenerateObjectBase<SelectedLanguageModel extends LanguageModel = LanguageModel> = Omit<
-  RequestInput<SelectedLanguageModel>,
-  "tools" | "toolChoice"
->
+type GenerateObjectBase = Omit<RequestInput, "tools" | "toolChoice">

 export class GenerateObjectResponse<T> {
  constructor(
@@ -88,15 +80,11 @@ export class GenerateObjectResponse<T> {
  }
 }

-export interface GenerateObjectOptions<
-  S extends ToolSchema<any>,
-  SelectedLanguageModel extends LanguageModel = LanguageModel,
-> extends GenerateObjectBase<SelectedLanguageModel> {
+export interface GenerateObjectOptions<S extends ToolSchema<any>> extends GenerateObjectBase {
  readonly schema: S
 }

-export interface GenerateObjectDynamicOptions<SelectedLanguageModel extends LanguageModel = LanguageModel>
-  extends GenerateObjectBase<SelectedLanguageModel> {
+export interface GenerateObjectDynamicOptions extends GenerateObjectBase {
  /** Raw JSON Schema object describing the expected output shape. */
  readonly jsonSchema: JsonSchema.JsonSchema
 }
@@ -115,7 +103,7 @@ const runGenerateObject = Effect.fn("LLM.generateObject")(function* (
    (event) => LLMEvent.is.toolCall(event) && event.name === GENERATE_OBJECT_TOOL_NAME,
  )
  if (!call || !LLMEvent.is.toolCall(call))
-    return yield* new AIError({
+    return yield* new LLMError({
      module: "LLM",
      method: "generateObject",
      reason: new InvalidProviderOutputReason({
@@ -125,7 +113,7 @@ const runGenerateObject = Effect.fn("LLM.generateObject")(function* (
  const object = yield* tool._decode(call.input).pipe(
    Effect.mapError(
      (error) =>
-        new AIError({
+        new LLMError({
          module: "LLM",
          method: "generateObject",
          reason: new InvalidProviderOutputReason({
@@ -145,16 +133,16 @@ const runGenerateObject = Effect.fn("LLM.generateObject")(function* (
 * Two input modes:
 *
 * 1. `schema: EffectSchema<T>` — `.object` is decoded and typed as `T`.
- *    Decode failures surface as `AIError`.
+ *    Decode failures surface as `LLMError`.
 * 2. `jsonSchema: JsonSchema.JsonSchema` — `.object` is `unknown`. Use when
 *    the schema is only available at runtime (MCP, plugin manifests). Caller validates.
 */
-export function generateObject<const SelectedLanguageModel extends LanguageModel, S extends ToolSchema<any>>(
-  options: GenerateObjectOptions<S, SelectedLanguageModel>,
-): Effect.Effect<GenerateObjectResponse<Schema.Schema.Type<S>>, AIError>
-export function generateObject<const SelectedLanguageModel extends LanguageModel>(
-  options: GenerateObjectDynamicOptions<SelectedLanguageModel>,
-): Effect.Effect<GenerateObjectResponse<unknown>, AIError>
+export function generateObject<S extends ToolSchema<any>>(
+  options: GenerateObjectOptions<S>,
+): Effect.Effect<GenerateObjectResponse<Schema.Schema.Type<S>>, LLMError>
+export function generateObject(
+  options: GenerateObjectDynamicOptions,
+): Effect.Effect<GenerateObjectResponse<unknown>, LLMError>
 export function generateObject(options: GenerateObjectOptions<ToolSchema<any>> | GenerateObjectDynamicOptions) {
  if ("schema" in options) {
    const { schema, ...rest } = options
@@ -6,7 +6,7 @@ import { Endpoint } from "../route/endpoint"
 import { Framing } from "../route/framing"
 import { Protocol } from "../route/protocol"
 import {
-  AIError,
+  LLMError,
  LLMEvent,
  mergeJsonRecords,
  Usage,
@@ -242,14 +242,16 @@ const AnthropicUsage = Schema.StructWithRest(
    cache_creation_input_tokens: optionalNull(Schema.Number),
    cache_read_input_tokens: optionalNull(Schema.Number),
    server_tool_use: optionalNull(
-      Schema.StructWithRest(Schema.Struct({ web_search_requests: Schema.optional(Schema.Number) }), [
-        Schema.Record(Schema.String, Schema.Unknown),
-      ]),
+      Schema.StructWithRest(
+        Schema.Struct({ web_search_requests: Schema.optional(Schema.Number) }),
+        [Schema.Record(Schema.String, Schema.Unknown)],
+      ),
    ),
    output_tokens_details: optionalNull(
-      Schema.StructWithRest(Schema.Struct({ thinking_tokens: Schema.optional(Schema.Number) }), [
-        Schema.Record(Schema.String, Schema.Unknown),
-      ]),
+      Schema.StructWithRest(
+        Schema.Struct({ thinking_tokens: Schema.optional(Schema.Number) }),
+        [Schema.Record(Schema.String, Schema.Unknown)],
+      ),
    ),
  }),
  [Schema.Record(Schema.String, Schema.Unknown)],
@@ -723,7 +725,8 @@ const mergeUsage = (left: Usage | undefined, right: Usage | undefined) => {
    reasoningTokens,
    totalTokens: ProviderShared.totalTokens(inputTokens, outputTokens, undefined),
    providerMetadata: {
-      anthropic: mergeJsonRecords(left.providerMetadata?.["anthropic"], right.providerMetadata?.["anthropic"]) ?? {},
+      anthropic:
+        mergeJsonRecords(left.providerMetadata?.["anthropic"], right.providerMetadata?.["anthropic"]) ?? {},
    },
  })
 }
@@ -813,8 +816,7 @@ const onContentBlockStart = (state: ParserState, event: AnthropicEvent): StepRes
  if (block.type === "thinking" && block.thinking !== undefined) {
    const events: LLMEvent[] = []
    const id = `reasoning-${event.index ?? 0}`
-    const providerMetadata =
-      block.signature === undefined ? undefined : anthropicMetadata({ signature: block.signature })
+    const providerMetadata = block.signature === undefined ? undefined : anthropicMetadata({ signature: block.signature })
    const lifecycle = Lifecycle.reasoningStart(state.lifecycle, events, id, providerMetadata)
    return [
      {
@@ -978,7 +980,7 @@ const providerErrorMessage = (event: AnthropicEvent): string => {
 }

 const onError = (event: AnthropicEvent) =>
-  new AIError({
+  new LLMError({
    module: ADAPTER,
    method: "stream",
    reason: classifyProviderFailure({ message: providerErrorMessage(event), code: event.error?.type }),
@@ -3,7 +3,7 @@ import { Route } from "../route/client"
 import { Endpoint } from "../route/endpoint"
 import { Protocol } from "../route/protocol"
 import {
-  AIError,
+  LLMError,
  LLMEvent,
  Usage,
  type CacheHint,
@@ -11,7 +11,7 @@ import {
  type FinishReasonDetails,
  type JsonSchema,
  type LLMRequest,
-  type LanguageModelToolSchemaCompatibility,
+  type ModelToolSchemaCompatibility,
  type ProviderMetadata,
  type ReasoningPart,
  type ToolCallPart,
@@ -232,7 +232,7 @@ const lowerToolSpec = (tool: ToolDefinition, inputSchema: JsonSchema): BedrockTo
 })

 const lowerTools = (
-  compatibility: LanguageModelToolSchemaCompatibility | undefined,
+  compatibility: ModelToolSchemaCompatibility | undefined,
  breakpoints: BedrockCache.Breakpoints,
  tools: ReadonlyArray<ToolDefinition>,
 ): BedrockTool[] => {
@@ -274,7 +274,9 @@ const reasoningSignature = (part: ReasoningPart) => {

 const reasoningRedactedData = (part: ReasoningPart) => {
  const bedrock = part.providerMetadata?.bedrock
-  return ProviderShared.isRecord(bedrock) && typeof bedrock.redactedData === "string" ? bedrock.redactedData : undefined
+  return ProviderShared.isRecord(bedrock) && typeof bedrock.redactedData === "string"
+    ? bedrock.redactedData
+    : undefined
 }

 const lowerToolCall = (part: ToolCallPart): BedrockToolUseBlock => ({
@@ -654,7 +656,7 @@ const step = (state: ParserState, event: BedrockEvent) =>
      ] as const
    ).find((entry) => entry[1] !== undefined)
    if (exception) {
-      return yield* new AIError({
+      return yield* new LLMError({
        module: ADAPTER,
        method: "stream",
        reason: classifyProviderFailure({
@@ -29,30 +29,9 @@ export const DEFAULT_BASE_URL = "https://generativelanguage.googleapis.com/v1bet

 export interface OptionsInput {
  readonly [key: string]: unknown
-  readonly cachedContent?: string
-  readonly safetySettings?: ReadonlyArray<{
-    readonly category:
-      | "HARM_CATEGORY_UNSPECIFIED"
-      | "HARM_CATEGORY_HATE_SPEECH"
-      | "HARM_CATEGORY_DANGEROUS_CONTENT"
-      | "HARM_CATEGORY_HARASSMENT"
-      | "HARM_CATEGORY_SEXUALLY_EXPLICIT"
-      | "HARM_CATEGORY_CIVIC_INTEGRITY"
-      | (string & {})
-    readonly threshold:
-      | "HARM_BLOCK_THRESHOLD_UNSPECIFIED"
-      | "BLOCK_LOW_AND_ABOVE"
-      | "BLOCK_MEDIUM_AND_ABOVE"
-      | "BLOCK_ONLY_HIGH"
-      | "BLOCK_NONE"
-      | "OFF"
-      | (string & {})
-  }>
-  readonly serviceTier?: "standard" | "flex" | "priority" | (string & {})
  readonly thinkingConfig?: {
    readonly thinkingBudget?: number
    readonly includeThoughts?: boolean
-    readonly thinkingLevel?: "minimal" | "low" | "medium" | "high" | (string & {})
  }
 }

@@ -132,12 +111,6 @@ const GeminiToolConfig = Schema.Struct({
 const GeminiThinkingConfig = Schema.Struct({
  thinkingBudget: Schema.optional(Schema.Number),
  includeThoughts: Schema.optional(Schema.Boolean),
-  thinkingLevel: Schema.optional(Schema.String),
-})
-
-const GeminiSafetySetting = Schema.Struct({
-  category: Schema.String,
-  threshold: Schema.String,
 })

 const GeminiGenerationConfig = Schema.Struct({
@@ -150,10 +123,7 @@ const GeminiGenerationConfig = Schema.Struct({
 })

 const GeminiBodyFields = {
-  cachedContent: Schema.optional(Schema.String),
  contents: Schema.Array(GeminiContent),
-  safetySettings: optionalArray(GeminiSafetySetting),
-  serviceTier: Schema.optional(Schema.String),
  systemInstruction: Schema.optional(GeminiSystemInstruction),
  tools: optionalArray(GeminiTool),
  toolConfig: Schema.optional(GeminiToolConfig),
@@ -346,38 +316,17 @@ const lowerMessages = Effect.fn("Gemini.lowerMessages")(function* (request: LLMR
 })

 const resolveOptions = (request: LLMRequest) => {
-  const input = request.providerOptions?.gemini
-  const value = input?.thinkingConfig
+  const value = request.providerOptions?.gemini?.thinkingConfig
+  if (!ProviderShared.isRecord(value)) return {}
  const thinkingConfig = {
-    thinkingBudget:
-      ProviderShared.isRecord(value) && typeof value.thinkingBudget === "number" ? value.thinkingBudget : undefined,
-    includeThoughts:
-      ProviderShared.isRecord(value) && typeof value.includeThoughts === "boolean"
-        ? value.includeThoughts
-        : ProviderShared.isRecord(value)
-          ? true
-          : undefined,
-    thinkingLevel:
-      ProviderShared.isRecord(value) && typeof value.thinkingLevel === "string" ? value.thinkingLevel : undefined,
+    thinkingBudget: typeof value.thinkingBudget === "number" ? value.thinkingBudget : undefined,
+    includeThoughts: typeof value.includeThoughts === "boolean" ? value.includeThoughts : undefined,
  }
  return {
-    cachedContent: typeof input?.cachedContent === "string" ? input.cachedContent : undefined,
-    safetySettings: mapSafetySettings(input?.safetySettings),
-    serviceTier: typeof input?.serviceTier === "string" ? input.serviceTier : undefined,
    thinkingConfig: Object.values(thinkingConfig).some((item) => item !== undefined) ? thinkingConfig : undefined,
  }
 }

-function mapSafetySettings(value: unknown) {
-  if (!Array.isArray(value)) return undefined
-  const settings = value.flatMap((item) =>
-    ProviderShared.isRecord(item) && typeof item.category === "string" && typeof item.threshold === "string"
-      ? [{ category: item.category, threshold: item.threshold }]
-      : [],
-  )
-  return settings
-}
-
 const fromRequest = Effect.fn("Gemini.fromRequest")(function* (request: LLMRequest) {
  const hasTools = request.tools.length > 0
  const generation = request.generation
@@ -393,10 +342,7 @@ const fromRequest = Effect.fn("Gemini.fromRequest")(function* (request: LLMReque
  }

  return {
-    cachedContent: options.cachedContent,
    contents: yield* lowerMessages(request),
-    safetySettings: options.safetySettings,
-    serviceTier: options.serviceTier,
    systemInstruction:
      request.system.length === 0 ? undefined : { parts: [{ text: ProviderShared.joinText(request.system) }] },
    tools: hasTools
@@ -11,7 +11,7 @@ import {
 import { Auth, type Definition as AuthDefinition } from "../route/auth"
 import {
  InvalidProviderOutputReason,
-  AIError,
+  LLMError,
  Usage,
  mergeHttpOptions,
  mergeJsonRecords,
@@ -125,7 +125,7 @@ const nativeOptions = (options: GoogleImageOptions | undefined) => {
 }

 const invalidOutput = (message: string, providerMetadata?: ProviderMetadata) =>
-  new AIError({
+  new LLMError({
    module: ADAPTER,
    method: "generate",
    reason: new InvalidProviderOutputReason({ message, route: ADAPTER, providerMetadata }),
@@ -285,7 +285,7 @@ export const model = (input: ModelInput) => {
  return ImageModel.make<GoogleImageOptions>({ id: input.id, provider: "google", route, http: input.http })
 }

-const googleImagePart = (image: ImageInput): Effect.Effect<Record<string, unknown>, AIError> => {
+const googleImagePart = (image: ImageInput): Effect.Effect<Record<string, unknown>, LLMError> => {
  if (image.type === "bytes")
    return Effect.succeed({ inlineData: { mimeType: image.mediaType, data: Encoding.encodeBase64(image.data) } })
  if (image.type === "file-uri") return Effect.succeed({ fileData: { mimeType: image.mediaType, fileUri: image.uri } })
@@ -3,7 +3,7 @@ import type { Content } from "@opencode-ai/schema/tool"
 import { HttpTransport } from "../route/transport"
 import { Protocol } from "../route/protocol"
 import {
-  AIError,
+  LLMError,
  LLMEvent,
  Usage,
  type FinishReason,
@@ -434,7 +434,10 @@ const lowerMessages = Effect.fn("OpenResponses.lowerMessages")(function* (reques
        const groups = content.reduce<Array<{ phase: MessagePhase | null | undefined; parts: TextPart[] }>>(
          (groups, part) => {
            const metadata = part.providerMetadata?.[providerMetadataKey]
-            const phase = ProviderShared.isRecord(metadata) ? messagePhase(metadata.phase, extension) : undefined
+            const phase =
+              ProviderShared.isRecord(metadata)
+                ? messagePhase(metadata.phase, extension)
+                : undefined
            const group = groups.at(-1)
            if (group && group.phase === phase) group.parts.push(part)
            else groups.push({ phase, parts: [part] })
@@ -643,7 +646,10 @@ const onOutputTextDelta = (state: ParserState, event: Event, id: string): StepRe
  const phase = state.messagePhases[id]
  const metadata = phase === undefined ? undefined : providerMetadata(state, { phase })
  const lifecycle = Lifecycle.textStart(state.lifecycle, events, id, metadata)
-  return [{ ...state, lifecycle: Lifecycle.textDelta(lifecycle, events, id, event.delta) }, events]
+  return [
+    { ...state, lifecycle: Lifecycle.textDelta(lifecycle, events, id, event.delta) },
+    events,
+  ]
 }

 const onOutputTextDone = (state: ParserState, event: Event, id: string): StepResult => {
@@ -969,7 +975,7 @@ const providerErrorMessage = (event: Event, fallback: string): string => {
 const providerError = (state: ParserState, event: Event, fallback: string) => {
  const code = event.code || event.error?.code || event.response?.error?.code || undefined
  const message = providerErrorMessage(event, fallback)
-  return new AIError({
+  return new LLMError({
    module: state.id,
    method: "stream",
    reason: classifyProviderFailure({ message, code }),
@@ -6,12 +6,11 @@ import { Endpoint } from "../route/endpoint"
 import { HttpTransport } from "../route/transport"
 import { Protocol } from "../route/protocol"
 import {
-  AIError,
+  LLMError,
  LLMEvent,
  Usage,
  type FinishReason,
  type FinishReasonDetails,
-  type CacheHint,
  type JsonSchema,
  type LLMRequest,
  type MediaPart,
@@ -39,11 +38,6 @@ export const PATH = "/chat/completions"
 // The body schema is the provider-native JSON body. `fromRequest` below builds
 // this shape from the common `LLMRequest`, then `Route.make` validates and
 // JSON-encodes it before transport.
-const OpenAIChatCacheControl = Schema.Struct({
-  type: Schema.Literal("ephemeral"),
-  ttl: Schema.optional(Schema.String),
-})
-
 const OpenAIChatFunction = Schema.Struct({
  name: Schema.String,
  description: Schema.String,
@@ -53,7 +47,6 @@ const OpenAIChatFunction = Schema.Struct({
 const OpenAIChatTool = Schema.Struct({
  type: Schema.tag("function"),
  function: OpenAIChatFunction,
-  cache_control: Schema.optional(OpenAIChatCacheControl),
 })
 type OpenAIChatTool = Schema.Schema.Type<typeof OpenAIChatTool>

@@ -68,11 +61,7 @@ const OpenAIChatAssistantToolCall = Schema.Struct({
 type OpenAIChatAssistantToolCall = Schema.Schema.Type<typeof OpenAIChatAssistantToolCall>

 const OpenAIChatUserContent = Schema.Union([
-  Schema.Struct({
-    type: Schema.Literal("text"),
-    text: Schema.String,
-    cache_control: Schema.optional(OpenAIChatCacheControl),
-  }),
+  Schema.Struct({ type: Schema.Literal("text"), text: Schema.String }),
  Schema.Struct({
    type: Schema.Literal("image_url"),
    image_url: Schema.Struct({ url: Schema.String }),
@@ -80,10 +69,7 @@ const OpenAIChatUserContent = Schema.Union([
 ])

 const OpenAIChatMessage = Schema.Union([
-  Schema.Struct({
-    role: Schema.Literal("system"),
-    content: Schema.Union([Schema.String, Schema.Array(OpenAIChatUserContent)]),
-  }),
+  Schema.Struct({ role: Schema.Literal("system"), content: Schema.String }),
  Schema.Struct({
    role: Schema.Literal("user"),
    content: Schema.Union([Schema.String, Schema.Array(OpenAIChatUserContent)]),
@@ -97,16 +83,10 @@ const OpenAIChatMessage = Schema.Union([
      reasoning: Schema.optional(Schema.String),
      reasoning_text: Schema.optional(Schema.String),
      reasoning_details: Schema.optional(Schema.Unknown),
-      cache_control: Schema.optional(OpenAIChatCacheControl),
    }),
    [Schema.Record(Schema.String, Schema.Unknown)],
  ),
-  Schema.Struct({
-    role: Schema.Literal("tool"),
-    tool_call_id: Schema.String,
-    content: Schema.String,
-    cache_control: Schema.optional(OpenAIChatCacheControl),
-  }),
+  Schema.Struct({ role: Schema.Literal("tool"), tool_call_id: Schema.String, content: Schema.String }),
 ]).pipe(Schema.toTaggedUnion("role"))
 type OpenAIChatMessage = Schema.Schema.Type<typeof OpenAIChatMessage>

@@ -127,7 +107,6 @@ export const bodyFields = {
  stream_options: Schema.optional(Schema.Struct({ include_usage: Schema.Boolean })),
  store: Schema.optional(Schema.Boolean),
  reasoning_effort: Schema.optional(OpenAIOptions.OpenAIReasoningEffort),
-  max_completion_tokens: Schema.optional(Schema.Number),
  max_tokens: Schema.optional(Schema.Number),
  temperature: Schema.optional(Schema.Number),
  top_p: Schema.optional(Schema.Number),
@@ -230,20 +209,13 @@ export interface ParserState {
 // Lowering is the only place that knows how common LLM messages map onto the
 // OpenAI Chat wire format. Keep provider quirks here instead of leaking native
 // fields into `LLMRequest`.
-interface LoweringOptions {
-  readonly cacheControl?: (
-    cache: CacheHint | undefined,
-  ) => Schema.Schema.Type<typeof OpenAIChatCacheControl> | undefined
-}
-
-const lowerTool = (tool: ToolDefinition, inputSchema: JsonSchema, options: LoweringOptions): OpenAIChatTool => ({
+const lowerTool = (tool: ToolDefinition, inputSchema: JsonSchema): OpenAIChatTool => ({
  type: "function",
  function: {
    name: tool.name,
    description: tool.description,
    parameters: ToolSchemaProjection.openAI(inputSchema),
  },
-  cache_control: options.cacheControl?.(tool.cache),
 })

 const lowerToolChoice = (toolChoice: NonNullable<LLMRequest["toolChoice"]>) =>
@@ -285,14 +257,11 @@ const reasoningDetails = (parts: ReadonlyArray<ReasoningPart>, native: unknown)
  if (isRecord(native) && Array.isArray(native.reasoning_details)) return native.reasoning_details
 }

-const lowerUserMessage = Effect.fn("OpenAIChat.lowerUserMessage")(function* (
-  message: OpenAIChatRequestMessage,
-  options: LoweringOptions,
-) {
+const lowerUserMessage = Effect.fn("OpenAIChat.lowerUserMessage")(function* (message: OpenAIChatRequestMessage) {
  const content: Array<Schema.Schema.Type<typeof OpenAIChatUserContent>> = []
  for (const part of message.content) {
    if (part.type === "text") {
-      content.push({ type: "text", text: part.text, cache_control: options.cacheControl?.(part.cache) })
+      content.push({ type: "text", text: part.text })
      continue
    }
    if (part.type === "media") {
@@ -301,18 +270,14 @@ const lowerUserMessage = Effect.fn("OpenAIChat.lowerUserMessage")(function* (
    }
    return yield* ProviderShared.unsupportedContent("OpenAI Chat", "user", ["text", "media"])
  }
-  if (content.every((part) => part.type === "text" && part.cache_control === undefined))
-    return {
-      role: "user" as const,
-      content: content.map((part) => (part.type === "text" ? part.text : "")).join(""),
-    }
+  if (content.every((part) => part.type === "text"))
+    return { role: "user" as const, content: content.map((part) => part.text).join("") }
  return { role: "user" as const, content }
 })

 const lowerAssistantMessage = Effect.fn("OpenAIChat.lowerAssistantMessage")(function* (
  message: OpenAIChatRequestMessage,
  configuredField?: string,
-  options: LoweringOptions = {},
 ) {
  const content: TextPart[] = []
  const reasoning: ReasoningPart[] = []
@@ -350,44 +315,29 @@ const lowerAssistantMessage = Effect.fn("OpenAIChat.lowerAssistantMessage")(func
    if (reasoning.length === 0) return nativeReasoning
    return text
  })()
-  const cached = message.content.findLast((part) => "cache" in part && part.cache !== undefined)
  const result = {
    role: "assistant" as const,
    content: content.length === 0 ? null : ProviderShared.joinText(content),
    tool_calls: toolCalls.length === 0 ? undefined : toolCalls,
    reasoning_details: details,
-    cache_control: options.cacheControl?.(cached && "cache" in cached ? cached.cache : undefined),
  }
  if (field === undefined || reasoningText === undefined) return result
  return { ...result, [field]: reasoningText }
 })

-const lowerToolMessages = Effect.fn("OpenAIChat.lowerToolMessages")(function* (
-  message: OpenAIChatRequestMessage,
-  options: LoweringOptions,
-) {
+const lowerToolMessages = Effect.fn("OpenAIChat.lowerToolMessages")(function* (message: OpenAIChatRequestMessage) {
  const messages: OpenAIChatMessage[] = []
  const images: Array<Schema.Schema.Type<typeof OpenAIChatUserContent>> = []
  for (const part of message.content) {
    if (!ProviderShared.supportsContent(part, ["tool-result"]))
      return yield* ProviderShared.unsupportedContent("OpenAI Chat", "tool", ["tool-result"])
    if (part.result.type !== "content") {
-      messages.push({
-        role: "tool",
-        tool_call_id: part.id,
-        content: ProviderShared.toolResultText(part),
-        cache_control: options.cacheControl?.(part.cache),
-      })
+      messages.push({ role: "tool", tool_call_id: part.id, content: ProviderShared.toolResultText(part) })
      continue
    }
    const content: ReadonlyArray<Tool.Content> = part.result.value
    const text = content.filter((item) => item.type === "text").map((item) => item.text)
-    messages.push({
-      role: "tool",
-      tool_call_id: part.id,
-      content: text.join("\n"),
-      cache_control: options.cacheControl?.(part.cache),
-    })
+    messages.push({ role: "tool", tool_call_id: part.id, content: text.join("\n") })
    const files = content.filter((item) => item.type === "file")
    images.push(
      ...(yield* Effect.forEach(files, (item) =>
@@ -401,29 +351,15 @@ const lowerToolMessages = Effect.fn("OpenAIChat.lowerToolMessages")(function* (
 const lowerMessage = Effect.fn("OpenAIChat.lowerMessage")(function* (
  message: OpenAIChatRequestMessage,
  reasoningField?: string,
-  options: LoweringOptions = {},
 ) {
-  if (message.role === "user") return [yield* lowerUserMessage(message, options)]
-  if (message.role === "assistant") return [yield* lowerAssistantMessage(message, reasoningField, options)]
-  return (yield* lowerToolMessages(message, options)).messages
+  if (message.role === "user") return [yield* lowerUserMessage(message)]
+  if (message.role === "assistant") return [yield* lowerAssistantMessage(message, reasoningField)]
+  return (yield* lowerToolMessages(message)).messages
 })

-const lowerMessages = Effect.fn("OpenAIChat.lowerMessages")(function* (request: LLMRequest, options: LoweringOptions) {
+const lowerMessages = Effect.fn("OpenAIChat.lowerMessages")(function* (request: LLMRequest) {
  const system: OpenAIChatMessage[] =
-    request.system.length === 0
-      ? []
-      : request.system.some((part) => part.cache !== undefined) && options.cacheControl !== undefined
-        ? [
-            {
-              role: "system",
-              content: request.system.map((part) => ({
-                type: "text",
-                text: part.text,
-                cache_control: options.cacheControl?.(part.cache),
-              })),
-            },
-          ]
-        : [{ role: "system", content: ProviderShared.joinText(request.system) }]
+    request.system.length === 0 ? [] : [{ role: "system", content: ProviderShared.joinText(request.system) }]
  const messages = [...system]
  const pendingImages: Array<Schema.Schema.Type<typeof OpenAIChatUserContent>> = []
  const flushImages = () => {
@@ -434,53 +370,28 @@ const lowerMessages = Effect.fn("OpenAIChat.lowerMessages")(function* (request:
    if (message.role === "system") {
      const part = yield* ProviderShared.wrappedSystemUpdate("OpenAI Chat", message)
      if (pendingImages.length > 0) {
-        messages.push({
-          role: "user",
-          content: [
-            ...pendingImages.splice(0),
-            { type: "text", text: part.text, cache_control: options.cacheControl?.(part.cache) },
-          ],
-        })
+        messages.push({ role: "user", content: [...pendingImages.splice(0), { type: "text", text: part.text }] })
        continue
      }
      const previous = messages.at(-1)
      if (previous?.role === "user" && typeof previous.content === "string")
-        messages[messages.length - 1] = options.cacheControl?.(part.cache)
-          ? {
-              role: "user",
-              content: [
-                { type: "text", text: previous.content },
-                { type: "text", text: part.text, cache_control: options.cacheControl(part.cache) },
-              ],
-            }
-          : { role: "user", content: `${previous.content}\n${part.text}` }
+        messages[messages.length - 1] = { role: "user", content: `${previous.content}\n${part.text}` }
      else if (previous?.role === "user" && Array.isArray(previous.content))
        messages[messages.length - 1] = {
          role: "user",
-          content: [
-            ...previous.content,
-            { type: "text", text: part.text, cache_control: options.cacheControl?.(part.cache) },
-          ],
+          content: [...previous.content, { type: "text", text: part.text }],
        }
-      else
-        messages.push(
-          options.cacheControl?.(part.cache)
-            ? {
-                role: "user",
-                content: [{ type: "text", text: part.text, cache_control: options.cacheControl(part.cache) }],
-              }
-            : { role: "user", content: part.text },
-        )
+      else messages.push({ role: "user", content: part.text })
      continue
    }
    if (message.role === "tool") {
-      const lowered = yield* lowerToolMessages(message, options)
+      const lowered = yield* lowerToolMessages(message)
      messages.push(...lowered.messages)
      pendingImages.push(...lowered.images)
      continue
    }
    flushImages()
-    messages.push(...(yield* lowerMessage(message, request.model.compatibility?.reasoningField, options)))
+    messages.push(...(yield* lowerMessage(message, request.model.compatibility?.reasoningField)))
  }
  flushImages()
  return messages
@@ -494,10 +405,7 @@ const lowerOptions = (request: LLMRequest) => {
  }
 }

-export const fromRequest = Effect.fn("OpenAIChat.fromRequest")(function* (
-  request: LLMRequest,
-  options: LoweringOptions = {},
-) {
+const fromRequest = Effect.fn("OpenAIChat.fromRequest")(function* (request: LLMRequest) {
  // `fromRequest` returns the provider body only. Endpoint, auth, framing,
  // validation, and HTTP execution are composed by `Route.make`.
  const reasoningField = request.model.compatibility?.reasoningField
@@ -507,26 +415,19 @@ export const fromRequest = Effect.fn("OpenAIChat.fromRequest")(function* (
    )
  const generation = request.generation
  const toolSchemaCompatibility = request.model.compatibility?.toolSchema
-  const maxTokensField = request.model.compatibility?.maxTokensField ?? "max_tokens"
  return {
    model: request.model.id,
-    messages: yield* lowerMessages(request, options),
+    messages: yield* lowerMessages(request),
    tools:
      request.tools.length === 0
        ? undefined
        : request.tools.map((tool) =>
-            lowerTool(
-              tool,
-              ToolSchemaProjection.modelCompatibility(tool.inputSchema, toolSchemaCompatibility),
-              options,
-            ),
+            lowerTool(tool, ToolSchemaProjection.modelCompatibility(tool.inputSchema, toolSchemaCompatibility)),
          ),
    tool_choice: request.toolChoice ? yield* lowerToolChoice(request.toolChoice) : undefined,
    stream: true as const,
    stream_options: { include_usage: true },
-    ...(maxTokensField === "max_completion_tokens"
-      ? { max_completion_tokens: generation?.maxTokens }
-      : { max_tokens: generation?.maxTokens }),
+    max_tokens: generation?.maxTokens,
    temperature: generation?.temperature,
    top_p: generation?.topP,
    frequency_penalty: generation?.frequencyPenalty,
@@ -555,7 +456,7 @@ const mapFinishReason = (reason: string | null | undefined): FinishReason => {
 // OpenAI Chat reports `prompt_tokens` (inclusive total) with a
 // cached-read and cache-write subsets, and `completion_tokens` (inclusive
 // total) with a `reasoning_tokens` subset. We pass the inclusive totals
-// through and derive the non-cached breakdown so the `AI.Usage` contract is
+// through and derive the non-cached breakdown so the `LLM.Usage` contract is
 // satisfied on both sides.
 const mapUsage = (usage: OpenAIChatEvent["usage"]): Usage | undefined => {
  if (!usage) return undefined
@@ -645,7 +546,7 @@ const reasoningMetadata = (field: ParserState["reasoningField"], details?: Reado
 const step = (state: ParserState, event: OpenAIChatEvent) =>
  Effect.gen(function* () {
    if (event.error)
-      return yield* new AIError({
+      return yield* new LLMError({
        module: ADAPTER,
        method: "stream",
        reason: classifyProviderFailure({
@@ -1,11 +1,11 @@
-import { Route, type RouteRoutedLanguageModelInput } from "../route/client"
+import { Route, type RouteRoutedModelInput } from "../route/client"
 import { Endpoint } from "../route/endpoint"
 import { Framing } from "../route/framing"
 import * as OpenAIChat from "./openai-chat"

 const ADAPTER = "openai-compatible-chat"

-export type OpenAICompatibleChatLanguageModelInput = RouteRoutedLanguageModelInput
+export type OpenAICompatibleChatModelInput = RouteRoutedModelInput

 /**
 * Route for non-OpenAI providers that expose an OpenAI Chat-compatible
@@ -1,10 +1,10 @@
-import { Route, type RouteRoutedLanguageModelInput } from "../route/client"
+import { Route, type RouteRoutedModelInput } from "../route/client"
 import { Endpoint } from "../route/endpoint"
 import { OpenResponses } from "./open-responses"

 const ADAPTER = "openai-compatible-responses"

-export type OpenAICompatibleResponsesLanguageModelInput = RouteRoutedLanguageModelInput
+export type OpenAICompatibleResponsesModelInput = RouteRoutedModelInput

 /**
 * Deployment adapter for providers that expose an Open Responses-compatible
@@ -11,7 +11,7 @@ import {
 import { Auth, type Definition as AuthDefinition } from "../route/auth"
 import {
  InvalidProviderOutputReason,
-  AIError,
+  LLMError,
  Usage,
  mergeHttpOptions,
  mergeJsonRecords,
@@ -85,7 +85,7 @@ const nativeOptions = (options: OpenAIImageOptions | undefined) => {
 }

 const invalidOutput = (message: string) =>
-  new AIError({
+  new LLMError({
    module: ADAPTER,
    method: "generate",
    reason: new InvalidProviderOutputReason({ message, route: ADAPTER }),
@@ -6,7 +6,7 @@ import { Headers, HttpClientRequest } from "effect/unstable/http"
 import {
  InvalidProviderOutputReason,
  InvalidRequestReason,
-  AIError,
+  LLMError,
  type ContentPart,
  type LLMRequest,
  type MediaPart,
@@ -41,7 +41,7 @@ export interface ToolAccumulator {
 * when at least one is defined. Returns `undefined` when neither input nor
 * output is known so routes don't publish a misleading `0`.
 *
- * Under the additive `AI.Usage` contract, `inputTokens` and `outputTokens`
+ * Under the additive `LLM.Usage` contract, `inputTokens` and `outputTokens`
 * are the non-cached input and visible output only. The provider-supplied
 * `total` is the source of truth when present; the computed fallback
 * under-counts cache and reasoning by design and exists mainly so
@@ -88,7 +88,7 @@ export const sumTokens = (...values: ReadonlyArray<number | undefined>): number
 }

 export const eventError = (route: string, message: string, raw?: string) =>
-  new AIError({
+  new LLMError({
    module: "ProviderShared",
    method: "stream",
    reason: new InvalidProviderOutputReason({ route, message, raw }),
@@ -238,9 +238,9 @@ export const errorText = (error: unknown) => {
 * `decodeChunk` sees one JSON string per element. The SSE channel emits a
 * `Retry` control event on its error channel; we drop it here (we don't
 * implement client-driven retries) so the public error channel stays
- * `AIError`.
+ * `LLMError`.
 */
-export const sseFraming = (bytes: Stream.Stream<Uint8Array, AIError>): Stream.Stream<string, AIError> =>
+export const sseFraming = (bytes: Stream.Stream<Uint8Array, LLMError>): Stream.Stream<string, LLMError> =>
  bytes.pipe(
    Stream.decodeText(),
    Stream.pipeThroughChannel(Sse.decode()),
@@ -257,7 +257,7 @@ export const sseFraming = (bytes: Stream.Stream<Uint8Array, AIError>): Stream.St
 * lands here.
 */
 export const invalidRequest = (message: string) =>
-  new AIError({
+  new LLMError({
    module: "ProviderShared",
    method: "request",
    reason: new InvalidRequestReason({ message }),
@@ -304,7 +304,7 @@ export const unsupportedContent = (
 * Build a `validate` step from a Schema decoder. Replaces the per-route
 * lambda body `(payload) => decode(payload).pipe(Effect.mapError((e) =>
 * invalid(e.message)))`. Any decode error is translated into
- * `AIError` carrying the original parse-error message.
+ * `LLMError` carrying the original parse-error message.
 */
 export const validateWith =
  <A, I, E extends { readonly message: string }>(decode: (input: I) => Effect.Effect<A, E>) =>
@@ -1,9 +1,9 @@
 import { Effect, Encoding } from "effect"
 import type { ImageInput } from "../../image"
-import { InvalidRequestReason, AIError } from "../../schema"
+import { InvalidRequestReason, LLMError } from "../../schema"

 const invalid = (module: string, message: string) =>
-  new AIError({
+  new LLMError({
    module,
    method: "generate",
    reason: new InvalidRequestReason({ message }),
@@ -15,7 +15,7 @@ export const dataUrl = (input: Extract<ImageInput, { readonly type: "bytes" }>)
 export const decodeDataUrl = (
  url: string,
  module: string,
-): Effect.Effect<{ readonly mediaType: string; readonly data: Uint8Array } | undefined, AIError> => {
+): Effect.Effect<{ readonly mediaType: string; readonly data: Uint8Array } | undefined, LLMError> => {
  if (!url.startsWith("data:")) return Effect.succeed(undefined)
  const match = /^data:([^;,]+);base64,(.*)$/s.exec(url)
  if (!match) return Effect.fail(invalid(module, "Image data URLs must contain a MIME type and base64 data"))
@@ -1,4 +1,4 @@
-import type { JsonSchema, LanguageModelToolSchemaCompatibility } from "../../schema"
+import type { JsonSchema, ModelToolSchemaCompatibility } from "../../schema"
 import { isRecord } from "../../utils/record"
 import { GeminiToolSchema } from "./gemini-tool-schema"

@@ -69,7 +69,7 @@ const gemini = (schema: JsonSchema): JsonSchema => GeminiToolSchema.convert(sche

 const modelCompatibility = (
  schema: JsonSchema,
-  compatibility: LanguageModelToolSchemaCompatibility | undefined,
+  compatibility: ModelToolSchemaCompatibility | undefined,
 ): JsonSchema => {
  if (compatibility === undefined) return schema
  switch (compatibility) {
@@ -1,5 +1,5 @@
 import { Effect } from "effect"
-import { AIError, LLMEvent, type ProviderMetadata, type ToolCall, type ToolInputError } from "../../schema"
+import { LLMError, LLMEvent, type ProviderMetadata, type ToolCall, type ToolInputError } from "../../schema"
 import { eventError, parseToolInput, type ToolAccumulator } from "../shared"

 type StreamKey = string | number
@@ -112,8 +112,8 @@ const appendTool = <K extends StreamKey>(
  }
 }

-export const isError = <K extends StreamKey>(result: AppendOutcome<K> | AIError): result is AIError =>
-  result instanceof AIError
+export const isError = <K extends StreamKey>(result: AppendOutcome<K> | LLMError): result is LLMError =>
+  result instanceof LLMError

 /**
 * Register a tool call whose start event arrived before any argument deltas.
@@ -138,7 +138,7 @@ export const appendOrStart = <K extends StreamKey>(
  key: K,
  delta: { readonly id?: string; readonly name?: string; readonly text: string },
  missingToolMessage: string,
-): AppendOutcome<K> | AIError => {
+): AppendOutcome<K> | LLMError => {
  const current = tools[key]
  const id = current?.id ?? delta.id
  const name = current?.name ?? delta.name
@@ -167,7 +167,7 @@ export const appendExisting = <K extends StreamKey>(
  key: K,
  text: string,
  missingToolMessage: string,
-): AppendOutcome<K> | AIError => {
+): AppendOutcome<K> | LLMError => {
  const current = tools[key]
  if (!current) return eventError(route, missingToolMessage)
  if (text.length === 0) return { tools, tool: current, events: [] }
@@ -4,7 +4,7 @@ import { GeneratedImage, ImageModel, ImageResponse, type ImageRequestFor, type I
 import { Auth, type Definition as AuthDefinition } from "../route/auth"
 import {
  InvalidProviderOutputReason,
-  AIError,
+  LLMError,
  Usage,
  mergeHttpOptions,
  mergeJsonRecords,
@@ -95,7 +95,7 @@ const nativeOptions = (options: XAIImageOptions | undefined) => {
 }

 const invalidOutput = (message: string) =>
-  new AIError({
+  new LLMError({
    module: ADAPTER,
    method: "generate",
    reason: new InvalidProviderOutputReason({ message, route: ADAPTER }),
@@ -2,7 +2,7 @@ import { Effect, Schema } from "effect"
 import { Headers, HttpClientRequest } from "effect/unstable/http"
 import { GeneratedImage, ImageModel, ImageResponse, type ImageRequestFor, type ImageRoute } from "../image"
 import { Auth, type Definition as AuthDefinition } from "../route/auth"
-import { InvalidProviderOutputReason, AIError, mergeHttpOptions, mergeJsonRecords, type HttpOptions } from "../schema"
+import { InvalidProviderOutputReason, LLMError, mergeHttpOptions, mergeJsonRecords, type HttpOptions } from "../schema"
 import { ProviderShared } from "./shared"
 import { ImageInputs } from "./utils/image-input"

@@ -58,7 +58,7 @@ const nativeOptions = (options: ZAIImageOptions | undefined) => {
 }

 const invalidOutput = (message: string) =>
-  new AIError({
+  new LLMError({
    module: ADAPTER,
    method: "generate",
    reason: new InvalidProviderOutputReason({ message, route: ADAPTER }),
@@ -3,7 +3,7 @@ import {
  AuthenticationReason,
  ContentPolicyReason,
  InvalidRequestReason,
-  AIError,
+  LLMError,
  ProviderErrorEvent,
  ProviderInternalReason,
  QuotaExceededReason,
@@ -16,6 +16,7 @@ import {

 const patterns = [
  /prompt is too long/i,
+  /request_too_large/i,
  /input is too long for requested model/i,
  /exceeds the context window/i,
  /exceeds (?:the )?(?:model'?s )?maximum context length(?: of [\d,]+ tokens?|\s*\([\d,]+\))/i,
@@ -32,6 +33,7 @@ const patterns = [
  /context window exceeds limit/i,
  /exceeded model token limit/i,
  /context[_ ]length[_ ]exceeded/i,
+  /request entity too large/i,
  /context length is only \d+ tokens/i,
  /input length.*exceeds.*context length/i,
  /prompt too long; exceeded (?:max )?context length/i,
@@ -42,18 +44,14 @@ const patterns = [
  /token limit exceeded/i,
 ]

-const payloadPatterns = [/request_too_large/i, /request entity too large/i, /payload too large/i, /request too large/i]
-
 const exclusions = [/^(throttling error|service unavailable):/i, /rate limit/i, /too many requests/i]

 export const isContextOverflow = (message: string) =>
  !exclusions.some((pattern) => pattern.test(message)) &&
-  (patterns.some((pattern) => pattern.test(message)) || /^400\s*(status code)?\s*\(no body\)/i.test(message))
-
-export const isPayloadTooLarge = (message: string) => payloadPatterns.some((pattern) => pattern.test(message))
+  (patterns.some((pattern) => pattern.test(message)) || /^4(00|13)\s*(status code)?\s*\(no body\)/i.test(message))

 export const isContextOverflowFailure = (failure: unknown) =>
-  failure instanceof AIError
+  failure instanceof LLMError
    ? failure.reason._tag === "InvalidRequest" && failure.reason.classification === "context-overflow"
    : Schema.is(ProviderErrorEvent)(failure) && failure.classification === "context-overflow"

@@ -86,7 +84,7 @@ export interface ProviderFailure {

 // Keep HTTP failures and provider-reported stream failures on one typed path so
 // session retry policy never needs provider-specific string matching.
-export function classifyProviderFailure(input: ProviderFailure): AIError["reason"] {
+export function classifyProviderFailure(input: ProviderFailure): LLMError["reason"] {
  const body = input.http?.body ?? ""
  const codes = [input.code, ...providerCodes(body), ...providerCodes(input.message)]
    .filter((code): code is string => code !== undefined)
@@ -102,8 +100,6 @@ export function classifyProviderFailure(input: ProviderFailure): AIError["reason
      isContextOverflow(text))
  )
    return new InvalidRequestReason({ ...common, classification: "context-overflow" })
-  if (input.status === 413 || isPayloadTooLarge(text))
-    return new InvalidRequestReason({ ...common, classification: "payload-too-large" })
  if (CONTENT_POLICY_TEXT.test(text)) return new ContentPolicyReason(common)
  if (codes.some((code) => QUOTA_CODES.has(code)) || (input.status === 429 && QUOTA_TEXT.test(text)))
    return new QuotaExceededReason(common)
@@ -146,7 +142,12 @@ export function classifyProviderFailure(input: ProviderFailure): AIError["reason
      retryAfterMs: input.retryAfterMs,
    })
  if (codes.some((code) => INVALID_REQUEST_CODES.has(code))) return new InvalidRequestReason(common)
-  if (input.status === 400 || input.status === 404 || input.status === 413 || input.status === 422)
+  if (
+    input.status === 400 ||
+    input.status === 404 ||
+    input.status === 413 ||
+    input.status === 422
+  )
    return new InvalidRequestReason(common)
  return new UnknownProviderReason({ ...common, status: input.status })
 }
@@ -1,21 +1,16 @@
-import type { LanguageModel, ProviderOptions } from "./schema"
+import type { Model } from "./schema"

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

-export interface Definition<
-  ProviderSettings extends Settings = Settings,
-  Options extends ProviderOptions = ProviderOptions,
-> {
-  readonly model: (modelID: string, settings: ProviderSettings) => LanguageModel<Options>
+export interface Definition<ProviderSettings extends Settings = Settings> {
+  readonly model: (modelID: string, settings: ProviderSettings) => Model
 }

 export * as ProviderPackage from "./provider-package"
@@ -1,6 +1,6 @@
-import type { LanguageModel, ModelID, ProviderID } from "./schema"
+import type { Model, ModelID, ProviderID } from "./schema"

-export type LanguageModelOptions = Pick<LanguageModel.Input, "defaults" | "compatibility">
+export type ModelOptions = Pick<Model.Input, "defaults" | "compatibility">

 /**
 * Advanced structural provider definition helper. Built-in providers should
@@ -8,23 +8,23 @@ export type LanguageModelOptions = Pick<LanguageModel.Input, "defaults" | "compa
 * chosen before model selection. The optional `apis` map remains for external
 * structural providers that expose multiple route selectors behind one provider.
 */
-export type LanguageModelFactory<Options extends LanguageModelOptions = LanguageModelOptions> = (
+export type ModelFactory<Options extends ModelOptions = ModelOptions> = (
  id: string | ModelID,
  options?: Options,
-) => LanguageModel
+) => Model

-type AnyLanguageModelFactory = (...args: never[]) => LanguageModel
+type AnyModelFactory = (...args: never[]) => Model

-export interface Definition<Factory extends AnyLanguageModelFactory = LanguageModelFactory> {
+export interface Definition<Factory extends AnyModelFactory = ModelFactory> {
  readonly id: ProviderID
  readonly model: Factory
-  readonly apis?: Record<string, AnyLanguageModelFactory>
+  readonly apis?: Record<string, AnyModelFactory>
 }

 type DefinitionShape = {
  readonly id: ProviderID
-  readonly model: (...args: never[]) => LanguageModel
-  readonly apis?: Record<string, (...args: never[]) => LanguageModel>
+  readonly model: (...args: never[]) => Model
+  readonly apis?: Record<string, (...args: never[]) => Model>
 }

 type NoExtraFields<Input, Shape> = Input & Record<Exclude<keyof Input, keyof Shape>, never>
@@ -47,7 +47,7 @@ export const configure = (input: Config) => {
  })
  return {
    id: ProviderID.make(provider),
-    model: (modelID: string | ModelID) => route.model<AnthropicMessages.ProviderOptionsInput>({ id: modelID }),
+    model: (modelID: string | ModelID) => route.model({ id: modelID }),
    configure,
  }
 }
@@ -57,10 +57,7 @@ export const provider = {
  configure,
 }

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

 export const provider = configure()
-export const model: ProviderPackage.Definition<Settings, AnthropicMessages.ProviderOptionsInput>["model"] = (
-  modelID,
-  settings,
-) => {
+export const model: ProviderPackage.Definition<Settings>["model"] = (modelID, settings) => {
  if (settings.apiKey !== undefined && settings.authToken !== undefined)
    throw new Error("Anthropic apiKey cannot be combined with authToken")
  return configure({
@@ -14,7 +14,7 @@ const routeAuth = Auth.remove("authorization")
 // (helper builds the URL) or `baseURL` directly.
 type AzureURL = AtLeastOne<{ readonly resourceName: string; readonly baseURL: string }>

-export type LanguageModelOptions = AzureURL &
+export type ModelOptions = AzureURL &
  RouteDefaultsInput &
  ProviderAuthOption<"optional"> & {
    readonly apiVersion?: string
@@ -22,7 +22,7 @@ export type LanguageModelOptions = AzureURL &
    readonly useCompletionUrls?: boolean
    readonly providerOptions?: OpenAIProviderOptionsInput
  }
-export type Config = LanguageModelOptions
+export type Config = ModelOptions

 export type Settings = ProviderPackage.Settings &
  AzureURL & {
@@ -99,14 +99,10 @@ export const configure = (input: Config) => {
  const modelDefaults = defaults(input)

  const responses = (modelID: string | ModelID) =>
-    configuredResponsesRoute
-      .with(withOpenAIOptions(modelID, modelDefaults))
-      .model<OpenAIProviderOptionsInput>({ id: modelID })
+    configuredResponsesRoute.with(withOpenAIOptions(modelID, modelDefaults)).model({ id: modelID })

  const chat = (modelID: string | ModelID) =>
-    configuredChatRoute
-      .with(withOpenAIOptions(modelID, modelDefaults))
-      .model<OpenAIProviderOptionsInput>({ id: modelID })
+    configuredChatRoute.with(withOpenAIOptions(modelID, modelDefaults)).model({ id: modelID })

  return {
    id,
@@ -137,12 +133,8 @@ const config = (settings: Settings): Config => {
  throw new Error("Azure requires resourceName or baseURL")
 }

-export const responsesModel: ProviderPackage.Definition<Settings, OpenAIProviderOptionsInput>["model"] = (
-  modelID,
-  settings,
-) => configure(config(settings)).responses(modelID)
-export const chatModel: ProviderPackage.Definition<Settings, OpenAIProviderOptionsInput>["model"] = (
-  modelID,
-  settings,
-) => configure(config(settings)).chat(modelID)
+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 model = responsesModel
@@ -4,7 +4,6 @@ import { Auth } from "../route/auth"
 import { AuthOptions, type AtLeastOne, type ProviderAuthOption } from "../route/auth-options"
 import type { RouteDefaultsInput } from "../route/client"
 import { ProviderID, type ModelID } from "../schema"
-import type { OpenAIProviderOptionsInput } from "./openai-options"

 export const aiGatewayID = ProviderID.make("cloudflare-ai-gateway")
 export const workersAIID = ProviderID.make("cloudflare-workers-ai")
@@ -21,11 +20,10 @@ type GatewayURL = AtLeastOne<{
 }

 export type AIGatewayOptions = GatewayURL &
-  Omit<RouteDefaultsInput, "providerOptions"> &
+  RouteDefaultsInput &
  ProviderAuthOption<"optional"> & {
    /** Cloudflare AI Gateway authentication token. Sent as `cf-aig-authorization`. */
    readonly gatewayApiKey?: CloudflareSecret
-    readonly providerOptions?: OpenAIProviderOptionsInput
  }

 type WorkersAIURL = AtLeastOne<{
@@ -33,11 +31,7 @@ type WorkersAIURL = AtLeastOne<{
  readonly baseURL: string
 }>

-export type WorkersAIOptions = WorkersAIURL &
-  Omit<RouteDefaultsInput, "providerOptions"> &
-  ProviderAuthOption<"optional"> & {
-    readonly providerOptions?: OpenAIProviderOptionsInput
-  }
+export type WorkersAIOptions = WorkersAIURL & RouteDefaultsInput & ProviderAuthOption<"optional">

 export const aiGatewayBaseURL = (input: GatewayURL) => {
  if (input.baseURL) return input.baseURL
@@ -104,7 +98,7 @@ const configureAIGateway = (options: AIGatewayOptions) => {
  })
  return {
    id: aiGatewayID,
-    model: (modelID: string | ModelID) => route.model<OpenAIProviderOptionsInput>({ id: modelID }),
+    model: (modelID: string | ModelID) => route.model({ id: modelID }),
    configure: configureAIGateway,
  }
 }
@@ -117,7 +111,7 @@ const configureWorkersAI = (options: WorkersAIOptions) => {
  })
  return {
    id: workersAIID,
-    model: (modelID: string | ModelID) => route.model<OpenAIProviderOptionsInput>({ id: modelID }),
+    model: (modelID: string | ModelID) => route.model({ id: modelID }),
    configure: configureWorkersAI,
  }
 }
@@ -9,14 +9,14 @@ export const id = ProviderID.make("github-copilot")

 // GitHub Copilot has no canonical public URL — callers (opencode, etc.) must
 // supply `baseURL` explicitly.
-export type LanguageModelOptions = Omit<RouteDefaultsInput, "providerOptions"> &
+export type ModelOptions = Omit<RouteDefaultsInput, "providerOptions"> &
  ProviderAuthOption<"optional"> & {
    readonly baseURL: string
    readonly endpoint?: "chat" | "responses"
    readonly providerOptions?: OpenAIProviderOptionsInput
  }

-export const shouldUseResponsesApi = (modelID: string | ModelID, endpoint?: LanguageModelOptions["endpoint"]) => {
+export const shouldUseResponsesApi = (modelID: string | ModelID, endpoint?: ModelOptions["endpoint"]) => {
  if (endpoint) return endpoint === "responses"
  const model = String(modelID)
  const match = /^gpt-(\d+)/.exec(model)
@@ -29,32 +29,30 @@ export const routes = [OpenAIResponses.route, OpenAIChat.route]
 const chatRoute = OpenAIChat.route.with({ provider: id })
 const responsesRoute = OpenAIResponses.route.with({ provider: id })

-const defaults = (options: LanguageModelOptions) => {
+const defaults = (options: ModelOptions) => {
  const { apiKey: _, auth: _auth, baseURL: _baseURL, endpoint: _endpoint, ...rest } = options
  return rest
 }

-const configuredResponsesRoute = (options: LanguageModelOptions) =>
+const configuredResponsesRoute = (options: ModelOptions) =>
  responsesRoute.with({
    endpoint: { baseURL: options.baseURL },
    auth: AuthOptions.bearer(options, []),
  })

-const configuredChatRoute = (options: LanguageModelOptions) =>
+const configuredChatRoute = (options: ModelOptions) =>
  chatRoute.with({
    endpoint: { baseURL: options.baseURL },
    auth: AuthOptions.bearer(options, []),
  })

-export const configure = (options: LanguageModelOptions) => {
+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<OpenAIProviderOptionsInput>({ id: modelID })
+    responsesRoute.with(withOpenAIOptions(modelID, defaults(options))).model({ id: modelID })
  const chat = (modelID: string | ModelID) =>
-    chatRoute.with(withOpenAIOptions(modelID, defaults(options))).model<OpenAIProviderOptionsInput>({ id: modelID })
+    chatRoute.with(withOpenAIOptions(modelID, defaults(options))).model({ id: modelID })
  return {
    id,
    model: (modelID: string | ModelID) =>
@@ -1,9 +1,8 @@
 import type { ProviderPackage } from "../provider-package"
 import { OpenAICompatibleChat } from "../protocols/openai-compatible-chat"
 import type { RouteDefaultsInput } from "../route/client"
-import { ProviderID, type ModelID } from "../schema"
+import { ProviderID, type ModelID, type ProviderOptions } from "../schema"
 import { GoogleVertexShared } from "./google-vertex-shared"
-import type { OpenAIProviderOptionsInput } from "./openai-options"

 export const id = ProviderID.make("google-vertex")

@@ -12,7 +11,6 @@ export type Config = RouteDefaultsInput &
    readonly baseURL?: string
    readonly location?: string
    readonly project?: string
-    readonly providerOptions?: OpenAIProviderOptionsInput
  }

 export interface Settings extends ProviderPackage.Settings {
@@ -21,7 +19,7 @@ export interface Settings extends ProviderPackage.Settings {
  readonly baseURL?: string
  readonly location?: string
  readonly project?: string
-  readonly providerOptions?: OpenAIProviderOptionsInput
+  readonly providerOptions?: ProviderOptions
 }

 const route = OpenAICompatibleChat.route.with({
@@ -58,7 +56,7 @@ export const configure = (input: Config = {}) => {
  const route = configuredRoute(input)
  return {
    id,
-    model: (modelID: string | ModelID) => route.model<OpenAIProviderOptionsInput>({ id: modelID }),
+    model: (modelID: string | ModelID) => route.model({ id: modelID }),
    configure,
  }
 }
@@ -68,7 +66,7 @@ export const provider = {
  configure,
 }

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

-export const model: ProviderPackage.Definition<Settings, AnthropicMessages.ProviderOptionsInput>["model"] = (
-  modelID,
-  settings,
-) => {
+export const model: ProviderPackage.Definition<Settings>["model"] = (modelID, settings) => {
  if (settings.apiKey !== undefined) throw new Error("Google Vertex Messages does not support API keys")
  return configure({
    accessToken: settings.accessToken,
@@ -1,9 +1,8 @@
 import type { ProviderPackage } from "../provider-package"
 import { OpenAICompatibleResponses } from "../protocols/openai-compatible-responses"
 import type { RouteDefaultsInput } from "../route/client"
-import { ProviderID, type ModelID } from "../schema"
+import { ProviderID, type ModelID, type ProviderOptions } from "../schema"
 import { GoogleVertexShared } from "./google-vertex-shared"
-import type { OpenResponsesProviderOptionsInput } from "./open-responses-options"

 export const id = ProviderID.make("google-vertex")

@@ -12,7 +11,6 @@ export type Config = RouteDefaultsInput &
    readonly baseURL?: string
    readonly location?: string
    readonly project?: string
-    readonly providerOptions?: OpenResponsesProviderOptionsInput
  }

 export interface Settings extends ProviderPackage.Settings {
@@ -21,7 +19,7 @@ export interface Settings extends ProviderPackage.Settings {
  readonly baseURL?: string
  readonly location?: string
  readonly project?: string
-  readonly providerOptions?: OpenResponsesProviderOptionsInput
+  readonly providerOptions?: ProviderOptions
 }

 const route = OpenAICompatibleResponses.route.with({
@@ -60,7 +58,7 @@ export const configure = (input: Config = {}) => {
  const route = configuredRoute(input)
  return {
    id,
-    model: (modelID: string | ModelID) => route.model<OpenResponsesProviderOptionsInput>({ id: modelID }),
+    model: (modelID: string | ModelID) => route.model({ id: modelID }),
    configure,
  }
 }
@@ -70,10 +68,7 @@ export const provider = {
  configure,
 }

-export const model: ProviderPackage.Definition<Settings, OpenResponsesProviderOptionsInput>["model"] = (
-  modelID,
-  settings,
-) => {
+export const model: ProviderPackage.Definition<Settings>["model"] = (modelID, settings) => {
  if (settings.apiKey !== undefined) throw new Error("Google Vertex Responses does not support API keys")
  return configure({
    accessToken: settings.accessToken,
@@ -77,8 +77,7 @@ const configuredRoute = (input: Config, modelID: string | ModelID) => {
 export const configure = (input: Config = {}) => {
  return {
    id,
-    model: (modelID: string | ModelID) =>
-      configuredRoute(input, modelID).model<Gemini.ProviderOptionsInput>({ id: modelID }),
+    model: (modelID: string | ModelID) => configuredRoute(input, modelID).model({ id: modelID }),
    configure,
  }
 }
@@ -87,10 +86,7 @@ export const provider = {
  id,
  configure,
 }
-export const model: ProviderPackage.Definition<Settings, Gemini.ProviderOptionsInput>["model"] = (
-  modelID,
-  settings,
-) => {
+export const model: ProviderPackage.Definition<Settings>["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({
@@ -50,14 +50,14 @@ export const configure = (input: Config = {}) => {
    })
  return {
    id,
-    model: (modelID: string | ModelID) => route.model<Gemini.ProviderOptionsInput>({ id: modelID }),
+    model: (modelID: string | ModelID) => route.model({ id: modelID }),
    image,
    configure,
  }
 }

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

-export const model: ProviderPackage.Definition<Settings, OpenResponsesProviderOptionsInput>["model"] = (
-  modelID,
-  settings,
-) =>
+export const model: ProviderPackage.Definition<Settings>["model"] = (modelID, settings) =>
  configure({
    apiKey: settings.apiKey,
    baseURL: settings.baseURL,
@@ -4,15 +4,13 @@ import type { RouteDefaultsInput } from "../route/client"
 import { AuthOptions, type ProviderAuthOption } from "../route/auth-options"
 import type { ProviderPackage } from "../provider-package"
 import { profiles, type OpenAICompatibleProfile } from "./openai-compatible-profile"
-import type { OpenAIProviderOptionsInput } from "./openai-options"

 export const id = ProviderID.make("openai-compatible")

-type GenericModelOptions = Omit<RouteDefaultsInput, "providerOptions"> &
+type GenericModelOptions = RouteDefaultsInput &
  ProviderAuthOption<"optional"> & {
    readonly provider?: string
    readonly baseURL: string
-    readonly providerOptions?: OpenAIProviderOptionsInput
  }

 export interface Settings extends ProviderPackage.Settings {
@@ -21,10 +19,9 @@ export interface Settings extends ProviderPackage.Settings {
  readonly provider?: string
 }

-export type FamilyModelOptions = Omit<RouteDefaultsInput, "providerOptions"> &
+export type FamilyModelOptions = RouteDefaultsInput &
  ProviderAuthOption<"optional"> & {
    readonly baseURL?: string
-    readonly providerOptions?: OpenAIProviderOptionsInput
  }

 export const routes = [OpenAICompatibleChat.route]
@@ -40,8 +37,7 @@ export const configure = (input: GenericModelOptions) => {
  })
  return {
    id: ProviderID.make(provider),
-    model: (modelID: string | ModelID) =>
-      route.model<OpenAIProviderOptionsInput>({ id: modelID, provider: ProviderID.make(provider) }),
+    model: (modelID: string | ModelID) => route.model({ id: modelID, provider: ProviderID.make(provider) }),
    configure,
  }
 }
@@ -67,7 +63,7 @@ export const provider = {
  configure,
 }

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

-export const model: ProviderPackage.Definition<Settings, OpenAIProviderOptionsInput>["model"] = (modelID, settings) => {
+export const model: ProviderPackage.Definition<Settings>["model"] = (modelID, settings) => {
  const configured = configure(config(settings))
  if (settings.transport === undefined || settings.transport === "http") return configured.responses(modelID)
  if (settings.transport === "websocket") return configured.responsesWebSocket(modelID)
  throw new Error(`Unsupported OpenAI Responses transport: ${String(settings.transport)}`)
 }

-export const chatModel: ProviderPackage.Definition<Settings, OpenAIProviderOptionsInput>["model"] = (
-  modelID,
-  settings,
-) => configure(config(settings)).chat(modelID)
+export const chatModel: ProviderPackage.Definition<Settings>["model"] = (modelID, settings) =>
+  configure(config(settings)).chat(modelID)
 export const responses = provider.responses
 export const responsesWebSocket = provider.responsesWebSocket
 export const chat = provider.chat
@@ -4,90 +4,32 @@ import { Endpoint } from "../route/endpoint"
 import { Framing } from "../route/framing"
 import { Protocol } from "../route/protocol"
 import { AuthOptions, type ProviderAuthOption } from "../route/auth-options"
-import { ProviderID, type CacheHint, type ModelID, type ProviderOptions } from "../schema"
-import type { ProviderPackage } from "../provider-package"
+import { ProviderID, type ModelID, type ProviderOptions } from "../schema"
 import * as OpenAICompatibleProfiles from "./openai-compatible-profile"
 import * as OpenAIChat from "../protocols/openai-chat"
-import { newBreakpoints, ttlBucket } from "../protocols/utils/cache"
 import { isRecord } from "../protocols/shared"

 export const profile = OpenAICompatibleProfiles.profiles.openrouter
 export const id = ProviderID.make(profile.provider)
 const ADAPTER = "openrouter"

-type OpenRouterString<Known extends string> = Known | (string & {})
-
-export interface OpenRouterProviderRouting {
-  readonly [key: string]: unknown
-  readonly order?: ReadonlyArray<string>
-  readonly allow_fallbacks?: boolean
-  readonly require_parameters?: boolean
-  readonly data_collection?: OpenRouterString<"allow" | "deny">
-  readonly only?: ReadonlyArray<string>
-  readonly ignore?: ReadonlyArray<string>
-  readonly quantizations?: ReadonlyArray<string>
-  readonly sort?: OpenRouterString<"price" | "throughput" | "latency">
-  readonly max_price?: Readonly<{
-    prompt?: number | string
-    completion?: number | string
-    image?: number | string
-    audio?: number | string
-    request?: number | string
-  }>
-  readonly zdr?: boolean
-}
-
-export type OpenRouterPlugin =
-  | Readonly<{
-      id: "web"
-      max_results?: number
-      search_prompt?: string
-      engine?: OpenRouterString<"native" | "exa">
-    }>
-  | Readonly<{ id: "file-parser"; max_files?: number; pdf?: { engine?: string } }>
-  | Readonly<{ id: "moderation" }>
-  | Readonly<{ id: "response-healing" }>
-  | Readonly<{ id: "auto-router"; allowed_models?: ReadonlyArray<string> }>
-  | Readonly<{ id: string & {}; [key: string]: unknown }>
-
 export interface OpenRouterOptions {
  readonly [key: string]: unknown
-  readonly debug?: Readonly<{ echo_upstream_body?: boolean }>
-  readonly models?: ReadonlyArray<string>
-  readonly plugins?: ReadonlyArray<OpenRouterPlugin>
+  readonly usage?: boolean | Record<string, unknown>
+  readonly reasoning?: Record<string, unknown>
  readonly promptCacheKey?: string
-  readonly provider?: OpenRouterProviderRouting
-  readonly reasoning?: Readonly<{
-    enabled?: boolean
-    exclude?: boolean
-    effort?: OpenRouterString<"none" | "minimal" | "low" | "medium" | "high" | "xhigh" | "max">
-    max_tokens?: number
-  }>
-  readonly usage?: boolean | Readonly<{ include: boolean }>
-  readonly user?: string
-  readonly web_search_options?: Readonly<{
-    max_results?: number
-    search_prompt?: string
-    engine?: OpenRouterString<"native" | "exa">
-  }>
 }

 export type OpenRouterProviderOptionsInput = ProviderOptions & {
  readonly openrouter?: OpenRouterOptions
 }

-export type LanguageModelOptions = Omit<RouteDefaultsInput, "providerOptions"> &
+export type ModelOptions = Omit<RouteDefaultsInput, "providerOptions"> &
  ProviderAuthOption<"optional"> & {
    readonly baseURL?: string
    readonly providerOptions?: OpenRouterProviderOptionsInput
  }

-export interface Settings extends ProviderPackage.Settings {
-  readonly apiKey?: string
-  readonly baseURL?: string
-  readonly providerOptions?: OpenRouterProviderOptionsInput
-}
-
 const OpenRouterBody = Schema.StructWithRest(Schema.Struct(OpenAIChat.bodyFields), [
  Schema.Record(Schema.String, Schema.Any),
 ])
@@ -98,7 +40,7 @@ export const protocol = Protocol.make({
  body: {
    schema: OpenRouterBody,
    from: (request) =>
-      OpenAIChat.fromRequest(request, { cacheControl: cacheControl() }).pipe(
+      OpenAIChat.protocol.body.from(request).pipe(
        Effect.map((body) => {
          const sourceAssistants = request.messages.filter((message) => message.role === "assistant")
          let assistantIndex = 0
@@ -129,39 +71,16 @@ export const protocol = Protocol.make({
  stream: OpenAIChat.protocol.stream,
 })

-const cacheControl = () => {
-  const breakpoints = newBreakpoints(4)
-  return (cache: CacheHint | undefined) => {
-    if (cache === undefined || breakpoints.remaining === 0) return undefined
-    breakpoints.remaining -= 1
-    return {
-      type: "ephemeral" as const,
-      ...(ttlBucket(cache.ttlSeconds) === "1h" ? { ttl: "1h" } : {}),
-    }
-  }
-}
-
 const bodyOptions = (input: unknown) => {
  const openrouter = isRecord(input) ? input : {}
-  const { usage, models, provider, plugins, web_search_options, debug, user, reasoning, promptCacheKey, ...options } =
-    openrouter
  return {
-    ...options,
-    ...(usage === undefined || usage === true
+    ...(openrouter.usage === true
      ? { usage: { include: true } }
-      : usage === false
-        ? { usage: { include: false } }
-        : isRecord(usage)
-          ? { usage }
-          : {}),
-    ...(Array.isArray(models) ? { models } : {}),
-    ...(isRecord(provider) ? { provider } : {}),
-    ...(Array.isArray(plugins) ? { plugins } : {}),
-    ...(isRecord(web_search_options) ? { web_search_options } : {}),
-    ...(isRecord(debug) ? { debug } : {}),
-    ...(typeof user === "string" ? { user } : {}),
-    ...(isRecord(reasoning) ? { reasoning } : {}),
-    ...(typeof promptCacheKey === "string" ? { prompt_cache_key: promptCacheKey } : {}),
+      : isRecord(openrouter.usage)
+        ? { usage: openrouter.usage }
+        : {}),
+    ...(isRecord(openrouter.reasoning) ? { reasoning: openrouter.reasoning } : {}),
+    ...(typeof openrouter.promptCacheKey === "string" ? { prompt_cache_key: openrouter.promptCacheKey } : {}),
  }
 }

@@ -175,7 +94,7 @@ export const route = Route.make({

 export const routes = [route]

-const configuredRoute = (input: LanguageModelOptions) => {
+const configuredRoute = (input: ModelOptions) => {
  const { apiKey: _, auth: _auth, baseURL, ...rest } = input
  return route.with({
    ...rest,
@@ -184,25 +103,14 @@ const configuredRoute = (input: LanguageModelOptions) => {
  })
 }

-export const configure = (input: LanguageModelOptions = {}) => {
+export const configure = (input: ModelOptions = {}) => {
  const route = configuredRoute(input)
  return {
    id,
-    model: (modelID: string | ModelID) => route.model<OpenRouterProviderOptionsInput>({ id: modelID }),
+    model: (modelID: string | ModelID) => route.model({ id: modelID }),
    configure,
  }
 }

 export const provider = configure()
-export const model: ProviderPackage.Definition<Settings, OpenRouterProviderOptionsInput>["model"] = (
-  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 model = provider.model
@@ -1,81 +1,49 @@
 import { AuthOptions, type ProviderAuthOption } from "../route/auth-options"
-import { Route, type RouteDefaultsInput } from "../route/client"
-import { Endpoint } from "../route/endpoint"
-import { HttpOptions, ProviderID, type ModelID, type ProviderOptions } from "../schema"
+import type { RouteDefaultsInput } from "../route/client"
+import { HttpOptions, ProviderID, type ModelID } from "../schema"
 import * as OpenAICompatibleProfiles from "./openai-compatible-profile"
 import * as OpenAICompatibleChat from "../protocols/openai-compatible-chat"
-import * as OpenAIChat from "../protocols/openai-chat"
 import * as OpenAIResponses from "../protocols/openai-responses"
 import { XAIImages } from "../protocols/xai-images"
-import type { OpenAIOptionsInput } from "./openai-options"
-import type { ProviderPackage } from "../provider-package"

 export const id = ProviderID.make("xai")

-export type XAIProviderOptionsInput = ProviderOptions & {
-  readonly xai?: OpenAIOptionsInput
-}
-
-export type LanguageModelOptions = Omit<RouteDefaultsInput, "providerOptions"> &
+export type ModelOptions = RouteDefaultsInput &
  ProviderAuthOption<"optional"> & {
    readonly baseURL?: string
-    readonly providerOptions?: XAIProviderOptionsInput
  }

-export interface Settings extends ProviderPackage.Settings {
-  readonly apiKey?: string
-  readonly baseURL?: string
-  readonly providerOptions?: XAIProviderOptionsInput
-}
-
 export type { XAIImageOptions } from "../protocols/xai-images"

-const responsesRoute = Route.make({
-  id: "openai-responses",
-  provider: id,
-  providerMetadataKey: "xai",
-  protocol: OpenAIResponses.protocol,
-  endpoint: Endpoint.path("/responses", { baseURL: OpenAICompatibleProfiles.profiles.xai.baseURL }),
-  transport: OpenAIResponses.httpTransport,
-  defaults: { providerOptions: { xai: { store: false } } },
-})
-
-const chatRoute = Route.make({
-  id: "openai-compatible-chat",
-  provider: id,
-  providerMetadataKey: "xai",
-  protocol: OpenAIChat.protocol,
-  endpoint: Endpoint.path("/chat/completions", { baseURL: OpenAICompatibleProfiles.profiles.xai.baseURL }),
-  transport: OpenAICompatibleChat.route.transport,
-})
-
-export const routes = [responsesRoute, chatRoute]
+export const routes = [OpenAIResponses.route, OpenAICompatibleChat.route]

 const auth = (options: ProviderAuthOption<"optional">) => AuthOptions.bearer(options, "XAI_API_KEY")

-const configuredResponsesRoute = (input: LanguageModelOptions) => {
+const configuredResponsesRoute = (input: ModelOptions) => {
  const { apiKey: _, auth: _auth, baseURL, ...rest } = input
-  return responsesRoute.with({
+  return OpenAIResponses.route.with({
    ...rest,
+    provider: id,
    endpoint: { baseURL: baseURL ?? OpenAICompatibleProfiles.profiles.xai.baseURL },
    auth: auth(input),
  })
 }

-const configuredChatRoute = (input: LanguageModelOptions) => {
+const configuredChatRoute = (input: ModelOptions) => {
  const { apiKey: _, auth: _auth, baseURL, ...rest } = input
-  return chatRoute.with({
+  return OpenAICompatibleChat.route.with({
    ...rest,
+    provider: id,
    endpoint: { baseURL: baseURL ?? OpenAICompatibleProfiles.profiles.xai.baseURL },
    auth: auth(input),
  })
 }

-export const configure = (input: LanguageModelOptions = {}) => {
+export const configure = (input: ModelOptions = {}) => {
  const responsesRoute = configuredResponsesRoute(input)
  const chatRoute = configuredChatRoute(input)
-  const responses = (modelID: string | ModelID) => responsesRoute.model<XAIProviderOptionsInput>({ id: modelID })
-  const chat = (modelID: string | ModelID) => chatRoute.model<XAIProviderOptionsInput>({ id: modelID })
+  const responses = (modelID: string | ModelID) => responsesRoute.model({ id: modelID })
+  const chat = (modelID: string | ModelID) => chatRoute.model({ id: modelID })
  const image = (modelID: string | ModelID) =>
    XAIImages.model({
      id: modelID,
@@ -95,15 +63,7 @@ export const configure = (input: LanguageModelOptions = {}) => {
 }

 export const provider = configure()
-export const model: ProviderPackage.Definition<Settings, XAIProviderOptionsInput>["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 model = provider.model
 export const responses = provider.responses
 export const chat = provider.chat
 export const image = provider.image
@@ -22,17 +22,13 @@ export type ProviderAuthOption<Mode extends ApiKeyMode> =
  | AuthOverride
  | (Mode extends "optional" ? OptionalApiKeyAuth : RequiredApiKeyAuth)

-export type LanguageModelOptions<Base, Mode extends ApiKeyMode> = Omit<Base, "apiKey" | "auth"> &
-  ProviderAuthOption<Mode>
+export type ModelOptions<Base, Mode extends ApiKeyMode> = Omit<Base, "apiKey" | "auth"> & ProviderAuthOption<Mode>

-export type LanguageModelArgs<Base, Mode extends ApiKeyMode> = Mode extends "optional"
-  ? readonly [options?: LanguageModelOptions<Base, Mode>]
-  : readonly [options: LanguageModelOptions<Base, Mode>]
+export type ModelArgs<Base, Mode extends ApiKeyMode> = Mode extends "optional"
+  ? readonly [options?: ModelOptions<Base, Mode>]
+  : readonly [options: ModelOptions<Base, Mode>]

-export type LanguageModelFactory<Base, Mode extends ApiKeyMode, LanguageModel> = (
-  id: string,
-  ...args: LanguageModelArgs<Base, Mode>
-) => LanguageModel
+export type ModelFactory<Base, Mode extends ApiKeyMode, Model> = (id: string, ...args: ModelArgs<Base, Mode>) => Model

 /**
 * Require at least one of the keys in `T`. Use for option shapes where any
@@ -1,6 +1,6 @@
 import { Config, Effect, Redacted } from "effect"
 import { Headers } from "effect/unstable/http"
-import { AuthenticationReason, InvalidRequestReason, AIError, type HttpOptions } from "../schema"
+import { AuthenticationReason, InvalidRequestReason, LLMError, type HttpOptions } from "../schema"

 export class MissingCredentialError extends Error {
  readonly _tag = "MissingCredentialError"
@@ -11,7 +11,7 @@ export class MissingCredentialError extends Error {
 }

 export type CredentialError = MissingCredentialError | Config.ConfigError
-export type AuthError = CredentialError | AIError
+export type AuthError = CredentialError | LLMError
 type Secret = string | Redacted.Redacted | Config.Config<string | Redacted.Redacted>

 export interface AuthInput {
@@ -100,7 +100,7 @@ export const headers = (input: Headers.Input) =>

 export const remove = (name: string) => auth((input) => Effect.succeed(Headers.remove(input.headers, name)))

-export const custom = (apply: (input: AuthInput) => Effect.Effect<Headers.Headers, AIError>) => auth(apply)
+export const custom = (apply: (input: AuthInput) => Effect.Effect<Headers.Headers, LLMError>) => auth(apply)

 export const passthrough = none

@@ -134,9 +134,9 @@ export function bearerHeader(name: string, source?: Secret | Credential) {
  return render(source)
 }

-const toAIError = (error: AuthError): AIError => {
+const toLLMError = (error: AuthError): LLMError => {
  if (error instanceof MissingCredentialError || error instanceof Config.ConfigError) {
-    return new AIError({
+    return new LLMError({
      module: "Auth",
      method: "apply",
      reason:
@@ -150,7 +150,7 @@ const toAIError = (error: AuthError): AIError => {

 export const toEffect =
  (input: Definition) =>
-  (authInput: AuthInput): Effect.Effect<Headers.Headers, AIError> =>
-    input.apply(authInput).pipe(Effect.mapError(toAIError))
+  (authInput: AuthInput): Effect.Effect<Headers.Headers, LLMError> =>
+    input.apply(authInput).pipe(Effect.mapError(toLLMError))

 export * as Auth from "./auth"
@@ -5,21 +5,22 @@ import { Endpoint, type EndpointPatch } from "./endpoint"
 import { RequestExecutor } from "./executor"
 import { Framing } from "./framing"
 import { HttpTransport } from "./transport"
-import type { HttpRequestTransform, Transport, TransportRuntime } from "./transport"
+import type { Transport, TransportRuntime } from "./transport"
 import { WebSocketExecutor } from "./transport"
 import type { Protocol } from "./protocol"
 import { applyCachePolicy } from "../cache-policy"
 import * as ProviderShared from "../protocols/shared"
-import type { ProtocolID, ProviderOptions } from "../schema"
+import type { LLMError, PreparedRequestOf, ProtocolID, ProviderOptions } from "../schema"
 import {
-  AIError,
  GenerationOptions,
  HttpOptions,
  LLMRequest,
  LLMResponse,
-  LanguageModel,
-  LanguageModelLimits,
+  Model,
+  ModelLimits,
+  LLMError as LLMErrorClass,
  LLMEvent,
+  PreparedRequest,
  ProviderID,
  mergeGenerationOptions,
  mergeHttpOptions,
@@ -30,7 +31,7 @@ export interface RouteBody<Body> {
  /** Schema for the validated provider-native body sent as the JSON request. */
  readonly schema: Schema.Codec<Body, unknown>
  /** Build the provider-native body from a common `LLMRequest`. */
-  readonly from: (request: LLMRequest) => Effect.Effect<Body, AIError>
+  readonly from: (request: LLMRequest) => Effect.Effect<Body, LLMError>
 }

 export interface Route<Body, Prepared = unknown> {
@@ -45,19 +46,13 @@ export interface Route<Body, Prepared = unknown> {
  readonly defaults: RouteDefaults
  readonly body: RouteBody<Body>
  readonly with: (patch: RoutePatch<Body, Prepared>) => Route<Body, Prepared>
-  readonly model: <Options extends ProviderOptions = ProviderOptions>(
-    input: RouteMappedLanguageModelInput,
-  ) => LanguageModel<Options>
-  readonly prepareTransport: (
-    body: Body,
-    request: LLMRequest,
-    options?: StreamOptions,
-  ) => Effect.Effect<Prepared, AIError>
+  readonly model: (input: RouteMappedModelInput) => Model
+  readonly prepareTransport: (body: Body, request: LLMRequest) => Effect.Effect<Prepared, LLMError>
  readonly streamPrepared: (
    prepared: Prepared,
    request: LLMRequest,
    runtime: TransportRuntime,
-  ) => Stream.Stream<LLMEvent, AIError>
+  ) => Stream.Stream<LLMEvent, LLMError>
 }

 // Route registries intentionally erase body generics after construction.
@@ -68,13 +63,13 @@ export type AnyRoute = Route<any, any>

 export type HttpOptionsInput = HttpOptions.Input

-export type RouteLanguageModelInput = Omit<LanguageModel.Input, "provider" | "route">
+export type RouteModelInput = Omit<Model.Input, "provider" | "route">

-export type RouteRoutedLanguageModelInput = Omit<LanguageModel.Input, "route">
+export type RouteRoutedModelInput = Omit<Model.Input, "route">

 export interface RouteDefaults {
  readonly headers?: Record<string, string>
-  readonly limits?: LanguageModelLimits
+  readonly limits?: ModelLimits
  readonly generation?: GenerationOptions
  readonly providerOptions?: ProviderOptions
  readonly http?: HttpOptions
@@ -82,7 +77,7 @@ export interface RouteDefaults {

 export interface RouteDefaultsInput {
  readonly headers?: Record<string, string>
-  readonly limits?: LanguageModelLimits.Input
+  readonly limits?: ModelLimits.Input
  readonly generation?: GenerationOptions.Input
  readonly providerOptions?: ProviderOptions
  readonly http?: HttpOptions.Input
@@ -96,17 +91,14 @@ export interface RoutePatch<Body, Prepared> extends RouteDefaultsInput {
  readonly endpoint?: EndpointPatch<Body>
 }

-type RouteMappedLanguageModelInput = RouteLanguageModelInput | RouteRoutedLanguageModelInput
+type RouteMappedModelInput = RouteModelInput | RouteRoutedModelInput

-const makeRouteLanguageModel = <Options extends ProviderOptions = ProviderOptions>(
-  route: AnyRoute,
-  mapped: RouteMappedLanguageModelInput,
-) => {
+const makeRouteModel = (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 LanguageModel.make<Options>({
+  return Model.make({
    ...mapped,
    provider,
    route,
@@ -119,7 +111,7 @@ const mergeRouteDefaults = (base: RouteDefaults | undefined, patch: RouteDefault
    ...base,
    ...patch,
    headers,
-    limits: patch.limits === undefined ? base?.limits : LanguageModelLimits.make(patch.limits),
+    limits: patch.limits === undefined ? base?.limits : ModelLimits.make(patch.limits),
    generation: mergeGenerationOptions(generationOptions(base?.generation), generationOptions(patch.generation)),
    providerOptions: mergeProviderOptions(base?.providerOptions, patch.providerOptions),
    http: mergeHttpOptions(
@@ -150,20 +142,27 @@ export const httpOptions = (input: HttpOptionsInput | undefined) => {
 }

 export interface Interface {
+  /**
+   * Compile a request through protocol body construction, validation, and HTTP
+   * preparation without sending it. Returns the prepared request including the
+   * provider-native body.
+   *
+   * Pass a `Body` type argument to statically expose the route's body
+   * shape (e.g. `prepare<OpenAIChatBody>(...)`) — the runtime body is
+   * identical, so this is a type-level assertion the caller makes about which
+   * route the request will resolve to.
+   */
+  readonly prepare: <Body = unknown>(request: LLMRequest) => Effect.Effect<PreparedRequestOf<Body>, LLMError>
  readonly stream: StreamMethod
  readonly generate: GenerateMethod
 }

-export interface StreamOptions {
-  readonly transform?: HttpRequestTransform
-}
-
 export interface StreamMethod {
-  (request: LLMRequest, options?: StreamOptions): Stream.Stream<LLMEvent, AIError>
+  (request: LLMRequest): Stream.Stream<LLMEvent, LLMError>
 }

 export interface GenerateMethod {
-  (request: LLMRequest, options?: StreamOptions): Effect.Effect<LLMResponse, AIError>
+  (request: LLMRequest): Effect.Effect<LLMResponse, LLMError>
 }

 export class Service extends Context.Service<Service, Interface>()("@opencode/LLMClient") {}
@@ -227,11 +226,11 @@ export interface MakeTransportInput<Body, Prepared, Frame, Event, State> {

 const streamError = (route: string, message: string, cause: Cause.Cause<unknown>) => {
  const failed = cause.reasons.find(Cause.isFailReason)?.error
-  if (failed instanceof AIError) return failed
+  if (failed instanceof LLMErrorClass) return failed
  return ProviderShared.eventError(route, message, Cause.pretty(cause))
 }

-const requireTerminalEvent = (route: string) => (events: Stream.Stream<LLMEvent, AIError>) =>
+const requireTerminalEvent = (route: string) => (events: Stream.Stream<LLMEvent, LLMError>) =>
  Stream.suspend(() => {
    let terminal = false
    return events.pipe(
@@ -297,9 +296,8 @@ function makeFromTransport<Body, Prepared, Frame, Event, State>(
          defaults: mergeRouteDefaults(route.defaults, defaults),
        })
      },
-      model: <Options extends ProviderOptions = ProviderOptions>(input: RouteMappedLanguageModelInput) =>
-        makeRouteLanguageModel<Options>(route, input),
-      prepareTransport: (body, request, options) =>
+      model: (input) => makeRouteModel(route, input),
+      prepareTransport: (body, request) =>
        routeInput.transport.prepare({
          body,
          request,
@@ -307,7 +305,6 @@ function makeFromTransport<Body, Prepared, Frame, Event, State>(
          auth: routeInput.auth ?? Auth.none,
          encodeBody,
          headers: routeInput.headers,
-          transform: options?.transform,
        }),
      streamPrepared: (prepared: Prepared, request: LLMRequest, runtime: TransportRuntime) => {
        const route = `${request.model.provider}/${request.model.route.id}`
@@ -373,14 +370,17 @@ export function make<Body, Prepared, Frame, Event, State>(
  })
 }

-const compile = Effect.fn("LLM.compile")(function* (request: LLMRequest, options?: StreamOptions) {
+// `compile` is the important boundary: it turns a common `LLMRequest` into a
+// validated provider body plus transport-private prepared data, but does not
+// execute transport.
+const compile = Effect.fn("LLM.compile")(function* (request: LLMRequest) {
  const resolved = applyCachePolicy(resolveRequestOptions(request))
  const route = resolved.model.route

  const body = yield* route.body
    .from(resolved)
    .pipe(Effect.flatMap(ProviderShared.validateWith(Schema.decodeUnknownEffect(route.body.schema))))
-  const prepared = yield* route.prepareTransport(body, resolved, options)
+  const prepared = yield* route.prepareTransport(body, resolved)

  return {
    request: resolved,
@@ -390,30 +390,30 @@ const compile = Effect.fn("LLM.compile")(function* (request: LLMRequest, options
  }
 })

-/** @internal Test-only projection of the execution compiler; not exported from package barrels. */
-export const compileRequest = Effect.fn("LLM.compileRequest")(function* (request: LLMRequest) {
+const prepareWith = Effect.fn("LLMClient.prepare")(function* (request: LLMRequest) {
  const compiled = yield* compile(request)
-  return {
+
+  return new PreparedRequest({
    id: compiled.request.id ?? "request",
    route: compiled.route.id,
    protocol: compiled.route.protocol,
    model: compiled.request.model,
    body: compiled.body,
    metadata: { transport: compiled.route.transport.id },
-  }
+  })
 })

-const streamRequestWith = (runtime: TransportRuntime) => (request: LLMRequest, options?: StreamOptions) =>
+const streamRequestWith = (runtime: TransportRuntime) => (request: LLMRequest) =>
  Stream.unwrap(
    Effect.gen(function* () {
-      const compiled = yield* compile(request, options)
+      const compiled = yield* compile(request)
      return compiled.route.streamPrepared(compiled.prepared, compiled.request, runtime)
    }),
  )

 const generateWith = (stream: Interface["stream"]) =>
-  Effect.fn("LLM.generate")(function* (request: LLMRequest, options?: StreamOptions) {
-    const state = yield* stream(request, options).pipe(Stream.runFold(LLMResponse.empty, LLMResponse.reduce))
+  Effect.fn("LLM.generate")(function* (request: LLMRequest) {
+    const state = yield* stream(request).pipe(Stream.runFold(LLMResponse.empty, LLMResponse.reduce))
    const response = LLMResponse.complete(state)
    if (response) return response
    return yield* ProviderShared.eventError(
@@ -422,24 +422,27 @@ const generateWith = (stream: Interface["stream"]) =>
    )
  })

-export function stream(request: LLMRequest, options?: StreamOptions): Stream.Stream<LLMEvent, AIError> {
+export const prepare = <Body = unknown>(request: LLMRequest) =>
+  prepareWith(request) as Effect.Effect<PreparedRequestOf<Body>, LLMError>
+
+export function stream(request: LLMRequest): Stream.Stream<LLMEvent, LLMError> {
  return Stream.unwrap(
    Effect.gen(function* () {
-      return (yield* Service).stream(request, options)
+      return (yield* Service).stream(request)
    }),
-  ) as Stream.Stream<LLMEvent, AIError>
+  ) as Stream.Stream<LLMEvent, LLMError>
 }

-export function generate(request: LLMRequest, options?: StreamOptions): Effect.Effect<LLMResponse, AIError> {
+export function generate(request: LLMRequest): Effect.Effect<LLMResponse, LLMError> {
  return Effect.gen(function* () {
-    return yield* (yield* Service).generate(request, options)
-  }) as Effect.Effect<LLMResponse, AIError>
+    return yield* (yield* Service).generate(request)
+  }) as Effect.Effect<LLMResponse, LLMError>
 }

-export const streamRequest = (request: LLMRequest, options?: StreamOptions) =>
+export const streamRequest = (request: LLMRequest) =>
  Stream.unwrap(
    Effect.gen(function* () {
-      return (yield* Service).stream(request, options)
+      return (yield* Service).stream(request)
    }),
  )

@@ -450,7 +453,7 @@ export const layer: Layer.Layer<Service, never, RequestExecutor.Service> = Layer
      http: yield* RequestExecutor.Service,
      webSocket: Option.getOrUndefined(yield* Effect.serviceOption(WebSocketExecutor.Service)),
    })
-    return Service.of({ stream, generate: generateWith(stream) })
+    return Service.of({ prepare: prepareWith as Interface["prepare"], stream, generate: generateWith(stream) })
  }),
 )

@@ -459,6 +462,7 @@ export const Route = { make } as const
 export const LLMClient = {
  Service,
  layer,
+  prepare,
  stream,
  generate,
 } as const
@@ -12,7 +12,7 @@ import {
  HttpRateLimitDetails,
  HttpRequestDetails,
  HttpResponseDetails,
-  AIError,
+  LLMError,
  TransportReason,
 } from "../schema"
 import { classifyProviderFailure } from "../provider-error"
@@ -20,10 +20,10 @@ import { classifyProviderFailure } from "../provider-error"
 export interface Interface {
  readonly execute: (
    request: HttpClientRequest.HttpClientRequest,
-  ) => Effect.Effect<HttpClientResponse.HttpClientResponse, AIError>
+  ) => Effect.Effect<HttpClientResponse.HttpClientResponse, LLMError>
 }

-export class Service extends Context.Service<Service, Interface>()("@opencode/AI/RequestExecutor") {}
+export class Service extends Context.Service<Service, Interface>()("@opencode/LLM/RequestExecutor") {}

 const BODY_LIMIT = 16_384
 const REDACTED = "<redacted>"
@@ -220,7 +220,7 @@ const statusError =
      const retryAfter = retryAfterMs(headers)
      const rateLimit = rateLimitDetails(headers, retryAfter)
      const details = responseBody(body, request)
-      return yield* new AIError({
+      return yield* new LLMError({
        module: "RequestExecutor",
        method: "execute",
        reason: classifyProviderFailure({
@@ -246,7 +246,7 @@ const toHttpError = (redactedNames: ReadonlyArray<string | RegExp>) => (error: u
    readonly kind?: string | undefined
    readonly request?: HttpClientRequest.HttpClientRequest | undefined
  }) =>
-    new AIError({
+    new LLMError({
      module: "RequestExecutor",
      method: "execute",
      reason: new TransportReason({
@@ -1,6 +1,6 @@
 import type { Stream } from "effect"
 import * as ProviderShared from "../protocols/shared"
-import type { AIError } from "../schema"
+import type { LLMError } from "../schema"

 /**
 * Decode a streaming HTTP response body into provider-protocol frames.
@@ -18,7 +18,7 @@ import type { AIError } from "../schema"
 */
 export interface Definition<Frame> {
  readonly id: string
-  readonly frame: (bytes: Stream.Stream<Uint8Array, AIError>) => Stream.Stream<Frame, AIError>
+  readonly frame: (bytes: Stream.Stream<Uint8Array, LLMError>) => Stream.Stream<Frame, LLMError>
 }

 /** Server-Sent Events framing. Used by every JSON-streaming HTTP provider. */
@@ -1,14 +1,13 @@
 export { Route, LLMClient } from "./client"
 export type {
  Route as RouteShape,
-  RouteLanguageModelInput,
-  RouteRoutedLanguageModelInput,
+  RouteModelInput,
+  RouteRoutedModelInput,
  RouteDefaults,
  RouteDefaultsInput,
  AnyRoute,
  Interface as LLMClientShape,
  Service as LLMClientService,
-  StreamOptions,
 } from "./client"
 export * from "./executor"
 export { Auth } from "./auth"
@@ -23,4 +22,4 @@ export type { ApiKeyMode, AuthOverride, ProviderAuthOption } from "./auth-option
 export type { Definition as EndpointFn, EndpointInput } from "./endpoint"
 export type { Definition as FramingDef } from "./framing"
 export type { Protocol as ProtocolDef } from "./protocol"
-export type { HttpRequest, HttpRequestTransform, Transport as TransportDef, TransportRuntime } from "./transport"
+export type { Transport as TransportDef, TransportRuntime } from "./transport"
@@ -1,5 +1,5 @@
 import { Schema, type Effect } from "effect"
-import type { AIError, LLMEvent, LLMRequest, ProtocolID } from "../schema"
+import type { LLMError, LLMEvent, LLMRequest, ProtocolID } from "../schema"

 /**
 * The semantic API contract of one model server family.
@@ -47,7 +47,7 @@ export interface ProtocolBody<Body> {
  /** Schema for the validated provider-native body sent as the JSON request. */
  readonly schema: Schema.Codec<Body, unknown>
  /** Build the provider-native body from a common `LLMRequest`. */
-  readonly from: (request: LLMRequest) => Effect.Effect<Body, AIError>
+  readonly from: (request: LLMRequest) => Effect.Effect<Body, LLMError>
 }

 export interface ProtocolStream<Frame, Event, State> {
@@ -56,7 +56,7 @@ export interface ProtocolStream<Frame, Event, State> {
  /** Initial parser state. Called once per response with the resolved request. */
  readonly initial: (request: LLMRequest) => State
  /** Translate one event into emitted `LLMEvent`s plus the next state. */
-  readonly step: (state: State, event: Event) => Effect.Effect<readonly [State, ReadonlyArray<LLMEvent>], AIError>
+  readonly step: (state: State, event: Event) => Effect.Effect<readonly [State, ReadonlyArray<LLMEvent>], LLMError>
  /** Optional request-completion signal for transports that do not end naturally. */
  readonly terminal?: (event: Event) => boolean
  /** Optional flush emitted when the framed stream ends. */
@@ -28,9 +28,57 @@ const applyQuery = (url: string, query: Record<string, string> | undefined) => {
  return next.toString()
 }

+const PROTOCOL_BODY_OVERLAY_DENYLIST = new Set([
+  "anthropic_version",
+  "content",
+  "contents",
+  "frequencyPenalty",
+  "frequency_penalty",
+  "generationConfig",
+  "inferenceConfig",
+  "input",
+  "maxTokens",
+  "max_tokens",
+  "messages",
+  "model",
+  "presencePenalty",
+  "presence_penalty",
+  "responseFormat",
+  "response_format",
+  "seed",
+  "stop",
+  "stopSequences",
+  "stop_sequences",
+  "stream",
+  "streamOptions",
+  "stream_options",
+  "system",
+  "systemInstruction",
+  "system_instruction",
+  "temperature",
+  "thinking",
+  "toolChoice",
+  "toolConfig",
+  "tool_choice",
+  "tool_config",
+  "tools",
+  "topK",
+  "topP",
+  "top_k",
+  "top_p",
+])
+
+const forbiddenBodyOverlayKeys = (body: Record<string, unknown>) =>
+  Object.keys(body).filter((key) => PROTOCOL_BODY_OVERLAY_DENYLIST.has(key))
+
 const bodyWithOverlay = <Body>(body: Body, request: LLMRequest, encodeBody: (body: Body) => string) =>
  Effect.gen(function* () {
    if (request.http?.body === undefined) return { jsonBody: body, bodyText: encodeBody(body) }
+    const forbiddenKeys = forbiddenBodyOverlayKeys(request.http.body)
+    if (forbiddenKeys.length > 0)
+      return yield* ProviderShared.invalidRequest(
+        `http.body cannot overlay protocol-owned field(s): ${forbiddenKeys.join(", ")}`,
+      )
    if (ProviderShared.isRecord(body)) {
      const overlaid = mergeJsonRecords(body, request.http.body) ?? {}
      return { jsonBody: overlaid, bodyText: ProviderShared.encodeJson(overlaid) }
@@ -72,19 +120,14 @@ export const httpJson = <Body, Frame>(input: HttpJsonInput<Body, Frame>): HttpJs
  id: "http-json",
  with: (patch) => httpJson({ ...input, ...patch }),
  prepare: (prepareInput) =>
-    Effect.gen(function* () {
-      const parts = yield* jsonRequestParts({ ...prepareInput })
-      const request = { url: parts.url, method: "POST", headers: { ...parts.headers }, body: parts.bodyText }
-      yield* (prepareInput.transform?.(request) ?? Effect.void)
-      return {
-        request: ProviderShared.jsonPost({
-          url: request.url,
-          body: request.body ?? "",
-          headers: Headers.fromInput(request.headers),
-        }),
+    jsonRequestParts({
+      ...prepareInput,
+    }).pipe(
+      Effect.map((parts) => ({
+        request: ProviderShared.jsonPost({ url: parts.url, body: parts.bodyText, headers: parts.headers }),
        framing: input.framing,
-      }
-    }),
+      })),
+    ),
  frames: (prepared, request, runtime) =>
    Stream.unwrap(
      runtime.http
@@ -3,26 +3,21 @@ import { Endpoint } from "../endpoint"
 import { Auth } from "../auth"
 import type { Interface as RequestExecutorInterface } from "../executor"
 import type { Interface as WebSocketExecutorInterface } from "./websocket"
-import type { AIError, LLMRequest } from "../../schema"
+import type { LLMError, LLMRequest } from "../../schema"

 export interface TransportRuntime {
  readonly http: RequestExecutorInterface
  readonly webSocket?: WebSocketExecutorInterface
 }

-export interface HttpRequest {
-  url: string
-  readonly method: string
-  headers: Record<string, string>
-  body: string | undefined
-}
-
-export type HttpRequestTransform = (request: HttpRequest) => Effect.Effect<void>
-
 export interface Transport<Body, Prepared, Frame> {
  readonly id: string
-  readonly prepare: (input: TransportPrepareInput<Body>) => Effect.Effect<Prepared, AIError>
-  readonly frames: (prepared: Prepared, request: LLMRequest, runtime: TransportRuntime) => Stream.Stream<Frame, AIError>
+  readonly prepare: (input: TransportPrepareInput<Body>) => Effect.Effect<Prepared, LLMError>
+  readonly frames: (
+    prepared: Prepared,
+    request: LLMRequest,
+    runtime: TransportRuntime,
+  ) => Stream.Stream<Frame, LLMError>
 }

 export interface TransportPrepareInput<Body> {
@@ -32,7 +27,6 @@ export interface TransportPrepareInput<Body> {
  readonly auth: Auth.Definition
  readonly encodeBody: (body: Body) => string
  readonly headers?: (input: { readonly request: LLMRequest }) => Record<string, string>
-  readonly transform?: HttpRequestTransform
 }

 export * as HttpTransport from "./http"
@@ -1,6 +1,6 @@
 import { Cause, Context, Effect, Layer, Queue, Stream } from "effect"
 import { Headers } from "effect/unstable/http"
-import { AIError, TransportReason } from "../../schema"
+import { LLMError, TransportReason } from "../../schema"
 import * as HttpTransport from "./http"
 import type { Transport } from "./index"

@@ -10,13 +10,13 @@ export interface WebSocketRequest {
 }

 export interface WebSocketConnection {
-  readonly sendText: (message: string) => Effect.Effect<void, AIError>
-  readonly messages: Stream.Stream<string | Uint8Array, AIError>
+  readonly sendText: (message: string) => Effect.Effect<void, LLMError>
+  readonly messages: Stream.Stream<string | Uint8Array, LLMError>
  readonly close: Effect.Effect<void, never>
 }

 export interface Interface {
-  readonly open: (input: WebSocketRequest) => Effect.Effect<WebSocketConnection, AIError>
+  readonly open: (input: WebSocketRequest) => Effect.Effect<WebSocketConnection, LLMError>
 }

 type WebSocketConstructorWithHeaders = new (
@@ -24,14 +24,14 @@ type WebSocketConstructorWithHeaders = new (
  options?: { readonly headers?: Headers.Headers },
 ) => globalThis.WebSocket

-export class Service extends Context.Service<Service, Interface>()("@opencode/AI/WebSocketExecutor") {}
+export class Service extends Context.Service<Service, Interface>()("@opencode/LLM/WebSocketExecutor") {}

 const transportError = (
  method: string,
  message: string,
  input: { readonly url?: string; readonly kind?: string } = {},
 ) =>
-  new AIError({
+  new LLMError({
    module: "WebSocketExecutor",
    method,
    reason: new TransportReason({ message, url: input.url, kind: input.kind }),
@@ -59,7 +59,7 @@ const waitOpen = (ws: globalThis.WebSocket, input: WebSocketRequest) => {
      }),
    )
  }
-  return Effect.callback<void, AIError>((resume, signal) => {
+  return Effect.callback<void, LLMError>((resume, signal) => {
    const cleanup = () => {
      ws.removeEventListener("open", onOpen)
      ws.removeEventListener("error", onError)
@@ -138,10 +138,10 @@ export const layer: Layer.Layer<Service> = Layer.succeed(Service, Service.of({ o
 export const fromWebSocket = (
  ws: globalThis.WebSocket,
  input: WebSocketRequest,
-): Effect.Effect<WebSocketConnection, AIError> =>
+): Effect.Effect<WebSocketConnection, LLMError> =>
  Effect.gen(function* () {
    yield* waitOpen(ws, input)
-    const messages = yield* Queue.bounded<string | Uint8Array, AIError | Cause.Done<void>>(128)
+    const messages = yield* Queue.bounded<string | Uint8Array, LLMError | Cause.Done<void>>(128)

    const onMessage = (event: MessageEvent) => {
      if (typeof event.data === "string") return Queue.offerUnsafe(messages, event.data)
@@ -213,7 +213,7 @@ export interface JsonPrepared {
 }

 export interface JsonInput<Body, Message> {
-  readonly toMessage: (body: Body | Record<string, unknown>) => Effect.Effect<Message, AIError>
+  readonly toMessage: (body: Body | Record<string, unknown>) => Effect.Effect<Message, LLMError>
  readonly encodeMessage: (message: Message) => string
 }

@@ -2,28 +2,28 @@ import { Schema } from "effect"
 import { Tool } from "@opencode-ai/schema/tool"
 import { ModelID, ProviderID, ProviderMetadata, RouteID } from "./ids"

-export const ProviderFailureClassification = Schema.Literals(["context-overflow", "payload-too-large"])
+export const ProviderFailureClassification = Schema.Literal("context-overflow")
 export type ProviderFailureClassification = typeof ProviderFailureClassification.Type

-export class HttpRequestDetails extends Schema.Class<HttpRequestDetails>("AI.HttpRequestDetails")({
+export class HttpRequestDetails extends Schema.Class<HttpRequestDetails>("LLM.HttpRequestDetails")({
  method: Schema.String,
  url: Schema.String,
  headers: Schema.Record(Schema.String, Schema.String),
 }) {}

-export class HttpResponseDetails extends Schema.Class<HttpResponseDetails>("AI.HttpResponseDetails")({
+export class HttpResponseDetails extends Schema.Class<HttpResponseDetails>("LLM.HttpResponseDetails")({
  status: Schema.Number,
  headers: Schema.Record(Schema.String, Schema.String),
 }) {}

-export class HttpRateLimitDetails extends Schema.Class<HttpRateLimitDetails>("AI.HttpRateLimitDetails")({
+export class HttpRateLimitDetails extends Schema.Class<HttpRateLimitDetails>("LLM.HttpRateLimitDetails")({
  retryAfterMs: Schema.optional(Schema.Number),
  limit: Schema.optional(Schema.Record(Schema.String, Schema.String)),
  remaining: Schema.optional(Schema.Record(Schema.String, Schema.String)),
  reset: Schema.optional(Schema.Record(Schema.String, Schema.String)),
 }) {}

-export class HttpContext extends Schema.Class<HttpContext>("AI.HttpContext")({
+export class HttpContext extends Schema.Class<HttpContext>("LLM.HttpContext")({
  request: HttpRequestDetails,
  response: Schema.optional(HttpResponseDetails),
  body: Schema.optional(Schema.String),
@@ -32,7 +32,7 @@ export class HttpContext extends Schema.Class<HttpContext>("AI.HttpContext")({
  rateLimit: Schema.optional(HttpRateLimitDetails),
 }) {}

-export class InvalidRequestReason extends Schema.Class<InvalidRequestReason>("AI.Error.InvalidRequest")({
+export class InvalidRequestReason extends Schema.Class<InvalidRequestReason>("LLM.Error.InvalidRequest")({
  _tag: Schema.tag("InvalidRequest"),
  message: Schema.String,
  parameter: Schema.optional(Schema.String),
@@ -41,18 +41,18 @@ export class InvalidRequestReason extends Schema.Class<InvalidRequestReason>("AI
  http: Schema.optional(HttpContext),
 }) {}

-export class NoRouteReason extends Schema.Class<NoRouteReason>("AI.Error.NoRoute")({
+export class NoRouteReason extends Schema.Class<NoRouteReason>("LLM.Error.NoRoute")({
  _tag: Schema.tag("NoRoute"),
  route: RouteID,
  provider: ProviderID,
  model: ModelID,
 }) {
  get message() {
-    return `No AI route for ${this.provider}/${this.model} using ${this.route}`
+    return `No LLM route for ${this.provider}/${this.model} using ${this.route}`
  }
 }

-export class AuthenticationReason extends Schema.Class<AuthenticationReason>("AI.Error.Authentication")({
+export class AuthenticationReason extends Schema.Class<AuthenticationReason>("LLM.Error.Authentication")({
  _tag: Schema.tag("Authentication"),
  message: Schema.String,
  kind: Schema.Literals(["missing", "invalid", "expired", "insufficient-permissions", "unknown"]),
@@ -60,7 +60,7 @@ export class AuthenticationReason extends Schema.Class<AuthenticationReason>("AI
  http: Schema.optional(HttpContext),
 }) {}

-export class RateLimitReason extends Schema.Class<RateLimitReason>("AI.Error.RateLimit")({
+export class RateLimitReason extends Schema.Class<RateLimitReason>("LLM.Error.RateLimit")({
  _tag: Schema.tag("RateLimit"),
  message: Schema.String,
  retryAfterMs: Schema.optional(Schema.Number),
@@ -69,21 +69,21 @@ export class RateLimitReason extends Schema.Class<RateLimitReason>("AI.Error.Rat
  http: Schema.optional(HttpContext),
 }) {}

-export class QuotaExceededReason extends Schema.Class<QuotaExceededReason>("AI.Error.QuotaExceeded")({
+export class QuotaExceededReason extends Schema.Class<QuotaExceededReason>("LLM.Error.QuotaExceeded")({
  _tag: Schema.tag("QuotaExceeded"),
  message: Schema.String,
  providerMetadata: Schema.optional(ProviderMetadata),
  http: Schema.optional(HttpContext),
 }) {}

-export class ContentPolicyReason extends Schema.Class<ContentPolicyReason>("AI.Error.ContentPolicy")({
+export class ContentPolicyReason extends Schema.Class<ContentPolicyReason>("LLM.Error.ContentPolicy")({
  _tag: Schema.tag("ContentPolicy"),
  message: Schema.String,
  providerMetadata: Schema.optional(ProviderMetadata),
  http: Schema.optional(HttpContext),
 }) {}

-export class ProviderInternalReason extends Schema.Class<ProviderInternalReason>("AI.Error.ProviderInternal")({
+export class ProviderInternalReason extends Schema.Class<ProviderInternalReason>("LLM.Error.ProviderInternal")({
  _tag: Schema.tag("ProviderInternal"),
  message: Schema.String,
  status: Schema.optional(Schema.Number),
@@ -92,7 +92,7 @@ export class ProviderInternalReason extends Schema.Class<ProviderInternalReason>
  http: Schema.optional(HttpContext),
 }) {}

-export class TransportReason extends Schema.Class<TransportReason>("AI.Error.Transport")({
+export class TransportReason extends Schema.Class<TransportReason>("LLM.Error.Transport")({
  _tag: Schema.tag("Transport"),
  message: Schema.String,
  kind: Schema.optional(Schema.String),
@@ -101,7 +101,7 @@ export class TransportReason extends Schema.Class<TransportReason>("AI.Error.Tra
 }) {}

 export class InvalidProviderOutputReason extends Schema.Class<InvalidProviderOutputReason>(
-  "AI.Error.InvalidProviderOutput",
+  "LLM.Error.InvalidProviderOutput",
 )({
  _tag: Schema.tag("InvalidProviderOutput"),
  message: Schema.String,
@@ -110,7 +110,7 @@ export class InvalidProviderOutputReason extends Schema.Class<InvalidProviderOut
  providerMetadata: Schema.optional(ProviderMetadata),
 }) {}

-export class UnknownProviderReason extends Schema.Class<UnknownProviderReason>("AI.Error.UnknownProvider")({
+export class UnknownProviderReason extends Schema.Class<UnknownProviderReason>("LLM.Error.UnknownProvider")({
  _tag: Schema.tag("UnknownProvider"),
  message: Schema.String,
  status: Schema.optional(Schema.Number),
@@ -118,7 +118,7 @@ export class UnknownProviderReason extends Schema.Class<UnknownProviderReason>("
  http: Schema.optional(HttpContext),
 }) {}

-export const AIErrorReason = Schema.Union([
+export const LLMErrorReason = Schema.Union([
  InvalidRequestReason,
  NoRouteReason,
  AuthenticationReason,
@@ -130,12 +130,12 @@ export const AIErrorReason = Schema.Union([
  InvalidProviderOutputReason,
  UnknownProviderReason,
 ]).pipe(Schema.toTaggedUnion("_tag"))
-export type AIErrorReason = Schema.Schema.Type<typeof AIErrorReason>
+export type LLMErrorReason = Schema.Schema.Type<typeof LLMErrorReason>

-export class AIError extends Schema.TaggedErrorClass<AIError>()("AI.Error", {
+export class LLMError extends Schema.TaggedErrorClass<LLMError>()("LLM.Error", {
  module: Schema.String,
  method: Schema.String,
-  reason: AIErrorReason,
+  reason: LLMErrorReason,
 }) {
  override readonly cause = this.reason

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

@@ -48,7 +49,7 @@ import { ProviderFailureClassification } from "./errors"
 * — for fields we don't normalize and for billing-level audit trails.
 * Matches the same escape-hatch field on `LLMEvent`.
 */
-export class Usage extends Schema.Class<Usage>("AI.Usage")({
+export class Usage extends Schema.Class<Usage>("LLM.Usage")({
  inputTokens: Schema.optional(Schema.Number),
  outputTokens: Schema.optional(Schema.Number),
  nonCachedInputTokens: Schema.optional(Schema.Number),
@@ -313,6 +314,29 @@ export const LLMEvent = Object.assign(llmEventTagged, {
 })
 export type LLMEvent = Schema.Schema.Type<typeof llmEventTagged>

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

 export { ProviderMetadata }

@@ -12,10 +11,10 @@ export type ProtocolID = Schema.Schema.Type<typeof ProtocolID>
 export const RouteID = Schema.String
 export type RouteID = Schema.Schema.Type<typeof RouteID>

-export const ModelID = Schema.String.pipe(Schema.brand("AI.ModelID"))
+export const ModelID = Schema.String.pipe(Schema.brand("LLM.ModelID"))
 export type ModelID = typeof ModelID.Type

-export const ProviderID = Schema.String.pipe(Schema.brand("AI.ProviderID"))
+export const ProviderID = Schema.String.pipe(Schema.brand("LLM.ProviderID"))
 export type ProviderID = typeof ProviderID.Type

 export const ResponseID = Schema.String
@@ -1,7 +1,7 @@
 import { Schema } from "effect"
 import { Tool } from "@opencode-ai/schema/tool"
 import { JsonSchema, MessageRole, ProviderMetadata } from "./ids"
-import { CacheHint, CachePolicy, GenerationOptions, HttpOptions, LanguageModelSchema, ProviderOptions } from "./options"
+import { CacheHint, CachePolicy, GenerationOptions, HttpOptions, ModelSchema, ProviderOptions } from "./options"
 import { isRecord } from "../utils/record"

 const systemPartSchema = Schema.Struct({
@@ -124,7 +124,6 @@ export const ToolCallPart = Object.assign(
    name: Schema.String,
    input: Schema.Unknown,
    providerExecuted: Schema.optional(Schema.Boolean),
-    cache: Schema.optional(CacheHint),
    metadata: Schema.optional(Schema.Record(Schema.String, Schema.Unknown)),
    providerMetadata: Schema.optional(ProviderMetadata),
  }).annotate({ identifier: "LLM.Content.ToolCall" }),
@@ -169,7 +168,6 @@ export const ReasoningPart = Schema.Struct({
  type: Schema.Literal("reasoning"),
  text: Schema.String,
  encrypted: Schema.optional(Schema.String),
-  cache: Schema.optional(CacheHint),
  metadata: Schema.optional(Schema.Record(Schema.String, Schema.Unknown)),
  providerMetadata: Schema.optional(ProviderMetadata),
 }).annotate({ identifier: "LLM.Content.Reasoning" })
@@ -263,7 +261,7 @@ export namespace ToolChoice {

 export class LLMRequest extends Schema.Class<LLMRequest>("LLM.Request")({
  id: Schema.optional(Schema.String),
-  model: LanguageModelSchema,
+  model: ModelSchema,
  system: Schema.Array(SystemPart),
  messages: Schema.Array(Message),
  tools: Schema.Array(ToolDefinition),
@@ -50,7 +50,7 @@ export const mergeProviderOptions = (
  return Object.keys(result).length === 0 ? undefined : result
 }

-export class HttpOptions extends Schema.Class<HttpOptions>("AI.HttpOptions")({
+export class HttpOptions extends Schema.Class<HttpOptions>("LLM.HttpOptions")({
  body: Schema.optional(JsonSchema),
  headers: Schema.optional(Schema.Record(Schema.String, Schema.String)),
  query: Schema.optional(Schema.Record(Schema.String, Schema.String)),
@@ -121,32 +121,31 @@ export const mergeGenerationOptions = (...items: ReadonlyArray<GenerationOptions
  return Object.values(result).some((value) => value !== undefined) ? result : undefined
 }

-export class LanguageModelLimits extends Schema.Class<LanguageModelLimits>("LLM.LanguageModelLimits")({
+export class ModelLimits extends Schema.Class<ModelLimits>("LLM.ModelLimits")({
  context: Schema.optional(Schema.Number),
-  input: Schema.optional(Schema.Number),
  output: Schema.optional(Schema.Number),
 }) {}

-export namespace LanguageModelLimits {
-  export type Input = LanguageModelLimits | ConstructorParameters<typeof LanguageModelLimits>[0]
+export namespace ModelLimits {
+  export type Input = ModelLimits | ConstructorParameters<typeof ModelLimits>[0]

-  /** Normalize model limit input into the canonical `LanguageModelLimits` class. */
+  /** Normalize model limit input into the canonical `ModelLimits` class. */
  export const make = (input: Input | undefined) =>
-    input instanceof LanguageModelLimits ? input : new LanguageModelLimits(input ?? {})
+    input instanceof ModelLimits ? input : new ModelLimits(input ?? {})
 }

-export class LanguageModelDefaults extends Schema.Class<LanguageModelDefaults>("LLM.LanguageModelDefaults")({
-  limits: Schema.optional(LanguageModelLimits),
+export class ModelDefaults extends Schema.Class<ModelDefaults>("LLM.ModelDefaults")({
+  limits: Schema.optional(ModelLimits),
  generation: Schema.optional(GenerationOptions),
  providerOptions: Schema.optional(ProviderOptions),
  http: Schema.optional(HttpOptions),
 }) {}

-export namespace LanguageModelDefaults {
+export namespace ModelDefaults {
  export type Input =
-    | LanguageModelDefaults
+    | ModelDefaults
    | {
-        readonly limits?: LanguageModelLimits.Input
+        readonly limits?: ModelLimits.Input
        readonly generation?: GenerationOptions.Input
        readonly providerOptions?: ProviderOptions
        readonly http?: HttpOptions.Input
@@ -154,9 +153,9 @@ export namespace LanguageModelDefaults {

  /** Normalize selected-model request defaults without applying precedence. */
  export const make = (input: Input) => {
-    if (input instanceof LanguageModelDefaults) return input
-    return new LanguageModelDefaults({
-      limits: input.limits === undefined ? undefined : LanguageModelLimits.make(input.limits),
+    if (input instanceof ModelDefaults) return input
+    return new ModelDefaults({
+      limits: input.limits === undefined ? undefined : ModelLimits.make(input.limits),
      generation: input.generation === undefined ? undefined : GenerationOptions.make(input.generation),
      providerOptions: input.providerOptions,
      http: input.http === undefined ? undefined : HttpOptions.make(input.http),
@@ -164,39 +163,29 @@ export namespace LanguageModelDefaults {
  }
 }

-export const LanguageModelToolSchemaCompatibility = Schema.Literals(["gemini", "moonshot"])
-export type LanguageModelToolSchemaCompatibility = Schema.Schema.Type<typeof LanguageModelToolSchemaCompatibility>
+export const ModelToolSchemaCompatibility = Schema.Literals(["gemini", "moonshot"])
+export type ModelToolSchemaCompatibility = Schema.Schema.Type<typeof ModelToolSchemaCompatibility>

-export const LanguageModelMaxTokensFieldCompatibility = Schema.Literals(["max_completion_tokens", "max_tokens"])
-export type LanguageModelMaxTokensFieldCompatibility = Schema.Schema.Type<
-  typeof LanguageModelMaxTokensFieldCompatibility
->
-
-export class LanguageModelCompatibility extends Schema.Class<LanguageModelCompatibility>(
-  "LLM.LanguageModelCompatibility",
-)({
-  toolSchema: Schema.optional(LanguageModelToolSchemaCompatibility),
+export class ModelCompatibility extends Schema.Class<ModelCompatibility>("LLM.ModelCompatibility")({
+  toolSchema: Schema.optional(ModelToolSchemaCompatibility),
  reasoningField: Schema.optional(Schema.String),
-  maxTokensField: Schema.optional(LanguageModelMaxTokensFieldCompatibility),
 }) {}

-export namespace LanguageModelCompatibility {
-  export type Input = LanguageModelCompatibility | ConstructorParameters<typeof LanguageModelCompatibility>[0]
+export namespace ModelCompatibility {
+  export type Input = ModelCompatibility | ConstructorParameters<typeof ModelCompatibility>[0]

  /** Normalize model/upstream compatibility metadata without projecting requests. */
-  export const make = (input: Input) =>
-    input instanceof LanguageModelCompatibility ? input : new LanguageModelCompatibility(input)
+  export const make = (input: Input) => (input instanceof ModelCompatibility ? input : new ModelCompatibility(input))
 }

-export class LanguageModel<Options extends ProviderOptions = ProviderOptions> {
-  declare protected readonly _ProviderOptions: Options
+export class Model {
  readonly id: ModelID
  readonly provider: ProviderID
  readonly route: AnyRoute
-  readonly defaults?: LanguageModelDefaults
-  readonly compatibility?: LanguageModelCompatibility
+  readonly defaults?: ModelDefaults
+  readonly compatibility?: ModelCompatibility

-  constructor(input: LanguageModel.ConstructorInput) {
+  constructor(input: Model.ConstructorInput) {
    this.id = input.id
    this.provider = input.provider
    this.route = input.route
@@ -204,18 +193,17 @@ export class LanguageModel<Options extends ProviderOptions = ProviderOptions> {
    this.compatibility = input.compatibility
  }

-  static make<Options extends ProviderOptions = ProviderOptions>(input: LanguageModel.Input) {
-    return new LanguageModel<Options>({
+  static make(input: Model.Input) {
+    return new Model({
      id: ModelID.make(input.id),
      provider: ProviderID.make(input.provider),
      route: input.route,
-      defaults: input.defaults === undefined ? undefined : LanguageModelDefaults.make(input.defaults),
-      compatibility:
-        input.compatibility === undefined ? undefined : LanguageModelCompatibility.make(input.compatibility),
+      defaults: input.defaults === undefined ? undefined : ModelDefaults.make(input.defaults),
+      compatibility: input.compatibility === undefined ? undefined : ModelCompatibility.make(input.compatibility),
    })
  }

-  static input<Options extends ProviderOptions>(model: LanguageModel<Options>): LanguageModel.ConstructorInput {
+  static input(model: Model): Model.ConstructorInput {
    return {
      id: model.id,
      provider: model.provider,
@@ -225,40 +213,35 @@ export class LanguageModel<Options extends ProviderOptions = ProviderOptions> {
    }
  }

-  static update<Options extends ProviderOptions>(model: LanguageModel<Options>, patch: Partial<LanguageModel.Input>) {
+  static update(model: Model, patch: Partial<Model.Input>) {
    if (Object.keys(patch).length === 0) return model
-    return LanguageModel.make<Options>({
-      ...LanguageModel.input(model),
+    return Model.make({
+      ...Model.input(model),
      ...patch,
    })
  }
 }

-export namespace LanguageModel {
+export namespace Model {
  export type ConstructorInput = {
    readonly id: ModelID
    readonly provider: ProviderID
    readonly route: AnyRoute
-    readonly defaults?: LanguageModelDefaults
-    readonly compatibility?: LanguageModelCompatibility
+    readonly defaults?: ModelDefaults
+    readonly compatibility?: ModelCompatibility
  }

  export type Input = Omit<ConstructorInput, "id" | "provider" | "defaults" | "compatibility"> & {
    readonly id: string | ModelID
    readonly provider: string | ProviderID
-    readonly defaults?: LanguageModelDefaults.Input
-    readonly compatibility?: LanguageModelCompatibility.Input
+    readonly defaults?: ModelDefaults.Input
+    readonly compatibility?: ModelCompatibility.Input
  }
 }

-export type LanguageModelInput = LanguageModel.Input
+export type ModelInput = Model.Input

-export type LanguageModelProviderOptions<SelectedModel> =
-  SelectedModel extends LanguageModel<infer Options> ? Options : never
-
-export const LanguageModelSchema = Schema.declare((value): value is LanguageModel => value instanceof LanguageModel, {
-  expected: "LLM.LanguageModel",
-})
+export const ModelSchema = Schema.declare((value): value is Model => value instanceof Model, { expected: "LLM.Model" })

 export class CacheHint extends Schema.Class<CacheHint>("LLM.CacheHint")({
  type: Schema.Literals(["ephemeral", "persistent"]),
@@ -5,13 +5,13 @@ import {
  LLMEvent,
  LLMResponse,
  type FinishReasonDetails,
-  type AIError,
+  type LLMError,
  type LLMRequest,
  type UsageInput,
 } from "./schema"
 import { Context, Deferred, Effect, Latch, Layer, Queue, Scope, Stream } from "effect"

-export type Response = readonly LLMEvent[] | Stream.Stream<LLMEvent, AIError>
+export type Response = readonly LLMEvent[] | Stream.Stream<LLMEvent, LLMError>

 export type Gate = Readonly<{ started: Effect.Effect<void>; release: Effect.Effect<void> }>

@@ -63,7 +63,7 @@ export const textWithUsage = (value: string, id: string, inputTokens: number) =>

 export const tool = (id: string, name: string, input: unknown) => toolCalls(LLMEvent.toolCall({ id, name, input }))

-export const failAfter = (error: AIError, ...events: readonly LLMEvent[]) =>
+export const failAfter = (error: LLMError, ...events: readonly LLMEvent[]) =>
  Stream.fromIterable(events).pipe(Stream.concat(Stream.fail(error)))

 export const hangAfter = (...events: readonly LLMEvent[]) => Stream.concat(Stream.fromIterable(events), Stream.never)
@@ -99,6 +99,7 @@ export const layer = (options: LayerOptions = {}) =>
        )
      }) as LLMClientShape["stream"]
      const client = LLMClient.Service.of({
+        prepare: () => Effect.die("TestLLM does not prepare provider-native requests"),
        stream,
        generate: (request) =>
          stream(request).pipe(
@@ -2,12 +2,11 @@ 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 { LanguageModel } from "../src/schema"
+import { Model } from "../src/schema"
 import { testEffect } from "./lib/effect"
 import { dynamicResponse } from "./lib/http"

-const updateModel = (model: LanguageModel, patch: Partial<LanguageModel.Input>) => LanguageModel.update(model, patch)
+const updateModel = (model: Model, patch: Partial<Model.Input>) => Model.update(model, patch)

 const Json = Schema.fromJsonString(Schema.Unknown)
 const encodeJson = Schema.encodeSync(Json)
@@ -86,7 +85,7 @@ const configuredGemini = gemini.with({ endpoint: { baseURL: "https://fake.local"

 const request = LLM.request({
  id: "req_1",
-  model: LanguageModel.make({
+  model: Model.make({
    id: "fake-model",
    provider: "fake-provider",
    route: configuredFake,
@@ -140,7 +139,8 @@ describe("llm route", () => {

  it.effect("selects routes by model route value", () =>
    Effect.gen(function* () {
-      const prepared = yield* compileRequest(
+      const llm = yield* LLMClient.Service
+      const prepared = yield* llm.prepare(
        LLMRequest.update(request, { model: updateModel(request.model, { route: configuredGemini }) }),
      )

@@ -173,7 +173,7 @@ describe("llm route", () => {
        framing: fakeFraming,
      })

-      const prepared = yield* compileRequest(
+      const prepared = yield* (yield* LLMClient.Service).prepare(
        LLMRequest.update(request, { model: updateModel(request.model, { route: duplicate }) }),
      )

@@ -1,6 +1,6 @@
 import { Config } from "effect"
 import { Auth } from "../src/route"
-import type { LanguageModelFactory } from "../src/route/auth-options"
+import type { ModelFactory } from "../src/route/auth-options"
 import * as OpenAIChat from "../src/protocols/openai-chat"
 import * as AmazonBedrock from "../src/providers/amazon-bedrock"
 import * as Anthropic from "../src/providers/anthropic"
@@ -23,13 +23,13 @@ type BaseOptions = {
  readonly headers?: Record<string, string>
 }

-type LanguageModel = {
+type Model = {
  readonly id: string
 }

 declare const auth: Auth.Definition
-declare const optionalAuthModel: LanguageModelFactory<BaseOptions, "optional", LanguageModel>
-declare const requiredAuthModel: LanguageModelFactory<BaseOptions, "required", LanguageModel>
+declare const optionalAuthModel: ModelFactory<BaseOptions, "optional", Model>
+declare const requiredAuthModel: ModelFactory<BaseOptions, "required", Model>
 const configApiKey = Config.redacted("OPENAI_API_KEY")

 OpenAIChat.route.model({ id: "gpt-4.1-mini" })
@@ -4,12 +4,12 @@ import { Headers } from "effect/unstable/http"
 import { LLM } from "../src"
 import { Auth } from "../src/route/auth"
 import * as OpenAIChat from "../src/protocols/openai-chat"
-import { LanguageModel } from "../src/schema"
+import { Model } from "../src/schema"
 import { it } from "./lib/effect"

 const request = LLM.request({
  id: "req_auth",
-  model: LanguageModel.make({ id: "fake-model", provider: "fake", route: OpenAIChat.route }),
+  model: Model.make({ id: "fake-model", provider: "fake", route: OpenAIChat.route }),
  prompt: "hello",
 })

@@ -1,8 +1,7 @@
 import { describe, expect, test } from "bun:test"
 import { Effect } from "effect"
 import { CacheHint, LLM, Message } from "../src"
-import { Auth } from "../src/route"
-import { compileRequest } from "../src/route/client"
+import { Auth, LLMClient } from "../src/route"
 import { AmazonBedrock } from "../src/providers"
 import * as AnthropicMessages from "../src/protocols/anthropic-messages"
 import * as Gemini from "../src/protocols/gemini"
@@ -32,7 +31,7 @@ const geminiModel = Gemini.route
 describe("applyCachePolicy", () => {
  it.effect("undefined cache resolves to 'auto' (the recommended default)", () =>
    Effect.gen(function* () {
-      const prepared = yield* compileRequest(
+      const prepared = yield* LLMClient.prepare(
        LLM.request({
          model: anthropicModel,
          system: "You are concise.",
@@ -51,7 +50,7 @@ describe("applyCachePolicy", () => {

  it.effect("'auto' marks the last tool, first and last system parts, and final message boundary on Anthropic", () =>
    Effect.gen(function* () {
-      const prepared = yield* compileRequest(
+      const prepared = yield* LLMClient.prepare(
        LLM.request({
          model: anthropicModel,
          system: [
@@ -88,7 +87,7 @@ describe("applyCachePolicy", () => {

  it.effect("'auto' is a no-op on OpenAI (implicit caching protocol)", () =>
    Effect.gen(function* () {
-      const prepared = yield* compileRequest(
+      const prepared = yield* LLMClient.prepare(
        LLM.request({
          model: openaiModel,
          system: "Sys",
@@ -107,7 +106,7 @@ describe("applyCachePolicy", () => {

  it.effect("'auto' is a no-op on Gemini (out-of-band caching protocol)", () =>
    Effect.gen(function* () {
-      const prepared = yield* compileRequest(
+      const prepared = yield* LLMClient.prepare(
        LLM.request({
          model: geminiModel,
          system: "Sys",
@@ -124,7 +123,7 @@ describe("applyCachePolicy", () => {

  it.effect("'auto' on Bedrock emits cachePoint markers in the right places", () =>
    Effect.gen(function* () {
-      const prepared = yield* compileRequest(
+      const prepared = yield* LLMClient.prepare(
        LLM.request({
          model: bedrockModel,
          system: [
@@ -158,7 +157,7 @@ describe("applyCachePolicy", () => {

  it.effect("'none' disables auto placement even when manual hints exist", () =>
    Effect.gen(function* () {
-      const prepared = yield* compileRequest(
+      const prepared = yield* LLMClient.prepare(
        LLM.request({
          model: anthropicModel,
          system: "Sys",
@@ -177,7 +176,7 @@ describe("applyCachePolicy", () => {

  it.effect("granular object form: tools-only marks just tools", () =>
    Effect.gen(function* () {
-      const prepared = yield* compileRequest(
+      const prepared = yield* LLMClient.prepare(
        LLM.request({
          model: anthropicModel,
          system: "Sys",
@@ -196,7 +195,7 @@ describe("applyCachePolicy", () => {

  it.effect("auto policy preserves manual CacheHints on other parts", () =>
    Effect.gen(function* () {
-      const prepared = yield* compileRequest(
+      const prepared = yield* LLMClient.prepare(
        LLM.request({
          model: anthropicModel,
          system: [
@@ -242,7 +241,7 @@ describe("applyCachePolicy", () => {
      expect("cache" in tail ? tail.cache : undefined).toBeUndefined()
      expect(applyCachePolicy(applied)).toBe(applied)

-      const prepared = yield* compileRequest(request)
+      const prepared = yield* LLMClient.prepare(request)

      const body = prepared.body as {
        tools: Array<{ cache_control?: unknown }>
@@ -262,7 +261,7 @@ describe("applyCachePolicy", () => {

  it.effect("ttlSeconds in the policy flows through to wire markers", () =>
    Effect.gen(function* () {
-      const prepared = yield* compileRequest(
+      const prepared = yield* LLMClient.prepare(
        LLM.request({
          model: anthropicModel,
          system: "Sys",
@@ -279,7 +278,7 @@ describe("applyCachePolicy", () => {

  it.effect("messages: { tail: 2 } marks the last 2 message boundaries", () =>
    Effect.gen(function* () {
-      const prepared = yield* compileRequest(
+      const prepared = yield* LLMClient.prepare(
        LLM.request({
          model: anthropicModel,
          messages: [Message.user("u1"), Message.assistant("a1"), Message.user("u2"), Message.assistant("a2")],
@@ -297,7 +296,7 @@ describe("applyCachePolicy", () => {

  it.effect("'latest-assistant' marks the last assistant message", () =>
    Effect.gen(function* () {
-      const prepared = yield* compileRequest(
+      const prepared = yield* LLMClient.prepare(
        LLM.request({
          model: anthropicModel,
          messages: [Message.user("u1"), Message.assistant("a1"), Message.user("u2")],
@@ -1,12 +1,4 @@
-import {
-  LLM,
-  Message,
-  ToolCallPart,
-  ToolDefinition,
-  ToolResultPart,
-  type ContentPart,
-  type LanguageModel,
-} from "../src"
+import { LLM, Message, ToolCallPart, ToolDefinition, ToolResultPart, type ContentPart, type Model } from "../src"

 export const basicContinuation = ["system", "user-text", "assistant-text", "user-follow-up"] as const
 export const toolContinuation = ["tool-call", "tool-result"] as const
@@ -48,7 +40,7 @@ export const continuationTool = ToolDefinition.make({

 export function continuationRequest(input: {
  readonly id: string
-  readonly model: LanguageModel
+  readonly model: Model
  readonly features: ReadonlyArray<ContinuationFeature>
  readonly image?: string
 }) {
@@ -2,11 +2,11 @@ import { describe, expect, test } from "bun:test"
 import { LLM } from "../src"
 import * as OpenAIChat from "../src/protocols/openai-chat"
 import { Endpoint } from "../src/route"
-import { LanguageModel } from "../src/schema"
+import { Model } from "../src/schema"

 const request = () =>
  LLM.request({
-    model: LanguageModel.make({
+    model: Model.make({
      id: "model-1",
      provider: "test",
      route: OpenAIChat.route,
@@ -1,7 +1,7 @@
 import { describe, expect } from "bun:test"
 import { Effect, Layer, Ref } from "effect"
 import { Headers, HttpClient, HttpClientRequest, HttpClientResponse } from "effect/unstable/http"
-import { LLM, AIError } from "../src"
+import { LLM, LLMError } from "../src"
 import { LLMClient, RequestExecutor } from "../src/route"
 import * as OpenAIChat from "../src/protocols/openai-chat"
 import { dynamicResponse } from "./lib/http"
@@ -58,13 +58,13 @@ const countedResponsesLayer = (attempts: Ref.Ref<number>, responses: ReadonlyArr
    ),
  )

-const expectAIError = (error: unknown) => {
-  expect(error).toBeInstanceOf(AIError)
-  if (!(error instanceof AIError)) throw new Error("expected AIError")
+const expectLLMError = (error: unknown) => {
+  expect(error).toBeInstanceOf(LLMError)
+  if (!(error instanceof LLMError)) throw new Error("expected LLMError")
  return error
 }

-const errorHttp = (error: AIError) => ("http" in error.reason ? error.reason.http : undefined)
+const errorHttp = (error: LLMError) => ("http" in error.reason ? error.reason.http : undefined)

 describe("RequestExecutor", () => {
  it.effect("classifies context overflow responses", () =>
@@ -72,7 +72,7 @@ describe("RequestExecutor", () => {
      const executor = yield* RequestExecutor.Service
      const error = yield* executor.execute(request).pipe(Effect.flip)

-      expectAIError(error)
+      expectLLMError(error)
      expect(error.reason).toMatchObject({ _tag: "InvalidRequest", classification: "context-overflow" })
    }).pipe(
      Effect.provide(
@@ -85,17 +85,14 @@ describe("RequestExecutor", () => {
    ),
  )

-  it.effect("classifies generic HTTP 413 payload errors", () =>
+  it.effect("does not classify generic HTTP 413 payload errors as context overflow", () =>
    Effect.gen(function* () {
      const executor = yield* RequestExecutor.Service
      const error = yield* executor.execute(request).pipe(Effect.flip)

-      expectAIError(error)
-      expect(error.reason).toMatchObject({
-        _tag: "InvalidRequest",
-        classification: "payload-too-large",
-        http: { response: { status: 413 } },
-      })
+      expectLLMError(error)
+      expect(error.reason).toMatchObject({ _tag: "InvalidRequest" })
+      expect("classification" in error.reason ? error.reason.classification : undefined).toBeUndefined()
    }).pipe(Effect.provide(responsesLayer([new Response("request too large", { status: 413 })]))),
  )

@@ -104,7 +101,7 @@ describe("RequestExecutor", () => {
      const executor = yield* RequestExecutor.Service
      const error = yield* executor.execute(request).pipe(Effect.flip)

-      expectAIError(error)
+      expectLLMError(error)
      expect(error.reason).toMatchObject({ _tag: "InvalidRequest" })
      expect("classification" in error.reason ? error.reason.classification : undefined).toBeUndefined()
    }).pipe(Effect.provide(responsesLayer([new Response("invalid parameter", { status: 400 })]))),
@@ -117,7 +114,7 @@ describe("RequestExecutor", () => {
          const executor = yield* RequestExecutor.Service
          const error = yield* executor.execute(request).pipe(Effect.flip)

-          expectAIError(error)
+          expectLLMError(error)
          expect(error.reason).toMatchObject({ _tag: "RateLimit" })
        }).pipe(Effect.provide(responsesLayer([new Response(body, { status: 400 })])))

@@ -134,7 +131,7 @@ describe("RequestExecutor", () => {
          const executor = yield* RequestExecutor.Service
          const error = yield* executor.execute(request).pipe(Effect.flip)

-          expectAIError(error)
+          expectLLMError(error)
          expect(error.reason).toMatchObject({ _tag: "ProviderInternal" })
        }).pipe(Effect.provide(responsesLayer([new Response(body, { status: 400 })])))

@@ -148,7 +145,7 @@ describe("RequestExecutor", () => {
      const executor = yield* RequestExecutor.Service
      const error = yield* executor.execute(request).pipe(Effect.flip)

-      expectAIError(error)
+      expectLLMError(error)
      expect(error).toMatchObject({
        reason: {
          _tag: "RateLimit",
@@ -190,7 +187,7 @@ describe("RequestExecutor", () => {
      const executor = yield* RequestExecutor.Service
      const error = yield* executor.execute(request).pipe(Effect.flip)

-      expectAIError(error)
+      expectLLMError(error)
      expect(errorHttp(error)?.request.headers["x-safe"]).toBe("<redacted>")
      expect(errorHttp(error)?.response?.headers["x-safe"]).toBe("<redacted>")
    }).pipe(
@@ -204,7 +201,7 @@ describe("RequestExecutor", () => {
      const executor = yield* RequestExecutor.Service
      const error = yield* executor.execute(request).pipe(Effect.flip)

-      expectAIError(error)
+      expectLLMError(error)
      expect(error.reason).toMatchObject({ _tag: "RateLimit" })
      expect(error.reason._tag === "RateLimit" ? error.reason.rateLimit : undefined).toEqual({
        retryAfterMs: 0,
@@ -237,7 +234,7 @@ describe("RequestExecutor", () => {
      const executor = yield* RequestExecutor.Service
      const error = yield* executor.execute(request).pipe(Effect.flip)

-      expectAIError(error)
+      expectLLMError(error)
      expect(error.reason).toMatchObject({ _tag: "ProviderInternal" })
      expect(errorHttp(error)?.rateLimit).toEqual({
        retryAfterMs: 0,
@@ -280,7 +277,7 @@ describe("RequestExecutor", () => {
        ),
      )

-      expectAIError(error)
+      expectLLMError(error)
      expect(error.reason).toMatchObject({ _tag: "ProviderInternal", status: 503 })
      expect(yield* Ref.get(attempts)).toBe(1)
    }),
@@ -293,7 +290,7 @@ describe("RequestExecutor", () => {
          const executor = yield* RequestExecutor.Service
          const error = yield* executor.execute(request).pipe(Effect.flip)

-          expectAIError(error)
+          expectLLMError(error)
          expect(error.reason).toMatchObject({ _tag: "ProviderInternal", status })
        }).pipe(
          Effect.provide(
@@ -316,7 +313,7 @@ describe("RequestExecutor", () => {
      const executor = yield* RequestExecutor.Service
      const error = yield* executor.execute(request).pipe(Effect.flip)

-      expectAIError(error)
+      expectLLMError(error)
      expect(error.reason).toMatchObject({ _tag: "Authentication" })
      expect(errorHttp(error)?.bodyTruncated).toBe(true)
      expect(errorHttp(error)?.body).toHaveLength(16_384)
@@ -335,7 +332,7 @@ describe("RequestExecutor", () => {
      const executor = yield* RequestExecutor.Service
      const error = yield* executor.execute(request).pipe(Effect.flip)

-      expectAIError(error)
+      expectLLMError(error)
      expect(errorHttp(error)?.body).toContain('"key":"<redacted>"')
      expect(errorHttp(error)?.body).toContain("api_key=<redacted>")
      expect(errorHttp(error)?.body).not.toContain("body-secret")
@@ -356,7 +353,7 @@ describe("RequestExecutor", () => {
      const executor = yield* RequestExecutor.Service
      const error = yield* executor.execute(secretRequest).pipe(Effect.flip)

-      expectAIError(error)
+      expectLLMError(error)
      expect(errorHttp(error)?.body).toContain("provider echoed <redacted>")
      expect(errorHttp(error)?.body).toContain("authorization <redacted>")
      expect(errorHttp(error)?.body).not.toContain("query-secret-123")
@@ -395,7 +392,7 @@ describe("RequestExecutor", () => {
        Effect.flip,
      )

-      expectAIError(error)
+      expectLLMError(error)
      expect(error.reason).toMatchObject({ _tag: "InvalidProviderOutput" })
      expect(yield* Ref.get(attempts)).toBe(1)
    }),
@@ -1,5 +1,5 @@
 import { describe, expect, test } from "bun:test"
-import { AIError, ImageInput, LanguageModel, LLM, LLMClient, Provider } from "@opencode-ai/ai"
+import { ImageInput, LLM, LLMClient, Provider } from "@opencode-ai/ai"
 import { Route, Protocol } from "@opencode-ai/ai/route"
 import { Provider as ProviderSubpath } from "@opencode-ai/ai/provider"
 import {
@@ -26,8 +26,6 @@ describe("public exports", () => {
    expect(LLM.request).toBeFunction()
    expect(LLMClient.Service).toBeFunction()
    expect(LLMClient.layer).toBeDefined()
-    expect(AIError).toBeFunction()
-    expect(LanguageModel.make).toBeFunction()
    expect(ImageInput.bytes).toBeFunction()
    expect(Provider.make).toBeFunction()
    expect(ProviderSubpath.make).toBe(Provider.make)
@@ -55,7 +53,9 @@ describe("public exports", () => {
    expect(CloudflareWorkersAI.configure).toBeFunction()
    expect(CloudflareWorkersAI.configure({ accountId: "fixture", apiKey: "fixture" }).model).toBeFunction()
    expect(OpenRouter.model).toBeFunction()
+    expect(OpenRouter.provider.model).toBe(OpenRouter.model)
    expect(XAI.model).toBeFunction()
+    expect(XAI.provider.model).toBe(XAI.model)
    expect(XAI.provider.responses).toBe(XAI.responses)
    expect(XAI.provider.chat).toBe(XAI.chat)
    expect(XAI.configure({ apiKey: "fixture" }).responses("grok-4.3").route.id).toBe("openai-responses")
@@ -1,47 +0,0 @@
-import { Schema } from "effect"
-import { LLM, type LanguageModel, type LanguageModelProviderOptions, 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: LanguageModel
-LLM.request({ model: generic, prompt: "Hello", providerOptions: { arbitrary: { option: true } } })
-
-const options: LanguageModelProviderOptions<typeof model> = { example: { mode: "fast" } }
-void options
@@ -6,7 +6,7 @@ import {
  GenerationOptions,
  LLMRequest,
  Message,
-  LanguageModel,
+  Model,
  ToolCallPart,
  ToolChoice,
  ToolDefinition,
@@ -20,13 +20,13 @@ describe("llm constructors", () => {
  test("builds canonical schema classes from ergonomic input", () => {
    const request = LLM.request({
      id: "req_1",
-      model: LanguageModel.make({ id: "fake-model", provider: "fake", route: chatRoute }),
+      model: Model.make({ id: "fake-model", provider: "fake", route: chatRoute }),
      system: "You are concise.",
      prompt: "Say hello.",
    })

    expect(request).toBeInstanceOf(LLMRequest)
-    expect(request.model).toBeInstanceOf(LanguageModel)
+    expect(request.model).toBeInstanceOf(Model)
    expect(request.messages[0]).toBeInstanceOf(Message)
    expect(request.system).toEqual([{ type: "text", text: "You are concise." }])
    expect(request.messages[0]?.content).toEqual([{ type: "text", text: "Say hello." }])
@@ -37,7 +37,7 @@ describe("llm constructors", () => {
  test("updates requests without spreading schema class instances", () => {
    const base = LLM.request({
      id: "req_1",
-      model: LanguageModel.make({ id: "fake-model", provider: "fake", route: chatRoute }),
+      model: Model.make({ id: "fake-model", provider: "fake", route: chatRoute }),
      prompt: "Say hello.",
    })
    const updated = LLMRequest.update(base, {
@@ -54,7 +54,7 @@ describe("llm constructors", () => {

  test("keeps request options separate from route defaults", () => {
    const request = LLM.request({
-      model: LanguageModel.make({
+      model: Model.make({
        id: "fake-model",
        provider: "fake",
        route: chatRoute.with({
@@ -81,7 +81,7 @@ describe("llm constructors", () => {
  test("updates canonical requests from the request datatype", () => {
    const base = LLM.request({
      id: "req_1",
-      model: LanguageModel.make({ id: "fake-model", provider: "fake", route: chatRoute }),
+      model: Model.make({ id: "fake-model", provider: "fake", route: chatRoute }),
      prompt: "Say hello.",
    })
    const updated = LLMRequest.update(base, { messages: [...base.messages, Message.assistant("Hi.")] })
@@ -94,19 +94,19 @@ describe("llm constructors", () => {
  })

  test("updates canonical models from the model datatype", () => {
-    const base = LanguageModel.make({
+    const base = Model.make({
      id: "fake-model",
      provider: "fake",
      route: chatRoute,
    })
-    const updated = LanguageModel.update(base, {
+    const updated = Model.update(base, {
      route: responsesRoute,
      defaults: { generation: { maxTokens: 20 } },
      compatibility: { toolSchema: "gemini" },
    })
-    const updatedInput = LanguageModel.input(updated)
+    const updatedInput = Model.input(updated)

-    expect(updated).toBeInstanceOf(LanguageModel)
+    expect(updated).toBeInstanceOf(Model)
    expect(String(updated.id)).toBe("fake-model")
    expect(updated.route).toBe(responsesRoute)
    expect(updated.defaults?.generation).toEqual({ maxTokens: 20 })
@@ -114,7 +114,7 @@ describe("llm constructors", () => {
    expect(updatedInput.defaults).toBe(updated.defaults)
    expect(updatedInput.compatibility).toBe(updated.compatibility)
    expect(String(updatedInput.provider)).toBe("fake")
-    expect(LanguageModel.update(updated, {})).toBe(updated)
+    expect(Model.update(updated, {})).toBe(updated)
  })

  test("carries model defaults and compatibility through route model selection", () => {
@@ -155,7 +155,7 @@ describe("llm constructors", () => {
    expect(ToolChoice.make("required")).toEqual(new ToolChoice({ type: "required" }))
    expect(
      LLM.request({
-        model: LanguageModel.make({
+        model: Model.make({
          id: "fake-model",
          provider: "fake",
          route: chatRoute,
@@ -181,7 +181,7 @@ describe("llm constructors", () => {
      { type: "text", text: "Use parameterized SQL.", cache: new CacheHint({ type: "ephemeral" }) },
    ])
    const request = LLM.request({
-      model: LanguageModel.make({ id: "fake-model", provider: "fake", route: chatRoute }),
+      model: Model.make({ id: "fake-model", provider: "fake", route: chatRoute }),
      system: "Initial operator prompt.",
      messages: [Message.user("Review this."), update],
    })
@@ -4,7 +4,6 @@ import { HttpClientRequest } from "effect/unstable/http"
 import { LLM, mergeProviderOptions } from "../src"
 import { AnthropicMessages, OpenAIChat } from "../src/protocols"
 import { Auth, LLMClient } from "../src/route"
-import { compileRequest } from "../src/route/client"
 import { it } from "./lib/effect"
 import { dynamicResponse } from "./lib/http"
 import { deltaChunk } from "./lib/openai-chunks"
@@ -45,7 +44,7 @@ describe("request option precedence", () => {
    })
  })

-  it.effect("compiles bodies with route defaults, model defaults, and call options in order", () =>
+  it.effect("prepares bodies with route defaults, model defaults, and call options in order", () =>
    Effect.gen(function* () {
      const route = OpenAIChat.route.with({
        endpoint: { baseURL: "https://api.openai.test/v1/" },
@@ -60,7 +59,7 @@ describe("request option precedence", () => {
          providerOptions: { openai: { reasoningEffort: "medium" } },
        },
      })
-      const prepared = yield* compileRequest(
+      const prepared = yield* LLMClient.prepare<OpenAIChat.OpenAIChatBody>(
        LLM.request({
          model,
          prompt: "Say hello.",
@@ -137,61 +136,24 @@ describe("request option precedence", () => {
    ),
  )

-  it.effect("transforms the final HTTP request after serialization and authentication", () =>
-    LLMClient.generate(
-      LLM.request({
-        model: OpenAIChat.route
-          .with({ endpoint: { baseURL: "https://api.openai.test/v1/" }, auth: Auth.bearer("fresh-key") })
-          .model({ id: "gpt-4o-mini" }),
-        prompt: "Say hello.",
-      }),
-      {
-        transform: (request) =>
-          Effect.sync(() => {
-            expect(request.headers.authorization).toBe("Bearer fresh-key")
-            request.url = "https://proxy.test/v1/chat/completions"
-            request.headers["x-plugin"] = "transformed"
-            request.body = JSON.stringify({ transformed: true })
-          }),
-      },
-    ).pipe(
-      Effect.provide(
-        dynamicResponse((input) =>
-          Effect.gen(function* () {
-            const web = yield* HttpClientRequest.toWeb(input.request).pipe(Effect.orDie)
-            expect(web.url).toBe("https://proxy.test/v1/chat/completions")
-            expect(web.headers.get("x-plugin")).toBe("transformed")
-            expect(decodeJson(input.text)).toEqual({ transformed: true })
-            return input.respond(sseEvents(deltaChunk({}, "stop")), {
-              headers: { "content-type": "text/event-stream" },
-            })
-          }),
-        ),
-      ),
-    ),
-  )
+  it.effect("rejects raw body overlays for protocol-owned roots", () =>
+    Effect.gen(function* () {
+      const model = OpenAIChat.route
+        .with({ endpoint: { baseURL: "https://api.openai.test/v1/" }, auth: Auth.bearer("test") })
+        .model({ id: "gpt-4o-mini" })
+      const error = yield* LLMClient.prepare(
+        LLM.request({
+          model,
+          prompt: "Say hello.",
+          http: { body: { model: "gpt-5", messages: [], tools: [] } },
+        }),
+      ).pipe(Effect.flip)

-  it.effect("applies raw body overlays after protocol lowering", () =>
-    LLMClient.generate(
-      LLM.request({
-        model: OpenAIChat.route
-          .with({ endpoint: { baseURL: "https://api.openai.test/v1/" }, auth: Auth.bearer("test") })
-          .model({ id: "gpt-4o-mini" }),
-        prompt: "Say hello.",
-        http: { body: { model: "gpt-5", messages: [], tools: [] } },
-      }),
-    ).pipe(
-      Effect.provide(
-        dynamicResponse((input) =>
-          Effect.gen(function* () {
-            expect(decodeJson(input.text)).toMatchObject({ model: "gpt-5", messages: [], tools: [] })
-            return input.respond(sseEvents(deltaChunk({}, "stop")), {
-              headers: { "content-type": "text/event-stream" },
-            })
-          }),
-        ),
-      ),
-    ),
+      expect(error.reason).toMatchObject({
+        _tag: "InvalidRequest",
+        message: "http.body cannot overlay protocol-owned field(s): model, messages, tools",
+      })
+    }),
  )

  it.effect("uses model output limits after route limits and before call maxTokens", () =>
@@ -202,8 +164,10 @@ describe("request option precedence", () => {
        limits: { output: 128 },
      })
      const model = route.model({ id: "claude-sonnet-4-5", defaults: { limits: { output: 64 } } })
-      const withoutMaxTokens = yield* compileRequest(LLM.request({ model, prompt: "Say hello.", cache: "none" }))
-      const withMaxTokens = yield* compileRequest(
+      const withoutMaxTokens = yield* LLMClient.prepare<AnthropicMessages.AnthropicMessagesBody>(
+        LLM.request({ model, prompt: "Say hello.", cache: "none" }),
+      )
+      const withMaxTokens = yield* LLMClient.prepare<AnthropicMessages.AnthropicMessagesBody>(
        LLM.request({ model, prompt: "Say hello.", cache: "none", generation: { maxTokens: 32 } }),
      )

@@ -6,6 +6,7 @@ describe("provider error classification", () => {
  test("classifies provider token limit messages as context overflow", () => {
    const messages = [
      "tokens in request more than max tokens allowed",
+      '{"error":{"type":"request_too_large","message":"Request exceeds the maximum size"}}',
      "Requested token count exceeds the model's maximum context length of 131072 tokens.",
      "Input length (265330) exceeds model's maximum context length (262144).",
      "Input length 131393 exceeds the maximum allowed input length of 131040 tokens.",
@@ -18,24 +19,6 @@ describe("provider error classification", () => {
    expect(messages.every(isContextOverflow)).toBe(true)
  })

-  test("classifies request size failures separately from context overflow", () => {
-    const failures = [
-      classifyProviderFailure({ message: "request too large", status: 413 }),
-      classifyProviderFailure({
-        message: '{"error":{"type":"request_too_large","message":"Request exceeds the maximum size"}}',
-        status: 400,
-      }),
-      classifyProviderFailure({ message: "upstream request entity too large", status: 502 }),
-    ]
-
-    expect(failures).toEqual(
-      failures.map((failure) =>
-        expect.objectContaining({ _tag: "InvalidRequest", classification: "payload-too-large" }),
-      ),
-    )
-    expect(isContextOverflow("413 status code (no body)")).toBe(false)
-  })
-
  test("does not classify rate limits as context overflow", () => {
    const messages = [
      "Throttling error: Too many tokens, please wait before trying again.",
@@ -76,10 +59,9 @@ describe("provider error classification", () => {
  })

  test("classifies transient client statuses as provider internal", () => {
-    expect([408, 409].map((status) => classifyProviderFailure({ message: `HTTP ${status}`, status })._tag)).toEqual([
-      "ProviderInternal",
-      "ProviderInternal",
-    ])
+    expect(
+      [408, 409].map((status) => classifyProviderFailure({ message: `HTTP ${status}`, status })._tag),
+    ).toEqual(["ProviderInternal", "ProviderInternal"])
  })

  test("classifies nested provider codes when a top-level code is also present", () => {
@@ -1,13 +0,0 @@
-import { LLM } from "../../src"
-import { AnthropicCompatible } from "../../src/providers"
-
-const model = AnthropicCompatible.configure({ baseURL: "https://example.com" }).model("claude")
-
-LLM.request({ model, prompt: "Hello", providerOptions: { anthropic: { effort: "high" } } })
-
-LLM.request({
-  model,
-  prompt: "Hello",
-  // @ts-expect-error Anthropic effort must be a string.
-  providerOptions: { anthropic: { effort: 1 } },
-})
@@ -1,13 +0,0 @@
-import { LLM } from "../../src"
-import { Anthropic } from "../../src/providers"
-
-const model = Anthropic.provider.model("claude-sonnet-4-5")
-
-LLM.request({ model, prompt: "Hello", providerOptions: { anthropic: { thinking: { type: "adaptive" } } } })
-
-LLM.request({
-  model,
-  prompt: "Hello",
-  // @ts-expect-error Anthropic thinking modes are a fixed union.
-  providerOptions: { anthropic: { thinking: { type: "automatic" } } },
-})
@@ -1,13 +0,0 @@
-import { LLM } from "../../src"
-import { Azure } from "../../src/providers"
-
-const model = Azure.configure({ resourceName: "example" }).responses("deployment")
-
-LLM.request({ model, prompt: "Hello", providerOptions: { openai: { store: false } } })
-
-LLM.request({
-  model,
-  prompt: "Hello",
-  // @ts-expect-error Azure OpenAI store must be boolean.
-  providerOptions: { openai: { store: "false" } },
-})
@@ -1,13 +0,0 @@
-import { LLM } from "../../src"
-import { CloudflareWorkersAI } from "../../src/providers"
-
-const model = CloudflareWorkersAI.configure({ accountId: "account", apiKey: "test" }).model("model")
-
-LLM.request({ model, prompt: "Hello", providerOptions: { openai: { promptCacheKey: "cache" } } })
-
-LLM.request({
-  model,
-  prompt: "Hello",
-  // @ts-expect-error Cloudflare's OpenAI-compatible prompt cache key must be a string.
-  providerOptions: { openai: { promptCacheKey: 1 } },
-})
@@ -1,13 +0,0 @@
-import { LLM } from "../../src"
-import { GitHubCopilot } from "../../src/providers"
-
-const model = GitHubCopilot.configure({ baseURL: "https://example.com" }).model("gpt-5")
-
-LLM.request({ model, prompt: "Hello", providerOptions: { openai: { reasoningSummary: "auto" } } })
-
-LLM.request({
-  model,
-  prompt: "Hello",
-  // @ts-expect-error Copilot reasoning summaries use the OpenAI union.
-  providerOptions: { openai: { reasoningSummary: "full" } },
-})
@@ -1,13 +0,0 @@
-import { LLM } from "../../src"
-import { GoogleVertexChat } from "../../src/providers"
-
-const model = GoogleVertexChat.configure({ accessToken: "test", project: "project" }).model("gemini")
-
-LLM.request({ model, prompt: "Hello", providerOptions: { openai: { serviceTier: "priority" } } })
-
-LLM.request({
-  model,
-  prompt: "Hello",
-  // @ts-expect-error Vertex OpenAI-compatible service tiers use the OpenAI union.
-  providerOptions: { openai: { serviceTier: "premium" } },
-})
@@ -1,13 +0,0 @@
-import { LLM } from "../../src"
-import { GoogleVertexMessages } from "../../src/providers"
-
-const model = GoogleVertexMessages.configure({ accessToken: "test", project: "project" }).model("claude")
-
-LLM.request({ model, prompt: "Hello", providerOptions: { anthropic: { effort: "medium" } } })
-
-LLM.request({
-  model,
-  prompt: "Hello",
-  // @ts-expect-error Vertex Anthropic effort must be a string.
-  providerOptions: { anthropic: { effort: false } },
-})
@@ -1,13 +0,0 @@
-import { LLM } from "../../src"
-import { GoogleVertexResponses } from "../../src/providers"
-
-const model = GoogleVertexResponses.configure({ accessToken: "test", project: "project" }).model("gemini")
-
-LLM.request({ model, prompt: "Hello", providerOptions: { openresponses: { textVerbosity: "high" } } })
-
-LLM.request({
-  model,
-  prompt: "Hello",
-  // @ts-expect-error Vertex Responses verbosity uses the Open Responses union.
-  providerOptions: { openresponses: { textVerbosity: "verbose" } },
-})
@@ -1,17 +0,0 @@
-import { LLM } from "../../src"
-import { GoogleVertex } from "../../src/providers"
-
-const model = GoogleVertex.provider.configure({ apiKey: "test" }).model("gemini-2.5-pro")
-
-LLM.request({
-  model,
-  prompt: "Hello",
-  providerOptions: { gemini: { thinkingConfig: { includeThoughts: true } } },
-})
-
-LLM.request({
-  model,
-  prompt: "Hello",
-  // @ts-expect-error Vertex Gemini includeThoughts must be boolean.
-  providerOptions: { gemini: { thinkingConfig: { includeThoughts: "yes" } } },
-})
@@ -1,47 +0,0 @@
-import { LLM } from "../../src"
-import { Google } from "../../src/providers"
-
-const model = Google.provider.model("gemini-2.5-pro")
-
-LLM.request({
-  model,
-  prompt: "Hello",
-  providerOptions: { gemini: { thinkingConfig: { thinkingBudget: 1024 } } },
-})
-
-LLM.request({
-  model,
-  prompt: "Hello",
-  providerOptions: {
-    gemini: {
-      // @ts-expect-error Gemini safety settings require a threshold for every category.
-      safetySettings: [{ category: "HARM_CATEGORY_HATE_SPEECH" }],
-    },
-  },
-})
-
-LLM.request({
-  model,
-  prompt: "Hello",
-  providerOptions: {
-    gemini: {
-      cachedContent: "cachedContents/example",
-      safetySettings: [{ category: "HARM_CATEGORY_HATE_SPEECH", threshold: "BLOCK_ONLY_HIGH" }],
-      serviceTier: "future-tier",
-      thinkingConfig: { thinkingLevel: "high", includeThoughts: true },
-    },
-  },
-})
-
-LLM.request({
-  model,
-  prompt: "Hello",
-  // @ts-expect-error Gemini thinking budgets must be numeric.
-  providerOptions: { gemini: { thinkingConfig: { thinkingBudget: "large" } } },
-})
-
-LLM.request({
-  model,
-  prompt: "Hello",
-  providerOptions: { gemini: { thinkingConfig: { thinkingLevel: "maximum" } } },
-})
@@ -1,13 +0,0 @@
-import { LLM } from "../../src"
-import { OpenAICompatibleResponses } from "../../src/providers"
-
-const model = OpenAICompatibleResponses.configure({ baseURL: "https://example.com" }).model("model")
-
-LLM.request({ model, prompt: "Hello", providerOptions: { openresponses: { reasoningSummary: "detailed" } } })
-
-LLM.request({
-  model,
-  prompt: "Hello",
-  // @ts-expect-error Open Responses reasoning summaries use a fixed union.
-  providerOptions: { openresponses: { reasoningSummary: "full" } },
-})
@@ -1,13 +0,0 @@
-import { LLM } from "../../src"
-import { OpenAICompatible } from "../../src/providers"
-
-const model = OpenAICompatible.deepseek.model("deepseek-chat")
-
-LLM.request({ model, prompt: "Hello", providerOptions: { openai: { store: false } } })
-
-LLM.request({
-  model,
-  prompt: "Hello",
-  // @ts-expect-error OpenAI-compatible store must be boolean.
-  providerOptions: { openai: { store: "false" } },
-})
@@ -1,13 +0,0 @@
-import { LLM } from "../../src"
-import { OpenAI } from "../../src/providers"
-
-const model = OpenAI.responses("gpt-5")
-
-LLM.request({ model, prompt: "Hello", providerOptions: { openai: { reasoningEffort: "high" } } })
-
-LLM.request({
-  model,
-  prompt: "Hello",
-  // @ts-expect-error OpenAI reasoning effort must be a string.
-  providerOptions: { openai: { reasoningEffort: 1 } },
-})
@@ -1,35 +0,0 @@
-import { LLM } from "../../src"
-import { OpenRouter } from "../../src/providers"
-
-const model = OpenRouter.provider.model("anthropic/claude-sonnet-4.5")
-
-LLM.request({ model, prompt: "Hello", providerOptions: { openrouter: { usage: true } } })
-
-LLM.request({
-  model,
-  prompt: "Hello",
-  providerOptions: {
-    openrouter: {
-      models: ["google/gemini-3.1-pro"],
-      provider: {
-        order: ["anthropic"],
-        require_parameters: true,
-        data_collection: "future-policy",
-        sort: "future-sort",
-        max_price: { prompt: "0.50" },
-      },
-      reasoning: { effort: "future-effort", exclude: false },
-      plugins: [{ id: "future-plugin", enabled: true }],
-      web_search_options: { engine: "future-engine" },
-      debug: { echo_upstream_body: true },
-      user: "user_123",
-    },
-  },
-})
-
-LLM.request({
-  model,
-  prompt: "Hello",
-  // @ts-expect-error OpenRouter usage must be boolean or an option record.
-  providerOptions: { openrouter: { usage: "yes" } },
-})
@@ -1,13 +0,0 @@
-import { LLM } from "../../src"
-import { XAI } from "../../src/providers"
-
-const model = XAI.provider.model("grok-4")
-
-LLM.request({ model, prompt: "Hello", providerOptions: { xai: { reasoningEffort: "high" } } })
-
-LLM.request({
-  model,
-  prompt: "Hello",
-  // @ts-expect-error xAI's OpenAI-compatible reasoning effort must be a string.
-  providerOptions: { xai: { reasoningEffort: true } },
-})
@@ -21,8 +21,6 @@ describe("provider package entrypoints", () => {
      import("@opencode-ai/ai/providers/google-vertex/chat"),
      import("@opencode-ai/ai/providers/google-vertex/responses"),
      import("@opencode-ai/ai/providers/google-vertex/messages"),
-      import("@opencode-ai/ai/providers/openrouter"),
-      import("@opencode-ai/ai/providers/xai"),
    ])

    for (const module of modules) expect(module.model).toBeFunction()
@@ -31,35 +29,6 @@ describe("provider package entrypoints", () => {
    expect(modules[12].model).toBe(modules[13].model)
  })

-  test("maps OpenRouter and xAI package settings onto executable models", async () => {
-    const OpenRouter = await import("@opencode-ai/ai/providers/openrouter")
-    const XAI = await import("@opencode-ai/ai/providers/xai")
-    const settings = {
-      apiKey: "fixture",
-      baseURL: "https://provider.example.test/v1",
-      headers: { "x-application": "opencode" },
-      body: { service_tier: "priority" },
-      limits: { context: 200_000, output: 64_000 },
-    }
-    const openrouter = OpenRouter.model("anthropic/claude-sonnet-4", {
-      ...settings,
-      providerOptions: { openrouter: { usage: true } },
-    })
-    const xai = XAI.model("grok-4", {
-      ...settings,
-      providerOptions: { xai: { reasoningEffort: "high" } },
-    })
-
-    for (const selected of [openrouter, xai]) {
-      expect(selected.route.endpoint.baseURL).toBe(settings.baseURL)
-      expect(selected.route.defaults.headers).toEqual(settings.headers)
-      expect(selected.route.defaults.http?.body).toEqual(settings.body)
-      expect(selected.route.defaults.limits).toEqual(settings.limits)
-    }
-    expect(openrouter.route.defaults.providerOptions).toEqual({ openrouter: { usage: true } })
-    expect(xai.route.defaults.providerOptions).toMatchObject({ xai: { reasoningEffort: "high", store: false } })
-  })
-
  test("maps package settings onto the executable model", () => {
    const selected = model("gpt-5", {
      apiKey: "fixture",
@@ -1,9 +1,9 @@
 import { Provider } from "../src/provider"
-import { ProviderID, type LanguageModel } from "../src/schema"
+import { ProviderID, type Model } from "../src/schema"

-declare const model: (id: string) => LanguageModel
-declare const requiredModel: (id: string, options: { readonly baseURL: string }) => LanguageModel
-declare const chat: (id: string, options: { readonly apiKey: string }) => LanguageModel
+declare const model: (id: string) => Model
+declare const requiredModel: (id: string, options: { readonly baseURL: string }) => Model
+declare const chat: (id: string, options: { readonly apiKey: string }) => Model

 Provider.make({
  id: ProviderID.make("example"),
@@ -1,6 +1,6 @@
 import { describe, expect } from "bun:test"
 import { Effect } from "effect"
-import { LLM, AIError, Message, ToolCallPart } from "../../src"
+import { LLM, LLMError, Message, ToolCallPart } from "../../src"
 import { LLMClient } from "../../src/route"
 import * as Anthropic from "../../src/providers/anthropic"
 import { weatherToolName } from "../recorded-scenarios"
@@ -37,7 +37,7 @@ describe("Anthropic Messages sad-path recorded", () => {
    Effect.gen(function* () {
      const error = yield* LLMClient.generate(malformedToolOrderRequest).pipe(Effect.flip)

-      expect(error).toBeInstanceOf(AIError)
+      expect(error).toBeInstanceOf(LLMError)
      expect(error.reason).toMatchObject({ _tag: "InvalidRequest" })
      expect(error.message).toContain("HTTP 400")
    }),
@@ -1,9 +1,8 @@
 import { describe, expect } from "bun:test"
 import { Effect } from "effect"
 import { HttpClientRequest } from "effect/unstable/http"
-import { CacheHint, LLM, AIError, LLMRequest, Message, ToolCallPart, ToolDefinition, Usage } from "../../src"
+import { CacheHint, LLM, LLMError, LLMRequest, Message, ToolCallPart, ToolDefinition, Usage } from "../../src"
 import { Auth, LLMClient } from "../../src/route"
-import { compileRequest } from "../../src/route/client"
 import * as AnthropicMessages from "../../src/protocols/anthropic-messages"
 import { continuationRequest, nativeAnthropicMessagesContinuation } from "../continuation-scenarios"
 import { it } from "../lib/effect"
@@ -45,7 +44,7 @@ const expectToolResult = (body: AnthropicMessages.AnthropicMessagesBody): Anthro
 describe("Anthropic Messages route", () => {
  it.effect("prepares Anthropic Messages target", () =>
    Effect.gen(function* () {
-      const prepared = yield* compileRequest(request)
+      const prepared = yield* LLMClient.prepare(request)

      expect(prepared.body).toEqual({
        model: "claude-sonnet-4-5",
@@ -60,7 +59,7 @@ describe("Anthropic Messages route", () => {

  it.effect("lowers adaptive thinking settings with effort", () =>
    Effect.gen(function* () {
-      const prepared = yield* compileRequest(
+      const prepared = yield* LLMClient.prepare<AnthropicMessages.AnthropicMessagesBody>(
        LLMRequest.update(request, {
          providerOptions: {
            anthropic: { thinking: { type: "adaptive", display: "summarized" }, effort: "low" },
@@ -77,17 +76,17 @@ describe("Anthropic Messages route", () => {

  it.effect("normalizes enabled and disabled thinking settings", () =>
    Effect.gen(function* () {
-      const enabled = yield* compileRequest(
+      const enabled = yield* LLMClient.prepare<AnthropicMessages.AnthropicMessagesBody>(
        LLMRequest.update(request, {
          providerOptions: { anthropic: { thinking: { type: "enabled", budgetTokens: 1_024 } } },
        }),
      )
-      const legacy = yield* compileRequest(
+      const legacy = yield* LLMClient.prepare<AnthropicMessages.AnthropicMessagesBody>(
        LLMRequest.update(request, {
          providerOptions: { anthropic: { thinking: { type: "enabled", budget_tokens: 2_048 } } },
        }),
      )
-      const disabled = yield* compileRequest(
+      const disabled = yield* LLMClient.prepare<AnthropicMessages.AnthropicMessagesBody>(
        LLMRequest.update(request, {
          providerOptions: { anthropic: { thinking: { type: "disabled" } } },
        }),
@@ -101,7 +100,7 @@ describe("Anthropic Messages route", () => {

  it.effect("rejects enabled thinking without a budget", () =>
    Effect.gen(function* () {
-      const error = yield* compileRequest(
+      const error = yield* LLMClient.prepare(
        LLMRequest.update(request, {
          providerOptions: { anthropic: { thinking: { type: "enabled" } } },
        }),
@@ -113,7 +112,7 @@ describe("Anthropic Messages route", () => {

  it.effect("lowers chronological system updates natively for Claude Opus 4.8 with cache hints", () =>
    Effect.gen(function* () {
-      const prepared = yield* compileRequest(
+      const prepared = yield* LLMClient.prepare<AnthropicMessages.AnthropicMessagesBody>(
        LLM.request({
          model: opus48,
          messages: [
@@ -138,7 +137,7 @@ describe("Anthropic Messages route", () => {

  it.effect("lowers chronological system updates to wrapped user text for unsupported Anthropic models", () =>
    Effect.gen(function* () {
-      const prepared = yield* compileRequest(
+      const prepared = yield* LLMClient.prepare<AnthropicMessages.AnthropicMessagesBody>(
        LLM.request({
          model,
          messages: [
@@ -165,7 +164,7 @@ describe("Anthropic Messages route", () => {

  it.effect("rejects non-text chronological system update content before send", () =>
    Effect.gen(function* () {
-      const error = yield* compileRequest(
+      const error = yield* LLMClient.prepare(
        LLM.request({
          model: opus48,
          messages: [
@@ -182,7 +181,7 @@ describe("Anthropic Messages route", () => {
  it.effect("falls back for unsupported native chronological system update placement", () =>
    Effect.gen(function* () {
      expect(
-        (yield* compileRequest(
+        (yield* LLMClient.prepare<AnthropicMessages.AnthropicMessagesBody>(
          LLM.request({
            model: opus48,
            messages: [Message.assistant("Plain."), Message.system("After plain assistant.")],
@@ -197,11 +196,12 @@ describe("Anthropic Messages route", () => {
        },
      ])
      expect(
-        (yield* compileRequest(LLM.request({ model: opus48, messages: [Message.system("First.")], cache: "none" })))
-          .body.messages,
+        (yield* LLMClient.prepare<AnthropicMessages.AnthropicMessagesBody>(
+          LLM.request({ model: opus48, messages: [Message.system("First.")], cache: "none" }),
+        )).body.messages,
      ).toEqual([{ role: "user", content: [{ type: "text", text: "<system-update>\nFirst.\n</system-update>" }] }])
      expect(
-        (yield* compileRequest(
+        (yield* LLMClient.prepare<AnthropicMessages.AnthropicMessagesBody>(
          LLM.request({
            model: opus48,
            messages: [Message.user("Before."), Message.system("One."), Message.system("Two.")],
@@ -223,7 +223,7 @@ describe("Anthropic Messages route", () => {

  it.effect("rejects a system update between a local tool call and its result", () =>
    Effect.gen(function* () {
-      const error = yield* compileRequest(
+      const error = yield* LLMClient.prepare(
        LLM.request({
          model: opus48,
          messages: [
@@ -242,7 +242,7 @@ describe("Anthropic Messages route", () => {

  it.effect("prepares tool call and tool result messages", () =>
    Effect.gen(function* () {
-      const prepared = yield* compileRequest(
+      const prepared = yield* LLMClient.prepare<AnthropicMessages.AnthropicMessagesBody>(
        LLM.request({
          id: "req_tool_result",
          model,
@@ -273,7 +273,7 @@ describe("Anthropic Messages route", () => {

  it.effect("keeps tools and sends tool_choice none", () =>
    Effect.gen(function* () {
-      const prepared = yield* compileRequest(
+      const prepared = yield* LLMClient.prepare<AnthropicMessages.AnthropicMessagesBody>(
        LLM.request({
          id: "req_tool_choice_none",
          model,
@@ -303,7 +303,7 @@ describe("Anthropic Messages route", () => {
  // not JSON-stringified into `tool_result.content`.
  it.effect("lowers media tool-result content as structured blocks", () =>
    Effect.gen(function* () {
-      const prepared = yield* compileRequest(
+      const prepared = yield* LLMClient.prepare<AnthropicMessages.AnthropicMessagesBody>(
        LLM.request({
          id: "req_tool_result_image",
          model,
@@ -335,7 +335,7 @@ describe("Anthropic Messages route", () => {

  it.effect("lowers single-image tool-result content as a structured image block", () =>
    Effect.gen(function* () {
-      const prepared = yield* compileRequest(
+      const prepared = yield* LLMClient.prepare<AnthropicMessages.AnthropicMessagesBody>(
        LLM.request({
          id: "req_tool_result_image_only",
          model,
@@ -360,7 +360,7 @@ describe("Anthropic Messages route", () => {

  it.effect("rejects unsupported media in tool-result content with a clear error", () =>
    Effect.gen(function* () {
-      const error = yield* compileRequest(
+      const error = yield* LLMClient.prepare(
        LLM.request({
          id: "req_tool_result_unsupported_media",
          model,
@@ -384,7 +384,7 @@ describe("Anthropic Messages route", () => {

  it.effect("prepares the composed native continuation request", () =>
    Effect.gen(function* () {
-      const prepared = yield* compileRequest(
+      const prepared = yield* LLMClient.prepare<AnthropicMessages.AnthropicMessagesBody>(
        continuationRequest({
          id: "req_native_continuation_anthropic",
          model,
@@ -428,7 +428,7 @@ describe("Anthropic Messages route", () => {

  it.effect("lowers preserved Anthropic reasoning signature metadata", () =>
    Effect.gen(function* () {
-      const prepared = yield* compileRequest(
+      const prepared = yield* LLMClient.prepare(
        LLM.request({
          model,
          messages: [
@@ -447,7 +447,7 @@ describe("Anthropic Messages route", () => {

  it.effect("round-trips redacted thinking as redacted_thinking blocks", () =>
    Effect.gen(function* () {
-      const prepared = yield* compileRequest(
+      const prepared = yield* LLMClient.prepare(
        LLM.request({
          model,
          messages: [
@@ -628,7 +628,9 @@ describe("Anthropic Messages route", () => {
        { type: "reasoning", text: "", providerMetadata: { anthropic: { signature: "sig_1" } } },
      ])

-      const prepared = yield* compileRequest(LLM.request({ model, messages: [response.message], cache: "none" }))
+      const prepared = yield* LLMClient.prepare<AnthropicMessages.AnthropicMessagesBody>(
+        LLM.request({ model, messages: [response.message], cache: "none" }),
+      )
      expect(prepared.body.messages).toEqual([
        { role: "assistant", content: [{ type: "thinking", thinking: "", signature: "sig_1" }] },
      ])
@@ -684,7 +686,9 @@ describe("Anthropic Messages route", () => {
        ),
      )

-      expect(response.toolCalls).toMatchObject([{ id: "call_1", name: "lookup", input: { query: "weather" } }])
+      expect(response.toolCalls).toMatchObject([
+        { id: "call_1", name: "lookup", input: { query: "weather" } },
+      ])
    }),
  )

@@ -769,7 +773,7 @@ describe("Anthropic Messages route", () => {
          ),
        ),
      )
-      const prepared = yield* compileRequest(
+      const prepared = yield* LLMClient.prepare<AnthropicMessages.AnthropicMessagesBody>(
        LLM.request({
          model,
          messages: [
@@ -933,7 +937,7 @@ describe("Anthropic Messages route", () => {

      const error = yield* LLMClient.generate(request).pipe(Effect.provide(fixedResponse(body)), Effect.flip)

-      expect(error).toBeInstanceOf(AIError)
+      expect(error).toBeInstanceOf(LLMError)
      expect(error.message).toContain("Invalid JSON input for anthropic-messages tool call web_search")
    }),
  )
@@ -1007,7 +1011,7 @@ describe("Anthropic Messages route", () => {
        Effect.flip,
      )

-      expect(error).toBeInstanceOf(AIError)
+      expect(error).toBeInstanceOf(LLMError)
      expect(error.reason).toMatchObject({ _tag: "InvalidRequest" })
      expect(error.message).toContain("HTTP 400")
    }),
@@ -1129,7 +1133,7 @@ describe("Anthropic Messages route", () => {

  it.effect("round-trips provider-executed assistant content into server tool blocks", () =>
    Effect.gen(function* () {
-      const prepared = yield* compileRequest(
+      const prepared = yield* LLMClient.prepare(
        LLM.request({
          id: "req_round_trip",
          model,
@@ -1180,7 +1184,7 @@ describe("Anthropic Messages route", () => {

  it.effect("rejects round-trip for unknown server tool names", () =>
    Effect.gen(function* () {
-      const error = yield* compileRequest(
+      const error = yield* LLMClient.prepare(
        LLM.request({
          id: "req_unknown_server_tool",
          model,
@@ -1257,7 +1261,7 @@ describe("Anthropic Messages route", () => {

  it.effect("maps ttlSeconds >= 3600 to cache_control ttl: '1h'", () =>
    Effect.gen(function* () {
-      const prepared = yield* compileRequest(
+      const prepared = yield* LLMClient.prepare(
        LLM.request({
          model,
          system: { type: "text", text: "system", cache: new CacheHint({ type: "ephemeral", ttlSeconds: 3600 }) },
@@ -1273,7 +1277,7 @@ describe("Anthropic Messages route", () => {

  it.effect("emits cache_control on tool definitions and tool-result blocks", () =>
    Effect.gen(function* () {
-      const prepared = yield* compileRequest(
+      const prepared = yield* LLMClient.prepare(
        LLM.request({
          model,
          tools: [
@@ -1314,7 +1318,7 @@ describe("Anthropic Messages route", () => {
  it.effect("drops cache_control breakpoints past the 4-per-request cap", () =>
    Effect.gen(function* () {
      const hint = new CacheHint({ type: "ephemeral" })
-      const prepared = yield* compileRequest(
+      const prepared = yield* LLMClient.prepare(
        LLM.request({
          model,
          system: [
@@ -1340,7 +1344,7 @@ describe("Anthropic Messages route", () => {
  it.effect("spends breakpoint budget on tools before system before messages", () =>
    Effect.gen(function* () {
      const hint = new CacheHint({ type: "ephemeral" })
-      const prepared = yield* compileRequest(
+      const prepared = yield* LLMClient.prepare(
        LLM.request({
          model,
          tools: [
@@ -13,7 +13,6 @@ import {
  ToolDefinition,
 } from "../../src"
 import { LLMClient } from "../../src/route"
-import { compileRequest } from "../../src/route/client"
 import { AmazonBedrock } from "../../src/providers"
 import * as BedrockConverse from "../../src/protocols/bedrock-converse"
 import { it } from "../lib/effect"
@@ -102,7 +101,7 @@ const baseRequest = LLM.request({
 describe("Bedrock Converse route", () => {
  it.effect("prepares Converse target with system, inference config, and messages", () =>
    Effect.gen(function* () {
-      const prepared = yield* compileRequest(baseRequest)
+      const prepared = yield* LLMClient.prepare(baseRequest)

      expect(prepared.body).toEqual({
        modelId: "anthropic.claude-3-5-sonnet-20240620-v1:0",
@@ -115,7 +114,7 @@ describe("Bedrock Converse route", () => {

  it.effect("passes topK through additionalModelRequestFields as top_k", () =>
    Effect.gen(function* () {
-      const prepared = yield* compileRequest(
+      const prepared = yield* LLMClient.prepare<BedrockConverse.BedrockConverseBody>(
        LLMRequest.update(baseRequest, {
          generation: GenerationOptions.make({ maxTokens: 64, temperature: 0, topK: 40 }),
        }),
@@ -130,14 +129,14 @@ describe("Bedrock Converse route", () => {

  it.effect("omits additionalModelRequestFields when topK is unset", () =>
    Effect.gen(function* () {
-      const prepared = yield* compileRequest(baseRequest)
+      const prepared = yield* LLMClient.prepare<BedrockConverse.BedrockConverseBody>(baseRequest)
      expect(prepared.body.additionalModelRequestFields).toBeUndefined()
    }),
  )

  it.effect("lowers chronological system updates to wrapped user text in order", () =>
    Effect.gen(function* () {
-      const prepared = yield* compileRequest(
+      const prepared = yield* LLMClient.prepare<BedrockConverse.BedrockConverseBody>(
        LLM.request({
          model,
          messages: [Message.user("Before."), Message.system("Update."), Message.assistant("After.")],
@@ -154,7 +153,7 @@ describe("Bedrock Converse route", () => {

  it.effect("prepares tool config with toolSpec and toolChoice", () =>
    Effect.gen(function* () {
-      const prepared = yield* compileRequest(
+      const prepared = yield* LLMClient.prepare(
        LLMRequest.update(baseRequest, {
          tools: [
            ToolDefinition.make({
@@ -188,7 +187,7 @@ describe("Bedrock Converse route", () => {

  it.effect("keeps tools and omits the unsupported choice when tool choice is none", () =>
    Effect.gen(function* () {
-      const prepared = yield* compileRequest(
+      const prepared = yield* LLMClient.prepare(
        LLMRequest.update(baseRequest, {
          tools: [
            ToolDefinition.make({
@@ -218,7 +217,7 @@ describe("Bedrock Converse route", () => {

  it.effect("lowers assistant tool-call + tool-result message history", () =>
    Effect.gen(function* () {
-      const prepared = yield* compileRequest(
+      const prepared = yield* LLMClient.prepare(
        LLM.request({
          id: "req_history",
          model,
@@ -257,7 +256,7 @@ describe("Bedrock Converse route", () => {

  it.effect("lowers image content in tool-result messages", () =>
    Effect.gen(function* () {
-      const prepared = yield* compileRequest(
+      const prepared = yield* LLMClient.prepare(
        LLM.request({
          id: "req_tool_image",
          model,
@@ -492,7 +491,7 @@ describe("Bedrock Converse route", () => {
        providerMetadata: { bedrock: { signature: "sig_1" } },
      })

-      const prepared = yield* compileRequest(
+      const prepared = yield* LLMClient.prepare<BedrockConverse.BedrockConverseBody>(
        LLM.request({
          model,
          messages: [
@@ -547,7 +546,9 @@ describe("Bedrock Converse route", () => {
        },
      ])

-      const prepared = yield* compileRequest(LLM.request({ model, messages: [response.message], cache: "none" }))
+      const prepared = yield* LLMClient.prepare<BedrockConverse.BedrockConverseBody>(
+        LLM.request({ model, messages: [response.message], cache: "none" }),
+      )
      expect(prepared.body.messages).toEqual([
        {
          role: "assistant",
@@ -638,7 +639,7 @@ describe("Bedrock Converse route", () => {
        text: "",
        providerMetadata: { bedrock: { redactedData } },
      })
-      const prepared = yield* compileRequest(
+      const prepared = yield* LLMClient.prepare<BedrockConverse.BedrockConverseBody>(
        LLM.request({
          model,
          messages: [
@@ -753,7 +754,7 @@ describe("Bedrock Converse route", () => {
          secretAccessKey: "wJalrXUtnFEMI/K7MDENG/bPxRfiCYEXAMPLEKEY",
        },
      }).model("anthropic.claude-3-5-sonnet-20240620-v1:0")
-      const prepared = yield* compileRequest(LLMRequest.update(baseRequest, { model: signed }))
+      const prepared = yield* LLMClient.prepare(LLMRequest.update(baseRequest, { model: signed }))

      expect(prepared.route).toBe("bedrock-converse")
      expect(prepared.model).toBe(signed)
@@ -763,7 +764,7 @@ describe("Bedrock Converse route", () => {
  it.effect("emits cachePoint markers after system, user-text, and assistant-text with cache hints", () =>
    Effect.gen(function* () {
      const cache = new CacheHint({ type: "ephemeral" })
-      const prepared = yield* compileRequest(
+      const prepared = yield* LLMClient.prepare(
        LLM.request({
          id: "req_cache",
          model,
@@ -795,7 +796,7 @@ describe("Bedrock Converse route", () => {

  it.effect("does not emit cachePoint when no cache hint is set", () =>
    Effect.gen(function* () {
-      const prepared = yield* compileRequest(baseRequest)
+      const prepared = yield* LLMClient.prepare(baseRequest)
      expect(prepared.body).toMatchObject({
        system: [{ text: "You are concise." }],
        messages: [{ role: "user", content: [{ text: "Say hello." }] }],
@@ -805,7 +806,7 @@ describe("Bedrock Converse route", () => {

  it.effect("lowers image media into Bedrock image blocks", () =>
    Effect.gen(function* () {
-      const prepared = yield* compileRequest(
+      const prepared = yield* LLMClient.prepare(
        LLM.request({
          id: "req_image",
          model,
@@ -842,7 +843,7 @@ describe("Bedrock Converse route", () => {

  it.effect("base64-encodes Uint8Array image bytes", () =>
    Effect.gen(function* () {
-      const prepared = yield* compileRequest(
+      const prepared = yield* LLMClient.prepare(
        LLM.request({
          id: "req_image_bytes",
          model,
@@ -864,7 +865,7 @@ describe("Bedrock Converse route", () => {

  it.effect("lowers document media into Bedrock document blocks with format and name", () =>
    Effect.gen(function* () {
-      const prepared = yield* compileRequest(
+      const prepared = yield* LLMClient.prepare(
        LLM.request({
          id: "req_doc",
          model,
@@ -896,7 +897,7 @@ describe("Bedrock Converse route", () => {

  it.effect("requires names for document media", () =>
    Effect.gen(function* () {
-      const error = yield* compileRequest(
+      const error = yield* LLMClient.prepare(
        LLM.request({
          model,
          messages: [Message.user({ type: "media", mediaType: "application/pdf", data: "UERGREFUQQ==" })],
@@ -909,7 +910,7 @@ describe("Bedrock Converse route", () => {

  it.effect("passes named document-only messages through for provider validation", () =>
    Effect.gen(function* () {
-      const prepared = yield* compileRequest(
+      const prepared = yield* LLMClient.prepare<BedrockConverse.BedrockConverseBody>(
        LLM.request({
          model,
          cache: "none",
@@ -935,7 +936,7 @@ describe("Bedrock Converse route", () => {

  it.effect("lowers document media in tool results", () =>
    Effect.gen(function* () {
-      const prepared = yield* compileRequest(
+      const prepared = yield* LLMClient.prepare<BedrockConverse.BedrockConverseBody>(
        LLM.request({
          model,
          cache: "none",
@@ -987,7 +988,7 @@ describe("Bedrock Converse route", () => {

  it.effect("rejects unsupported image media types", () =>
    Effect.gen(function* () {
-      const error = yield* compileRequest(
+      const error = yield* LLMClient.prepare(
        LLM.request({
          id: "req_bad_image",
          model,
@@ -1001,7 +1002,7 @@ describe("Bedrock Converse route", () => {

  it.effect("rejects unsupported document media types", () =>
    Effect.gen(function* () {
-      const error = yield* compileRequest(
+      const error = yield* LLMClient.prepare(
        LLM.request({
          id: "req_bad_doc",
          model,
@@ -1016,7 +1017,7 @@ describe("Bedrock Converse route", () => {
  it.effect("maps ttlSeconds >= 3600 to cachePoint ttl: '1h'", () =>
    Effect.gen(function* () {
      const cache = new CacheHint({ type: "ephemeral", ttlSeconds: 3600 })
-      const prepared = yield* compileRequest(
+      const prepared = yield* LLMClient.prepare(
        LLM.request({
          model,
          system: [{ type: "text", text: "system", cache }],
@@ -1033,7 +1034,7 @@ describe("Bedrock Converse route", () => {
  it.effect("appends cachePoint after marked tool definitions and tool-result blocks", () =>
    Effect.gen(function* () {
      const cache = new CacheHint({ type: "ephemeral" })
-      const prepared = yield* compileRequest(
+      const prepared = yield* LLMClient.prepare(
        LLM.request({
          model,
          tools: [{ name: "lookup", description: "lookup", inputSchema: { type: "object", properties: {} }, cache }],
@@ -1065,7 +1066,7 @@ describe("Bedrock Converse route", () => {
  it.effect("drops cachePoint markers past the 4-per-request cap", () =>
    Effect.gen(function* () {
      const cache = new CacheHint({ type: "ephemeral" })
-      const prepared = yield* compileRequest(
+      const prepared = yield* LLMClient.prepare(
        LLM.request({
          model,
          system: [
@@ -3,7 +3,7 @@ import { ConfigProvider, Effect, Schema } from "effect"
 import { HttpClientRequest } from "effect/unstable/http"
 import { LLM, LLMEvent } from "../../src"
 import { CloudflareAIGateway, CloudflareWorkersAI } from "../../src/providers/cloudflare"
-import { compileRequest } from "../../src/route/client"
+import { LLMClient } from "../../src/route"
 import { it } from "../lib/effect"
 import { dynamicResponse } from "../lib/http"
 import { sseEvents } from "../lib/sse"
@@ -34,7 +34,7 @@ describe("Cloudflare", () => {
      })
      expect(model.route.endpoint.baseURL).toBe("https://gateway.ai.cloudflare.com/v1/test-account/test-gateway/compat")

-      const prepared = yield* compileRequest(LLM.request({ model, prompt: "Say hello." }))
+      const prepared = yield* LLMClient.prepare(LLM.request({ model, prompt: "Say hello." }))

      expect(prepared.route).toBe("cloudflare-ai-gateway")
      expect(prepared.body).toMatchObject({
@@ -129,7 +129,7 @@ describe("Cloudflare", () => {
        openai: { reasoningField: "reasoning", reasoningDetails: merged },
      })

-      const replay = yield* compileRequest(LLM.request({ model, messages: [response.message] }))
+      const replay = yield* LLMClient.prepare(LLM.request({ model, messages: [response.message] }))
      expect(replay.body.messages).toEqual([
        { role: "assistant", content: "Hello", reasoning: "Thinking", reasoning_details: merged },
      ])
@@ -180,7 +180,7 @@ describe("Cloudflare", () => {

  it.effect("allows a fully configured baseURL override", () =>
    Effect.gen(function* () {
-      const prepared = yield* compileRequest(
+      const prepared = yield* LLMClient.prepare(
        LLM.request({
          model: CloudflareAIGateway.configure({
            baseURL: "https://gateway.proxy.test/v1/custom/compat",
@@ -208,7 +208,7 @@ describe("Cloudflare", () => {
      })
      expect(model.route.endpoint.baseURL).toBe("https://api.cloudflare.com/client/v4/accounts/test-account/ai/v1")

-      const prepared = yield* compileRequest(LLM.request({ model, prompt: "Say hello." }))
+      const prepared = yield* LLMClient.prepare(LLM.request({ model, prompt: "Say hello." }))

      expect(prepared.route).toBe("cloudflare-workers-ai")
      expect(prepared.body).toMatchObject({
@@ -1,8 +1,7 @@
 import { describe, expect } from "bun:test"
 import { Effect } from "effect"
-import { LLM, AIError, LLMRequest, Message, ToolCallPart, ToolDefinition, Usage } from "../../src"
+import { LLM, LLMError, LLMRequest, Message, ToolCallPart, ToolDefinition, Usage } from "../../src"
 import { Auth, LLMClient } from "../../src/route"
-import { compileRequest } from "../../src/route/client"
 import * as Gemini from "../../src/protocols/gemini"
 import { ProviderShared } from "../../src/protocols/shared"
 import { it } from "../lib/effect"
@@ -27,7 +26,7 @@ const request = LLM.request({
 describe("Gemini route", () => {
  it.effect("prepares Gemini target", () =>
    Effect.gen(function* () {
-      const prepared = yield* compileRequest(request)
+      const prepared = yield* LLMClient.prepare(request)

      expect(prepared.body).toEqual({
        contents: [{ role: "user", parts: [{ text: "Say hello." }] }],
@@ -39,56 +38,28 @@ describe("Gemini route", () => {

  it.effect("normalizes Gemini thinking options", () =>
    Effect.gen(function* () {
-      const prepared = yield* compileRequest(
+      const prepared = yield* LLMClient.prepare<Gemini.GeminiBody>(
        LLMRequest.update(request, {
-          providerOptions: {
-            gemini: {
-              cachedContent: "cachedContents/example",
-              safetySettings: [{ category: "HARM_CATEGORY_HATE_SPEECH", threshold: "BLOCK_ONLY_HIGH" }],
-              serviceTier: "priority",
-              thinkingConfig: { thinkingBudget: 0, includeThoughts: false, thinkingLevel: "high" },
-            },
-          },
+          providerOptions: { gemini: { thinkingConfig: { thinkingBudget: 0, includeThoughts: false } } },
        }),
      )
-      const filtered = yield* compileRequest(
+      const filtered = yield* LLMClient.prepare<Gemini.GeminiBody>(
        LLMRequest.update(request, {
          providerOptions: { gemini: { thinkingConfig: { thinkingBudget: "invalid", includeThoughts: false } } },
        }),
      )
-      const defaulted = yield* compileRequest(
-        LLMRequest.update(request, {
-          providerOptions: { gemini: { thinkingConfig: { thinkingLevel: "high" } } },
-        }),
-      )
-      const emptySafetySettings = yield* compileRequest(
-        LLMRequest.update(request, {
-          providerOptions: { gemini: { safetySettings: [] } },
-        }),
-      )

      expect(prepared.body.generationConfig?.thinkingConfig).toEqual({
        thinkingBudget: 0,
        includeThoughts: false,
-        thinkingLevel: "high",
      })
-      expect(prepared.body.cachedContent).toBe("cachedContents/example")
-      expect(prepared.body.safetySettings).toEqual([
-        { category: "HARM_CATEGORY_HATE_SPEECH", threshold: "BLOCK_ONLY_HIGH" },
-      ])
-      expect(prepared.body.serviceTier).toBe("priority")
      expect(filtered.body.generationConfig?.thinkingConfig).toEqual({ includeThoughts: false })
-      expect(defaulted.body.generationConfig?.thinkingConfig).toEqual({
-        includeThoughts: true,
-        thinkingLevel: "high",
-      })
-      expect(emptySafetySettings.body.safetySettings).toEqual([])
    }),
  )

  it.effect("lowers chronological system updates to wrapped user text in order", () =>
    Effect.gen(function* () {
-      const prepared = yield* compileRequest(
+      const prepared = yield* LLMClient.prepare<Gemini.GeminiBody>(
        LLM.request({
          model,
          messages: [Message.user("Before."), Message.system("Update."), Message.assistant("After.")],
@@ -104,7 +75,7 @@ describe("Gemini route", () => {

  it.effect("prepares multimodal user input and tool history", () =>
    Effect.gen(function* () {
-      const prepared = yield* compileRequest(
+      const prepared = yield* LLMClient.prepare(
        LLM.request({
          id: "req_tool_result",
          model,
@@ -172,7 +143,7 @@ describe("Gemini route", () => {

  it.effect("continues media tool results as inline model input without base64 text", () =>
    Effect.gen(function* () {
-      const prepared = yield* compileRequest(
+      const prepared = yield* LLMClient.prepare<Gemini.GeminiBody>(
        LLM.request({
          model,
          messages: [
@@ -217,7 +188,7 @@ describe("Gemini route", () => {

  it.effect("strips matching data URLs to raw base64 inlineData", () =>
    Effect.gen(function* () {
-      const prepared = yield* compileRequest(
+      const prepared = yield* LLMClient.prepare<Gemini.GeminiBody>(
        LLM.request({
          model,
          messages: [
@@ -258,7 +229,7 @@ describe("Gemini route", () => {
  ] as const)
    it.effect(`rejects ${name}`, () =>
      Effect.gen(function* () {
-        const error = yield* compileRequest(
+        const error = yield* LLMClient.prepare(
          LLM.request({ model, messages: [Message.user({ type: "media", ...media })] }),
        ).pipe(Effect.flip)
        expect(error.message).toMatch(/does not support|does not match|valid base64/)
@@ -267,7 +238,7 @@ describe("Gemini route", () => {

  it.effect("rejects oversized image input", () =>
    Effect.gen(function* () {
-      const error = yield* compileRequest(
+      const error = yield* LLMClient.prepare(
        LLM.request({
          model,
          messages: [
@@ -285,7 +256,7 @@ describe("Gemini route", () => {

  it.effect("keeps tools and sends function calling mode NONE", () =>
    Effect.gen(function* () {
-      const prepared = yield* compileRequest(
+      const prepared = yield* LLMClient.prepare(
        LLM.request({
          id: "req_tool_choice_none",
          model,
@@ -305,7 +276,7 @@ describe("Gemini route", () => {

  it.effect("sanitizes integer enums, dangling required, untyped arrays, and scalar object keys", () =>
    Effect.gen(function* () {
-      const prepared = yield* compileRequest(
+      const prepared = yield* LLMClient.prepare(
        LLM.request({
          id: "req_schema_patch",
          model,
@@ -486,7 +457,7 @@ describe("Gemini route", () => {
        response.events.findIndex((event) => event.type === "tool-call"),
      )

-      const prepared = yield* compileRequest(
+      const prepared = yield* LLMClient.prepare<Gemini.GeminiBody>(
        LLM.request({
          model,
          messages: [
@@ -712,7 +683,7 @@ describe("Gemini route", () => {
        Effect.flip,
      )

-      expect(error).toBeInstanceOf(AIError)
+      expect(error).toBeInstanceOf(LLMError)
      expect(error.reason).toMatchObject({ _tag: "InvalidProviderOutput" })
      expect(error.message).toContain("Invalid google/gemini stream event")
    }),
@@ -720,7 +691,7 @@ describe("Gemini route", () => {

  it.effect("rejects unsupported assistant media content", () =>
    Effect.gen(function* () {
-      const error = yield* compileRequest(
+      const error = yield* LLMClient.prepare(
        LLM.request({
          id: "req_media",
          model,
@@ -4,7 +4,6 @@ import { HttpClientRequest } from "effect/unstable/http"
 import { LLM } from "../../src"
 import { GoogleVertex, GoogleVertexChat, GoogleVertexMessages, GoogleVertexResponses } from "../../src/providers"
 import { LLMClient } from "../../src/route"
-import { compileRequest } from "../../src/route/client"
 import { it } from "../lib/effect"
 import { dynamicResponse } from "../lib/http"
 import { deltaChunk, finishChunk } from "../lib/openai-chunks"
@@ -181,6 +180,23 @@ describe("Google Vertex providers", () => {
    }),
  )

+  it.effect("protects the Vertex Messages API version from body overlays", () =>
+    Effect.gen(function* () {
+      const error = yield* LLMClient.prepare(
+        LLM.request({
+          model: GoogleVertexMessages.configure({
+            accessToken: "vertex-token",
+            http: { body: { anthropic_version: "wrong" } },
+            project: "vertex-project",
+          }).model("claude-sonnet-4-6"),
+          prompt: "Say hello.",
+        }),
+      ).pipe(Effect.flip)
+
+      expect(error.message).toContain("http.body cannot overlay protocol-owned field(s): anthropic_version")
+    }),
+  )
+
  it.effect("routes tuned Gemini models through their deployed endpoint", () =>
    Effect.gen(function* () {
      const response = yield* LLMClient.generate(
@@ -1,18 +1,17 @@
 import { describe, expect } from "bun:test"
 import { Effect } from "effect"
-import { LLM, LLMEvent, LLMResponse, LanguageModel } from "../../src"
+import { LLM, LLMEvent, LLMResponse, Model } from "../../src"
 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"

 const cases = [
  {
    name: "OpenRouter",
-    model: LanguageModel.update(
+    model: Model.update(
      OpenRouter.configure({
        apiKey: process.env.OPENROUTER_API_KEY ?? "fixture",
        providerOptions: { openrouter: { reasoning: { max_tokens: 1024 } } },
@@ -25,7 +24,7 @@ const cases = [
  },
  {
    name: "Vercel AI Gateway",
-    model: LanguageModel.update(
+    model: Model.update(
      OpenAICompatible.configure({
        provider: "vercel-ai-gateway",
        baseURL: "https://ai-gateway.vercel.sh/v1",
@@ -85,7 +84,9 @@ for (const item of cases) {
            ),
          ).toBe(true)

-          const replay = yield* compileRequest(LLM.request({ model: item.model, messages: [response.message] }))
+          const replay = yield* LLMClient.prepare<OpenAIChat.OpenAIChatBody>(
+            LLM.request({ model: item.model, messages: [response.message] }),
+          )
          expect(replay.body.messages).toMatchObject([
            { role: "assistant", content: response.text, reasoning: response.reasoning },
          ])
@@ -4,11 +4,11 @@ import { HttpClientRequest } from "effect/unstable/http"
 import {
  HttpOptions,
  LLM,
-  AIError,
+  LLMError,
  LLMEvent,
  LLMRequest,
  Message,
-  LanguageModel,
+  Model,
  ToolCallPart,
  ToolDefinition,
  Usage,
@@ -18,7 +18,6 @@ import * as OpenAI from "../../src/providers/openai"
 import * as OpenAIChat from "../../src/protocols/openai-chat"
 import { ProviderShared } from "../../src/protocols/shared"
 import { Auth, LLMClient } from "../../src/route"
-import { compileRequest } from "../../src/route/client"
 import { it } from "../lib/effect"
 import { dynamicResponse, fixedResponse, truncatedStream } from "../lib/http"
 import { deltaChunk, usageChunk } from "../lib/openai-chunks"
@@ -43,7 +42,11 @@ const request = LLM.request({
 describe("OpenAI Chat route", () => {
  it.effect("prepares OpenAI Chat payload", () =>
    Effect.gen(function* () {
-      const prepared = yield* compileRequest(request)
+      // Pass the OpenAIChat payload type so `prepared.body` is statically
+      // typed to the route's native shape — the assertions below read field
+      // names without `unknown` casts.
+      const prepared = yield* LLMClient.prepare<OpenAIChat.OpenAIChatBody>(request)
+      const _typed: { readonly model: string; readonly stream: true } = prepared.body

      expect(prepared.body).toEqual({
        model: "gpt-4o-mini",
@@ -61,7 +64,7 @@ describe("OpenAI Chat route", () => {

  it.effect("lowers chronological system updates to escaped user wrappers in order", () =>
    Effect.gen(function* () {
-      const prepared = yield* compileRequest(
+      const prepared = yield* LLMClient.prepare<OpenAIChat.OpenAIChatBody>(
        LLM.request({
          model,
          messages: [
@@ -84,7 +87,7 @@ describe("OpenAI Chat route", () => {

  it.effect("replays canonical reasoning as OpenAI-compatible reasoning_content", () =>
    Effect.gen(function* () {
-      const prepared = yield* compileRequest(
+      const prepared = yield* LLMClient.prepare<OpenAIChat.OpenAIChatBody>(
        LLM.request({
          model,
          messages: [
@@ -102,9 +105,9 @@ describe("OpenAI Chat route", () => {

  it.effect("writes reasoning to a configured custom field on every assistant message", () =>
    Effect.gen(function* () {
-      const prepared = yield* compileRequest(
+      const prepared = yield* LLMClient.prepare<OpenAIChat.OpenAIChatBody>(
        LLM.request({
-          model: LanguageModel.update(model, { compatibility: { reasoningField: "vendor_reasoning" } }),
+          model: Model.update(model, { compatibility: { reasoningField: "vendor_reasoning" } }),
          messages: [
            Message.assistant([
              {
@@ -128,9 +131,9 @@ describe("OpenAI Chat route", () => {

  it.effect("rejects reasoning fields that conflict with assistant message fields", () =>
    Effect.gen(function* () {
-      const error = yield* compileRequest(
+      const error = yield* LLMClient.prepare(
        LLM.request({
-          model: LanguageModel.update(model, { compatibility: { reasoningField: "content" } }),
+          model: Model.update(model, { compatibility: { reasoningField: "content" } }),
          messages: [Message.assistant([{ type: "reasoning", text: "thinking" }])],
        }),
      ).pipe(Effect.flip)
@@ -141,7 +144,7 @@ describe("OpenAI Chat route", () => {

  it.effect("maps OpenAI provider options to Chat options", () =>
    Effect.gen(function* () {
-      const prepared = yield* compileRequest(
+      const prepared = yield* LLMClient.prepare<OpenAIChat.OpenAIChatBody>(
        LLM.request({
          model: OpenAI.configure({ baseURL: "https://api.openai.test/v1/", apiKey: "test" }).chat("gpt-4o-mini"),
          prompt: "think",
@@ -156,7 +159,7 @@ describe("OpenAI Chat route", () => {

  it.effect("passes through custom OpenAI-compatible reasoning effort strings", () =>
    Effect.gen(function* () {
-      const prepared = yield* compileRequest(
+      const prepared = yield* LLMClient.prepare<OpenAIChat.OpenAIChatBody>(
        LLM.request({
          model,
          prompt: "think",
@@ -171,9 +174,7 @@ describe("OpenAI Chat route", () => {
  it.effect("adds native query params to the Chat Completions URL", () =>
    LLMClient.generate(
      LLMRequest.update(request, {
-        model: LanguageModel.update(model, {
-          route: model.route.with({ endpoint: { query: { "api-version": "v1" } } }),
-        }),
+        model: Model.update(model, { route: model.route.with({ endpoint: { query: { "api-version": "v1" } } }) }),
      }),
    ).pipe(
      Effect.provide(
@@ -252,7 +253,7 @@ describe("OpenAI Chat route", () => {

  it.effect("prepares assistant tool-call and tool-result messages", () =>
    Effect.gen(function* () {
-      const prepared = yield* compileRequest(
+      const prepared = yield* LLMClient.prepare(
        LLM.request({
          id: "req_tool_result",
          model,
@@ -290,7 +291,7 @@ describe("OpenAI Chat route", () => {
  it.effect("preserves structured tool errors for the model", () =>
    Effect.gen(function* () {
      const error = { error: { type: "unknown", message: "Tool execution interrupted" } }
-      const prepared = yield* compileRequest(
+      const prepared = yield* LLMClient.prepare<OpenAIChat.OpenAIChatBody>(
        LLM.request({
          model,
          messages: [
@@ -310,7 +311,7 @@ describe("OpenAI Chat route", () => {

  it.effect("continues image tool results as vision input without base64 text", () =>
    Effect.gen(function* () {
-      const prepared = yield* compileRequest(
+      const prepared = yield* LLMClient.prepare<OpenAIChat.OpenAIChatBody>(
        LLM.request({
          model,
          messages: [
@@ -354,7 +355,7 @@ describe("OpenAI Chat route", () => {

  it.effect("orders parallel tool responses before one aggregated vision message", () =>
    Effect.gen(function* () {
-      const prepared = yield* compileRequest(
+      const prepared = yield* LLMClient.prepare<OpenAIChat.OpenAIChatBody>(
        LLM.request({
          model,
          messages: [
@@ -404,7 +405,7 @@ describe("OpenAI Chat route", () => {

  it.effect("aggregates consecutive tool images with a following system update", () =>
    Effect.gen(function* () {
-      const prepared = yield* compileRequest(
+      const prepared = yield* LLMClient.prepare<OpenAIChat.OpenAIChatBody>(
        LLM.request({
          model,
          messages: [
@@ -445,7 +446,7 @@ describe("OpenAI Chat route", () => {

  it.effect("appends system updates without replacing multipart user content", () =>
    Effect.gen(function* () {
-      const prepared = yield* compileRequest(
+      const prepared = yield* LLMClient.prepare<OpenAIChat.OpenAIChatBody>(
        LLM.request({
          model,
          messages: [
@@ -473,7 +474,7 @@ describe("OpenAI Chat route", () => {
  ] as const)
    it.effect(`rejects ${name}`, () =>
      Effect.gen(function* () {
-        const error = yield* compileRequest(
+        const error = yield* LLMClient.prepare(
          LLM.request({ model, messages: [Message.user({ type: "media", ...media })] }),
        ).pipe(Effect.flip)
        expect(error.message).toMatch(/does not support|does not match|valid base64/)
@@ -482,7 +483,7 @@ describe("OpenAI Chat route", () => {

  it.effect("rejects oversized image input", () =>
    Effect.gen(function* () {
-      const error = yield* compileRequest(
+      const error = yield* LLMClient.prepare(
        LLM.request({
          model,
          messages: [
@@ -500,7 +501,7 @@ describe("OpenAI Chat route", () => {

  it.effect("prepares raw and data URL image media as vision input", () =>
    Effect.gen(function* () {
-      const prepared = yield* compileRequest(
+      const prepared = yield* LLMClient.prepare<OpenAIChat.OpenAIChatBody>(
        LLM.request({
          id: "req_media",
          model,
@@ -527,7 +528,7 @@ describe("OpenAI Chat route", () => {

  it.effect("lowers reasoning-only assistant history", () =>
    Effect.gen(function* () {
-      const prepared = yield* compileRequest(
+      const prepared = yield* LLMClient.prepare<OpenAIChat.OpenAIChatBody>(
        LLM.request({
          id: "req_reasoning",
          model,
@@ -618,7 +619,9 @@ describe("OpenAI Chat route", () => {
          openai: { reasoningField: field },
        })

-        const replay = yield* compileRequest(LLM.request({ model, messages: [response.message] }))
+        const replay = yield* LLMClient.prepare<OpenAIChat.OpenAIChatBody>(
+          LLM.request({ model, messages: [response.message] }),
+        )
        expect(replay.body.messages).toEqual([{ role: "assistant", content: "Hello", [field]: "thinking" }])
      }
    }),
@@ -626,7 +629,7 @@ describe("OpenAI Chat route", () => {

  it.effect("parses and replays a configured custom reasoning field", () =>
    Effect.gen(function* () {
-      const custom = LanguageModel.update(model, { compatibility: { reasoningField: "vendor_reasoning" } })
+      const custom = Model.update(model, { compatibility: { reasoningField: "vendor_reasoning" } })
      const response = yield* LLMClient.generate(LLMRequest.update(request, { model: custom })).pipe(
        Effect.provide(
          fixedResponse(
@@ -644,7 +647,9 @@ describe("OpenAI Chat route", () => {
        openai: { reasoningField: "vendor_reasoning" },
      })

-      const replay = yield* compileRequest(LLM.request({ model: custom, messages: [response.message] }))
+      const replay = yield* LLMClient.prepare<OpenAIChat.OpenAIChatBody>(
+        LLM.request({ model: custom, messages: [response.message] }),
+      )
      expect(replay.body.messages).toEqual([{ role: "assistant", content: "Hello", vendor_reasoning: "thinking" }])
    }),
  )
@@ -687,7 +692,9 @@ describe("OpenAI Chat route", () => {
        openai: { reasoningField: "reasoning", reasoningDetails: details },
      })

-      const replay = yield* compileRequest(LLM.request({ model, messages: [response.message] }))
+      const replay = yield* LLMClient.prepare<OpenAIChat.OpenAIChatBody>(
+        LLM.request({ model, messages: [response.message] }),
+      )
      expect(replay.body.messages).toEqual([
        {
          role: "assistant",
@@ -730,7 +737,9 @@ describe("OpenAI Chat route", () => {
        openai: { reasoningDetails: details },
      })

-      const replay = yield* compileRequest(LLM.request({ model, messages: [response.message] }))
+      const replay = yield* LLMClient.prepare<OpenAIChat.OpenAIChatBody>(
+        LLM.request({ model, messages: [response.message] }),
+      )
      expect(replay.body.messages).toEqual([{ role: "assistant", content: "Hello", reasoning_details: details }])
    }),
  )
@@ -755,7 +764,9 @@ describe("OpenAI Chat route", () => {
        openai: { reasoningField: "reasoning", reasoningDetails: details },
      })

-      const replay = yield* compileRequest(LLM.request({ model, messages: [response.message] }))
+      const replay = yield* LLMClient.prepare<OpenAIChat.OpenAIChatBody>(
+        LLM.request({ model, messages: [response.message] }),
+      )
      expect(replay.body.messages).toEqual([
        { role: "assistant", content: "Hello", reasoning: "thinking", reasoning_details: details },
      ])
@@ -828,7 +839,9 @@ describe("OpenAI Chat route", () => {
        openai: { reasoningDetails: [] },
      })

-      const replay = yield* compileRequest(LLM.request({ model, messages: [response.message] }))
+      const replay = yield* LLMClient.prepare<OpenAIChat.OpenAIChatBody>(
+        LLM.request({ model, messages: [response.message] }),
+      )
      expect(replay.body.messages).toEqual([{ role: "assistant", content: "Hello", reasoning_details: [] }])
    }),
  )
@@ -876,7 +889,9 @@ describe("OpenAI Chat route", () => {
        response.events.findIndex(LLMEvent.is.textStart),
      )

-      const replay = yield* compileRequest(LLM.request({ model, messages: [response.message] }))
+      const replay = yield* LLMClient.prepare<OpenAIChat.OpenAIChatBody>(
+        LLM.request({ model, messages: [response.message] }),
+      )
      expect(replay.body.messages).toEqual([
        { role: "assistant", content: "Hello", reasoning: "thinking", reasoning_details: merged },
      ])
@@ -903,7 +918,9 @@ describe("OpenAI Chat route", () => {
      expect(response.events.filter(LLMEvent.is.reasoningStart)).toHaveLength(1)
      expect(response.events.filter(LLMEvent.is.reasoningEnd)).toHaveLength(1)

-      const replay = yield* compileRequest(LLM.request({ model, messages: [response.message] }))
+      const replay = yield* LLMClient.prepare<OpenAIChat.OpenAIChatBody>(
+        LLM.request({ model, messages: [response.message] }),
+      )
      expect(replay.body.messages).toEqual([{ role: "assistant", content: null, reasoning_details: details }])
    }),
  )
@@ -933,7 +950,7 @@ describe("OpenAI Chat route", () => {
    Effect.gen(function* () {
      const first = { type: "reasoning.text", text: "first", signature: "signed-0", index: 0 }
      const second = { type: "reasoning.text", text: "second", signature: "signed-1", index: 1 }
-      const replay = yield* compileRequest(
+      const replay = yield* LLMClient.prepare<OpenAIChat.OpenAIChatBody>(
        LLM.request({
          model,
          messages: [
@@ -962,7 +979,7 @@ describe("OpenAI Chat route", () => {
  it.effect("retains scalar replay for mixed structured reasoning parts", () =>
    Effect.gen(function* () {
      const detail = { type: "reasoning.encrypted", data: "opaque", index: 0 }
-      const replay = yield* compileRequest(
+      const replay = yield* LLMClient.prepare<OpenAIChat.OpenAIChatBody>(
        LLM.request({
          model,
          messages: [
@@ -987,7 +1004,7 @@ describe("OpenAI Chat route", () => {
  it.effect("replays native scalar reasoning alongside native details", () =>
    Effect.gen(function* () {
      const details = [{ type: "reasoning.encrypted", data: "opaque", index: 0 }]
-      const replay = yield* compileRequest(
+      const replay = yield* LLMClient.prepare<OpenAIChat.OpenAIChatBody>(
        LLM.request({
          model,
          messages: [
@@ -1174,7 +1191,7 @@ describe("OpenAI Chat route", () => {
        Effect.flip,
      )

-      expect(error).toBeInstanceOf(AIError)
+      expect(error).toBeInstanceOf(LLMError)
      expect(error.reason).toMatchObject({ _tag: "InvalidRequest" })
      expect(error.message).toContain("HTTP 400")
    }),
@@ -3,7 +3,6 @@ import { Effect, Schema } from "effect"
 import { HttpClientRequest } from "effect/unstable/http"
 import { LLM, LLMRequest, Message, ToolCallPart, ToolChoice, ToolDefinition } from "../../src"
 import { Auth, LLMClient } from "../../src/route"
-import { compileRequest } from "../../src/route/client"
 import * as OpenAICompatible from "../../src/providers/openai-compatible"
 import * as OpenAICompatibleChat from "../../src/protocols/openai-compatible-chat"
 import { it } from "../lib/effect"
@@ -53,7 +52,7 @@ const providerFamilies = [
 describe("OpenAI-compatible Chat route", () => {
  it.effect("prepares generic Chat target", () =>
    Effect.gen(function* () {
-      const prepared = yield* compileRequest(
+      const prepared = yield* LLMClient.prepare(
        LLMRequest.update(request, {
          tools: [ToolDefinition.make({ name: "lookup", description: "Lookup data", inputSchema: { type: "object" } })],
          toolChoice: ToolChoice.make({ type: "required" }),
@@ -128,7 +127,7 @@ describe("OpenAI-compatible Chat route", () => {

  it.effect("matches AI SDK compatible basic request body fixture", () =>
    Effect.gen(function* () {
-      const prepared = yield* compileRequest(request)
+      const prepared = yield* LLMClient.prepare(request)

      expect(prepared.body).toEqual({
        model: "deepseek-chat",
@@ -144,23 +143,9 @@ describe("OpenAI-compatible Chat route", () => {
    }),
  )

-  it.effect("configures the max tokens request field", () =>
-    Effect.gen(function* () {
-      const compatible = OpenAICompatibleChat.route
-        .with({ provider: "custom", endpoint: { baseURL: "https://api.custom.test/v1" } })
-        .model({ id: "custom-model", compatibility: { maxTokensField: "max_completion_tokens" } })
-      const prepared = yield* compileRequest(
-        LLM.request({ model: compatible, prompt: "Say hello.", generation: { maxTokens: 20 } }),
-      )
-
-      expect(prepared.body).toMatchObject({ max_completion_tokens: 20 })
-      expect(prepared.body).not.toHaveProperty("max_tokens")
-    }),
-  )
-
  it.effect("matches AI SDK compatible tool request body fixture", () =>
    Effect.gen(function* () {
-      const prepared = yield* compileRequest(
+      const prepared = yield* LLMClient.prepare(
        LLM.request({
          id: "req_tool_parity",
          model,
--- a/Show More
+++ b/Show More
Author	SHA1	Message	Date
Brendan Allan	3aaff6ffe2	refactor(app): establish v2 session controller	2026-07-29 14:25:07 +08:00
Brendan Allan	99cae9766b	refactor(app): extract session timeline controller	2026-07-29 14:24:28 +08:00