Option C: rebuild SDK as a thin shell over the generated typed core

Replace the OpenAI-SDK delegation with a hand-written shell over the
OpenAPI-generated core (otari._client), generated from the otari spec.
The gateway is only OpenAI-compatible and has endpoints (notably the
Anthropic-shaped /messages) the OpenAI SDK cannot represent; generating
from the otari spec models the real contract.

Core (unmodified generator output) replaces the old _control_plane
subpackage and covers every endpoint with typed request/response models.

Shell (the real work):
- Auth modes preserved: platform -> Authorization: Bearer <token>;
  non-platform -> Otari-Key: Bearer <key>; control-plane -> Authorization:
  Bearer <admin/master key>. Fed into the generated ApiClient default
  headers and the streaming shim's httpx requests.
- Ergonomic methods keep the existing public names/signatures
  (completion/response/embedding/moderation/rerank/list_models/batches)
  and add message(...) for the previously-missing /messages endpoint.
  control_plane accessor now backed by _client.
- SSE streaming shim (_streaming.py): the generated core buffers and
  cannot stream, so stream=True does a raw httpx streaming POST, parses
  text/event-stream framing, terminates on [DONE], and yields typed
  ChatCompletionChunk (chat) / parsed JSON (responses, messages).
- Typed error mapping: generated ApiException (.status/.body) -> the
  errors.py hierarchy (Authentication 401/403, InsufficientFunds 402,
  ModelNotFound 404, BatchNotComplete 409, RateLimit 429, GatewayTimeout
  504, UpstreamProvider 502/5xx, UnsupportedCapability cross-mode, generic
  OtariError). The streaming path adapts failed responses through the same
  mapper.
- Async client wraps the sync generated core via asyncio.to_thread and
  streams natively over httpx.AsyncClient.

Drop the openai dependency from inference. Update ruff/mypy excludes to
src/otari/_client. Rewrite unit tests to mock the generated transport
(RESTClientObject.request) and respx for SSE; control-plane integration
tests updated to import from _client and verified green against a live
gateway.

Note: chat streaming cannot be live-verified in this sandbox (no provider
key on the gateway); the SSE shim is unit-tested over mocked chunk bytes.

Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>

Author: njbrake

README.md

@@ -195,6 +195,21 @@ response = client.response(
 print(response.output_text)
 ```
 
+### Messages API (Anthropic-shaped)
+
+The gateway's `/messages` endpoint (Anthropic message shape) is exposed via
+`message(...)`. Set `stream=True` to iterate raw message-stream event dicts.
+
+```python
+message = client.message(
+    model="anthropic:claude-3-5-sonnet",
+    messages=[{"role": "user", "content": "Hello!"}],
+    max_tokens=256,
+)
+
+print(message.content)
+```
+
 ### Embeddings
 
 ```python

pyproject.toml

@@ -22,9 +22,9 @@ classifiers = [
     "Typing :: Typed",
 ]
 dependencies = [
-    "openai>=1.99.3",
+    # Streaming shim (raw SSE over httpx); the generated core cannot stream.
     "httpx>=0.25.0",
-    # Runtime deps of the generated control-plane client (otari._control_plane).
+    # Runtime deps of the OpenAPI-generated core (otari._client).
     "urllib3>=2.1.0",
     "python-dateutil>=2.8.2",
     "pydantic>=2.0.0",
@@ -40,6 +40,7 @@ Issues = "https://github.com/mozilla-ai/otari-sdk-python/issues"
 dev = [
     "pytest>=8.0",
     "pytest-asyncio>=0.24",
+    "respx>=0.21",
     "ruff>=0.8",
     "mypy>=1.13",
 ]
@@ -57,8 +58,8 @@ markers = [
 [tool.ruff]
 target-version = "py311"
 line-length = 120
-# Generated control-plane client (OpenAPI Generator output) is not hand-linted.
-extend-exclude = ["src/otari/_control_plane"]
+# Generated core client (OpenAPI Generator output) is not hand-linted.
+extend-exclude = ["src/otari/_client"]
 
 [tool.ruff.lint]
 select = ["E", "F", "I", "N", "W", "UP", "S", "B", "A", "C4", "DTZ", "T10", "ISC", "ICN", "PIE", "PT", "RSE", "RET", "SIM", "TID", "TCH", "ARG", "PLC", "PLE", "PLW", "TRY", "FLY", "PERF", "RUF"]
@@ -77,10 +78,10 @@ python_version = "3.11"
 strict = true
 warn_return_any = true
 warn_unused_configs = true
-# Generated control-plane client (OpenAPI Generator output) is not type-checked here.
-exclude = ["src/otari/_control_plane/"]
+# Generated core client (OpenAPI Generator output) is not type-checked here.
+exclude = ["src/otari/_client/"]
 
 [[tool.mypy.overrides]]
-module = ["otari._control_plane.*"]
+module = ["otari._client.*"]
 ignore_errors = true
 ignore_missing_imports = true

src/otari/__init__.py

@@ -9,7 +9,7 @@
         platform_token="your-token-here",
     )
 
-    response = await client.completion(
+    response = client.completion(
         model="openai:gpt-4o-mini",
         messages=[{"role": "user", "content": "Hello!"}],
     )
@@ -33,23 +33,20 @@
     UpstreamProviderError,
 )
 from otari.types import (
-    AsyncStream,
     BatchRequestItem,
     BatchResult,
     BatchResultError,
     BatchResultItem,
     ChatCompletion,
     ChatCompletionChunk,
-    ChatCompletionMessageParam,
     CreateBatchParams,
     CreateEmbeddingResponse,
-    EmbeddingCreateParams,
     ListBatchesOptions,
-    Model,
+    MessageResponse,
+    ModelObject,
+    ModerationResponse,
     OtariClientOptions,
-    Response,
-    ResponseStreamEvent,
-    Stream,
+    RerankResponse,
 )
 
 try:
@@ -60,7 +57,6 @@
 
 __all__ = [
     "AsyncOtariClient",
-    "AsyncStream",
     "AuthenticationError",
     "BatchNotCompleteError",
     "BatchRequestItem",
@@ -69,23 +65,21 @@
     "BatchResultItem",
     "ChatCompletion",
     "ChatCompletionChunk",
-    "ChatCompletionMessageParam",
     "ControlPlane",
     "CreateBatchParams",
     "CreateEmbeddingResponse",
-    "EmbeddingCreateParams",
     "GatewayTimeoutError",
     "InsufficientFundsError",
     "ListBatchesOptions",
-    "Model",
+    "MessageResponse",
     "ModelNotFoundError",
+    "ModelObject",
+    "ModerationResponse",
     "OtariClient",
     "OtariClientOptions",
     "OtariError",
     "RateLimitError",
-    "Response",
-    "ResponseStreamEvent",
-    "Stream",
+    "RerankResponse",
     "UnsupportedCapabilityError",
     "UpstreamProviderError",
 ]