fix(sdk/python): address Responses API review feedback

- Rename model and variant Responses client factories to get_responses_client
  to match existing get_*_client naming.
- Use FoundryLocalException for Responses API transport and parsing errors
  instead of exporting a dedicated ResponsesAPIError.
- Keep only the foundry-local-core version bump in requirements.txt and
  restore existing ORT dependency markers/order.

Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>

Author: MaanavD

sdk/python/requirements.txt

@@ -1,8 +1,9 @@
 pydantic>=2.0.0
 requests>=2.32.4
 openai>=2.24.0
+# Standard native binary packages from the ORT-Nightly PyPI feed.
 foundry-local-core==1.0.0
-onnxruntime-gpu==1.24.4; platform_system == "Linux"
-onnxruntime-core==1.24.4; platform_system != "Linux"
-onnxruntime-genai-cuda==0.13.1; platform_system == "Linux"
-onnxruntime-genai-core==0.13.1; platform_system != "Linux"
+onnxruntime-core==1.24.4; sys_platform != "linux"
+onnxruntime-gpu==1.24.4; sys_platform == "linux"
+onnxruntime-genai-core==0.13.1; sys_platform != "linux"
+onnxruntime-genai-cuda==0.13.1; sys_platform == "linux"

sdk/python/src/__init__.py

@@ -7,7 +7,7 @@
 
 from .configuration import Configuration
 from .foundry_local_manager import FoundryLocalManager
-from .openai.responses_client import ResponsesAPIError, ResponsesClient, ResponsesClientSettings
+from .openai.responses_client import ResponsesClient, ResponsesClientSettings
 from .openai.responses_types import (
     ContentPart,
     DeleteResponseResult,
@@ -72,7 +72,6 @@
     "ResponseObject",
     "ResponseOutputItem",
     "ResponseUsage",
-    "ResponsesAPIError",
     "ResponsesClient",
     "ResponsesClientSettings",
     "StreamingEvent",

sdk/python/src/detail/model.py

@@ -148,6 +148,6 @@ def get_embedding_client(self) -> EmbeddingClient:
         """Get an embedding client for the currently selected variant."""
         return self._selected_variant.get_embedding_client()
 
-    def create_responses_client(self, base_url: str) -> "ResponsesClient":
-        """Create a Responses API client for the currently selected variant."""
-        return self._selected_variant.create_responses_client(base_url)
+    def get_responses_client(self, base_url: str) -> "ResponsesClient":
+        """Get a Responses API client for the currently selected variant."""
+        return self._selected_variant.get_responses_client(base_url)

sdk/python/src/detail/model_variant.py

@@ -177,8 +177,8 @@ def get_embedding_client(self) -> EmbeddingClient:
         """Create an OpenAI-compatible ``EmbeddingClient`` for this variant."""
         return EmbeddingClient(self.id, self._core_interop)
 
-    def create_responses_client(self, base_url: str) -> ResponsesClient:
-        """Create a Responses API client for this variant.
+    def get_responses_client(self, base_url: str) -> ResponsesClient:
+        """Create an OpenAI-compatible ``ResponsesClient`` for this variant.
 
         :param base_url: Base URL of the running Foundry Local web service
             (e.g. ``manager.urls[0]``).

sdk/python/src/imodel.py

@@ -138,9 +138,9 @@ def get_embedding_client(self) -> 'EmbeddingClient':
         pass
 
     @abstractmethod
-    def create_responses_client(self, base_url: str) -> 'ResponsesClient':
+    def get_responses_client(self, base_url: str) -> 'ResponsesClient':
         """
-        Create an OpenAI Responses API client bound to the running web service.
+        Get an OpenAI Responses API client bound to the running web service.
 
         Unlike the other clients, the Responses API is HTTP-only and requires
         the Foundry Local web service to be started. Pass the base URL

sdk/python/src/openai/__init__.py

@@ -14,7 +14,7 @@
     LiveAudioTranscriptionResponse,
     TranscriptionContentPart,
 )
-from .responses_client import ResponsesClient, ResponsesClientSettings, ResponsesAPIError
+from .responses_client import ResponsesClient, ResponsesClientSettings
 
 __all__ = [
     "AudioClient",
@@ -25,8 +25,7 @@
     "LiveAudioTranscriptionOptions",
     "LiveAudioTranscriptionResponse",
     "LiveAudioTranscriptionSession",
-    "ResponsesAPIError",
     "ResponsesClient",
     "ResponsesClientSettings",
     "TranscriptionContentPart",
-]
+]

sdk/python/src/openai/responses_client.py

@@ -36,6 +36,7 @@
 
 import requests
 
+from ..exception import FoundryLocalException
 from .responses_types import (
     DeleteResponseResult,
     InputItemsListResponse,
@@ -105,20 +106,11 @@ def _serialize(self) -> Dict[str, Any]:
         return {k: v for k, v in raw.items() if v is not None}
 
 
-class ResponsesAPIError(Exception):
-    """Raised for HTTP/transport errors against the Responses API."""
-
-    def __init__(self, message: str, status_code: Optional[int] = None, body: Optional[str] = None):
-        super().__init__(message)
-        self.status_code = status_code
-        self.body = body
-
-
 class ResponsesClient:
     """Client for the OpenAI Responses API served by Foundry Local.
 
     Construct via ``manager.create_responses_client(model_id)`` or
-    ``model.create_responses_client(base_url)``.
+    ``model.get_responses_client(base_url)``.
     """
 
     def __init__(self, base_url: str, model_id: Optional[str] = None):
@@ -294,7 +286,7 @@ def _request_json(self, method: str, path: str, body: Optional[Dict[str, Any]] =
                     timeout=timeout,
                 )
         except requests.RequestException as e:
-            raise ResponsesAPIError(f"Network error calling {method} {path}: {e}") from e
+            raise FoundryLocalException(f"Network error calling {method} {path}: {e}") from e
 
         return self._handle_json_response(resp, method, path)
 
@@ -305,15 +297,13 @@ def _post_json(self, path: str, body: Dict[str, Any]) -> Dict[str, Any]:
     def _handle_json_response(resp: requests.Response, method: str, path: str) -> Dict[str, Any]:
         text = resp.text
         if not resp.ok:
-            raise ResponsesAPIError(
-                f"Responses API error ({resp.status_code}) for {method} {path}: {text[:500]}",
-                status_code=resp.status_code,
-                body=text,
+            raise FoundryLocalException(
+                f"Responses API error ({resp.status_code}) for {method} {path}: {text[:500]}"
             )
         try:
             return json.loads(text) if text else {}
         except json.JSONDecodeError as e:
-            raise ResponsesAPIError(
+            raise FoundryLocalException(
                 f"Failed to parse response JSON from {method} {path}: {text[:200]}"
             ) from e
 
@@ -332,15 +322,13 @@ def _post_stream(
                 timeout=(connect_timeout, None),
             )
         except requests.RequestException as e:
-            raise ResponsesAPIError(f"Network error calling POST {path}: {e}") from e
+            raise FoundryLocalException(f"Network error calling POST {path}: {e}") from e
 
         if not resp.ok:
             body_text = resp.text
             resp.close()
-            raise ResponsesAPIError(
-                f"Responses API error ({resp.status_code}) for POST {path}: {body_text[:500]}",
-                status_code=resp.status_code,
-                body=body_text,
+            raise FoundryLocalException(
+                f"Responses API error ({resp.status_code}) for POST {path}: {body_text[:500]}"
             )
 
         return _iter_sse_events(resp)
@@ -409,7 +397,7 @@ def _parse_sse_block(block: str) -> Any:
     try:
         parsed = json.loads(data)
     except json.JSONDecodeError as e:
-        raise ResponsesAPIError(f"Failed to parse streaming event JSON: {e}") from e
+        raise FoundryLocalException(f"Failed to parse streaming event JSON: {e}") from e
     if not isinstance(parsed, dict):
         return None
     return parse_streaming_event(parsed)
@@ -418,5 +406,4 @@ def _parse_sse_block(block: str) -> Any:
 __all__ = [
     "ResponsesClient",
     "ResponsesClientSettings",
-    "ResponsesAPIError",
 ]

sdk/python/test/openai/test_responses_client.py

@@ -18,8 +18,8 @@
 
 import pytest
 
+from foundry_local_sdk.exception import FoundryLocalException
 from foundry_local_sdk.openai.responses_client import (
-    ResponsesAPIError,
     ResponsesClient,
     ResponsesClientSettings,
     _parse_sse_block,
@@ -291,7 +291,7 @@ def test_multi_line_data(self):
 
     def test_invalid_json_raises(self):
         block = 'data: {not valid json'
-        with pytest.raises(ResponsesAPIError):
+        with pytest.raises(FoundryLocalException):
             _parse_sse_block(block)
 
     def test_empty_block_returns_none(self):
@@ -527,15 +527,15 @@ def test_delete_parses_result(self):
         assert result.deleted is True
         assert result.id == "resp_1"
 
-    def test_http_error_raises_responses_api_error(self):
+    def test_http_error_raises_foundry_local_exception(self):
         resp = MagicMock()
         resp.ok = False
         resp.status_code = 400
         resp.text = '{"error":{"message":"bad"}}'
         with patch("foundry_local_sdk.openai.responses_client.requests.request", return_value=resp):
-            with pytest.raises(ResponsesAPIError) as excinfo:
+            with pytest.raises(FoundryLocalException) as excinfo:
                 self.client.create("hi")
-        assert excinfo.value.status_code == 400
+        assert "400" in str(excinfo.value)
         assert "bad" in str(excinfo.value)
 
     def test_create_streaming_yields_events(self):
@@ -566,9 +566,9 @@ def test_streaming_http_error(self):
         resp.text = "boom"
         resp.close = MagicMock()
         with patch("foundry_local_sdk.openai.responses_client.requests.post", return_value=resp):
-            with pytest.raises(ResponsesAPIError) as excinfo:
+            with pytest.raises(FoundryLocalException) as excinfo:
                 list(self.client.create_streaming("hi"))
-        assert excinfo.value.status_code == 500
+        assert "500" in str(excinfo.value)
 
     def test_settings_merge_precedence(self):
         self.client.settings.temperature = 0.1