Store OpenAI Responses text part ID to prevent error with reasoning (#2882)

Author: DouweM

pydantic_ai_slim/pydantic_ai/_parts_manager.py

@@ -71,6 +71,7 @@ def handle_text_delta(
         *,
         vendor_part_id: VendorId | None,
         content: str,
+        id: str | None = None,
         thinking_tags: tuple[str, str] | None = None,
         ignore_leading_whitespace: bool = False,
     ) -> ModelResponseStreamEvent | None:
@@ -85,6 +86,7 @@ def handle_text_delta(
                 of text. If None, a new part will be created unless the latest part is already
                 a TextPart.
             content: The text content to append to the appropriate TextPart.
+            id: An optional id for the text part.
             thinking_tags: If provided, will handle content between the thinking tags as thinking parts.
             ignore_leading_whitespace: If True, will ignore leading whitespace in the content.
 
@@ -137,7 +139,7 @@ def handle_text_delta(
 
             # There is no existing text part that should be updated, so create a new one
             new_part_index = len(self._parts)
-            part = TextPart(content=content)
+            part = TextPart(content=content, id=id)
             if vendor_part_id is not None:
                 self._vendor_id_to_part_index[vendor_part_id] = new_part_index
             self._parts.append(part)

pydantic_ai_slim/pydantic_ai/messages.py

@@ -870,6 +870,9 @@ class TextPart:
 
     _: KW_ONLY
 
+    id: str | None = None
+    """An optional identifier of the text part."""
+
     part_kind: Literal['text'] = 'text'
     """Part type identifier, this is available on all parts as a discriminator."""
 

pydantic_ai_slim/pydantic_ai/models/openai.py

@@ -886,7 +886,7 @@ def _process_response(self, response: responses.Response) -> ModelResponse:
             elif isinstance(item, responses.ResponseOutputMessage):
                 for content in item.content:
                     if isinstance(content, responses.ResponseOutputText):  # pragma: no branch
-                        items.append(TextPart(content.text))
+                        items.append(TextPart(content.text, id=item.id))
             elif isinstance(item, responses.ResponseFunctionToolCall):
                 items.append(
                     ToolCallPart(item.name, item.arguments, tool_call_id=_combine_tool_call_ids(item.call_id, item.id))
@@ -1122,10 +1122,31 @@ async def _map_messages(  # noqa: C901
                     else:
                         assert_never(part)
             elif isinstance(message, ModelResponse):
+                message_item: responses.ResponseOutputMessageParam | None = None
                 reasoning_item: responses.ResponseReasoningItemParam | None = None
                 for item in message.parts:
                     if isinstance(item, TextPart):
-                        openai_messages.append(responses.EasyInputMessageParam(role='assistant', content=item.content))
+                        if item.id and item.id.startswith('msg_'):
+                            if message_item is None or message_item['id'] != item.id:  # pragma: no branch
+                                message_item = responses.ResponseOutputMessageParam(
+                                    role='assistant',
+                                    id=item.id or _utils.generate_tool_call_id(),
+                                    content=[],
+                                    type='message',
+                                    status='completed',
+                                )
+                                openai_messages.append(message_item)
+
+                            message_item['content'] = [
+                                *message_item['content'],
+                                responses.ResponseOutputTextParam(
+                                    text=item.content, type='output_text', annotations=[]
+                                ),
+                            ]
+                        else:
+                            openai_messages.append(
+                                responses.EasyInputMessageParam(role='assistant', content=item.content)
+                            )
                     elif isinstance(item, ToolCallPart):
                         openai_messages.append(self._map_tool_call(item))
                     elif isinstance(item, BuiltinToolCallPart | BuiltinToolReturnPart):
@@ -1436,7 +1457,9 @@ async def _get_event_iterator(self) -> AsyncIterator[ModelResponseStreamEvent]:
                 pass  # there's nothing we need to do here
 
             elif isinstance(chunk, responses.ResponseTextDeltaEvent):
-                maybe_event = self._parts_manager.handle_text_delta(vendor_part_id=chunk.item_id, content=chunk.delta)
+                maybe_event = self._parts_manager.handle_text_delta(
+                    vendor_part_id=chunk.item_id, content=chunk.delta, id=chunk.item_id
+                )
                 if maybe_event is not None:  # pragma: no branch
                     yield maybe_event
 

tests/models/cassettes/test_model_names/test_known_model_names.yaml

@@ -99,53 +99,19 @@ interactions:
       alt-svc:
       - h3=":443"; ma=86400
       content-length:
-      - '762'
+      - '99'
       content-type:
       - application/json
       referrer-policy:
       - strict-origin-when-cross-origin
       strict-transport-security:
       - max-age=3600; includeSubDomains
     parsed_body:
-      data:
-      - created: 0
-        id: qwen-3-235b-a22b-thinking-2507
-        object: model
-        owned_by: Cerebras
-      - created: 0
-        id: llama-3.3-70b
-        object: model
-        owned_by: Cerebras
-      - created: 0
-        id: qwen-3-235b-a22b-instruct-2507
-        object: model
-        owned_by: Cerebras
-      - created: 0
-        id: llama-4-scout-17b-16e-instruct
-        object: model
-        owned_by: Cerebras
-      - created: 0
-        id: gpt-oss-120b
-        object: model
-        owned_by: Cerebras
-      - created: 0
-        id: qwen-3-coder-480b
-        object: model
-        owned_by: Cerebras
-      - created: 0
-        id: llama3.1-8b
-        object: model
-        owned_by: Cerebras
-      - created: 0
-        id: llama-4-maverick-17b-128e-instruct
-        object: model
-        owned_by: Cerebras
-      - created: 0
-        id: qwen-3-32b
-        object: model
-        owned_by: Cerebras
-      object: list
+      code: wrong_api_key
+      message: Wrong API Key
+      param: api_key
+      type: invalid_request_error
     status:
-      code: 200
-      message: OK
+      code: 401
+      message: Unauthorized
 version: 1