Feat: Add id and finish_reason to ModelResponse

.gitignore

@@ -17,3 +17,5 @@ examples/pydantic_ai_examples/.chat_app_messages.sqlite
 /docs-site/.wrangler/
 /CLAUDE.md
 node_modules/
+**.idea/
+.coverage*

docs/agents.md

@@ -155,6 +155,7 @@ async def main():
                 model_name='gpt-4o',
                 timestamp=datetime.datetime(...),
                 kind='response',
+                vendor_id=None,
             )
         ),
         End(data=FinalResult(output='Paris', tool_name=None, tool_call_id=None)),
@@ -226,6 +227,7 @@ async def main():
                     model_name='gpt-4o',
                     timestamp=datetime.datetime(...),
                     kind='response',
+                    vendor_id=None,
                 )
             ),
             End(data=FinalResult(output='Paris', tool_name=None, tool_call_id=None)),
@@ -829,6 +831,7 @@ with capture_run_messages() as messages:  # (2)!
                 model_name='gpt-4o',
                 timestamp=datetime.datetime(...),
                 kind='response',
+                vendor_id=None,
             ),
             ModelRequest(
                 parts=[
@@ -862,6 +865,7 @@ with capture_run_messages() as messages:  # (2)!
                 model_name='gpt-4o',
                 timestamp=datetime.datetime(...),
                 kind='response',
+                vendor_id=None,
             ),
         ]
         """

docs/direct.md

@@ -95,6 +95,7 @@ async def main():
         model_name='gpt-4.1-nano',
         timestamp=datetime.datetime(...),
         kind='response',
+        vendor_id=None,
     )
     """
 ```

docs/message-history.md

@@ -74,6 +74,7 @@ print(result.all_messages())
         model_name='gpt-4o',
         timestamp=datetime.datetime(...),
         kind='response',
+        vendor_id=None,
     ),
 ]
 """
@@ -159,6 +160,7 @@ async def main():
                 model_name='gpt-4o',
                 timestamp=datetime.datetime(...),
                 kind='response',
+                vendor_id=None,
             ),
         ]
         """
@@ -225,6 +227,7 @@ print(result2.all_messages())
         model_name='gpt-4o',
         timestamp=datetime.datetime(...),
         kind='response',
+        vendor_id=None,
     ),
     ModelRequest(
         parts=[
@@ -254,6 +257,7 @@ print(result2.all_messages())
         model_name='gpt-4o',
         timestamp=datetime.datetime(...),
         kind='response',
+        vendor_id=None,
     ),
 ]
 """
@@ -367,6 +371,7 @@ print(result2.all_messages())
         model_name='gpt-4o',
         timestamp=datetime.datetime(...),
         kind='response',
+        vendor_id=None,
     ),
     ModelRequest(
         parts=[
@@ -396,6 +401,7 @@ print(result2.all_messages())
         model_name='gemini-1.5-pro',
         timestamp=datetime.datetime(...),
         kind='response',
+        vendor_id=None,
     ),
 ]
 """

docs/models/index.md

@@ -105,6 +105,7 @@ print(response.all_messages())
         model_name='claude-3-5-sonnet-latest',
         timestamp=datetime.datetime(...),
         kind='response',
+        vendor_id=None,
     ),
 ]
 """

docs/tools.md

@@ -106,6 +106,7 @@ print(dice_result.all_messages())
         model_name='gemini-1.5-flash',
         timestamp=datetime.datetime(...),
         kind='response',
+        vendor_id=None,
     ),
     ModelRequest(
         parts=[
@@ -139,6 +140,7 @@ print(dice_result.all_messages())
         model_name='gemini-1.5-flash',
         timestamp=datetime.datetime(...),
         kind='response',
+        vendor_id=None,
     ),
     ModelRequest(
         parts=[
@@ -170,6 +172,7 @@ print(dice_result.all_messages())
         model_name='gemini-1.5-flash',
         timestamp=datetime.datetime(...),
         kind='response',
+        vendor_id=None,
     ),
 ]
 """

mkdocs.yml

@@ -105,25 +105,24 @@ theme:
   custom_dir: docs/.overrides
   palette:
     - media: "(prefers-color-scheme)"
-      scheme: default
       primary: pink
       accent: pink
       toggle:
-        icon: material/lightbulb
+        icon: material/brightness-auto
         name: "Switch to light mode"
     - media: "(prefers-color-scheme: light)"
       scheme: default
       primary: pink
       accent: pink
       toggle:
-        icon: material/lightbulb-outline
+        icon: material/brightness-7
         name: "Switch to dark mode"
     - media: "(prefers-color-scheme: dark)"
       scheme: slate
       primary: pink
       accent: pink
       toggle:
-        icon: material/lightbulb-auto-outline
+        icon: material/brightness-4
         name: "Switch to system preference"
   features:
     - search.suggest

pydantic_ai_slim/pydantic_ai/agent.py

@@ -585,6 +585,7 @@ async def main():
                         model_name='gpt-4o',
                         timestamp=datetime.datetime(...),
                         kind='response',
+                        vendor_id=None,
                     )
                 ),
                 End(data=FinalResult(output='Paris', tool_name=None, tool_call_id=None)),
@@ -1854,6 +1855,7 @@ async def main():
                     model_name='gpt-4o',
                     timestamp=datetime.datetime(...),
                     kind='response',
+                    vendor_id=None,
                 )
             ),
             End(data=FinalResult(output='Paris', tool_name=None, tool_call_id=None)),
@@ -1999,6 +2001,7 @@ async def main():
                             model_name='gpt-4o',
                             timestamp=datetime.datetime(...),
                             kind='response',
+                            vendor_id=None,
                         )
                     ),
                     End(data=FinalResult(output='Paris', tool_name=None, tool_call_id=None)),

pydantic_ai_slim/pydantic_ai/direct.py

@@ -52,6 +52,7 @@ async def main():
             model_name='claude-3-5-haiku-latest',
             timestamp=datetime.datetime(...),
             kind='response',
+            vendor_id=None,
         )
         '''
     ```
@@ -108,6 +109,7 @@ def model_request_sync(
         model_name='claude-3-5-haiku-latest',
         timestamp=datetime.datetime(...),
         kind='response',
+        vendor_id=None,
     )
     '''
     ```

pydantic_ai_slim/pydantic_ai/messages.py

@@ -553,6 +553,9 @@ class ModelResponse:
     model_name: str | None = None
     """The name of the model that generated the response."""
 
+    finish_reasons: list[str] = field(default_factory=list)
+    """The reasons why the model finished generating the response, one for each part of the response."""
+
     timestamp: datetime = field(default_factory=_now_utc)
     """The timestamp of the response.
 
@@ -562,6 +565,16 @@ class ModelResponse:
     kind: Literal['response'] = 'response'
     """Message type identifier, this is available on all parts as a discriminator."""
 
+    vendor_details: dict[str, Any] | None = field(default=None, repr=False)
+    """Additional vendor-specific details in a serializable format.
+
+    This allows storing selected vendor-specific data that isn't mapped to standard ModelResponse fields.
+    For OpenAI models, this may include 'logprobs', 'finish_reason', etc.
+    """
+
+    vendor_id: str | None = None
+    """Vendor ID as specified by the model provider. This can be used to track the specific request to the model."""
+
     def otel_events(self) -> list[Event]:
         """Return OpenTelemetry events for the response."""
         result: list[Event] = []

pydantic_ai_slim/pydantic_ai/models/anthropic.py

@@ -262,7 +262,7 @@ def _process_response(self, response: AnthropicMessage) -> ModelResponse:
                     )
                 )
 
-        return ModelResponse(items, usage=_map_usage(response), model_name=response.model)
+        return ModelResponse(items, usage=_map_usage(response), model_name=response.model, vendor_id=response.id)
 
     async def _process_streamed_response(self, response: AsyncStream[RawMessageStreamEvent]) -> StreamedResponse:
         peekable_response = _utils.PeekableAsyncStream(response)

pydantic_ai_slim/pydantic_ai/models/bedrock.py

@@ -271,7 +271,8 @@ async def _process_response(self, response: ConverseResponseTypeDef) -> ModelRes
             response_tokens=response['usage']['outputTokens'],
             total_tokens=response['usage']['totalTokens'],
         )
-        return ModelResponse(items, usage=u, model_name=self.model_name)
+        vendor_id = response.get('ResponseMetadata', {}).get('RequestId', None)
+        return ModelResponse(items, usage=u, model_name=self.model_name, vendor_id=vendor_id)
 
     @overload
     async def _messages_create(

pydantic_ai_slim/pydantic_ai/models/gemini.py

@@ -273,9 +273,21 @@ def _process_response(self, response: _GeminiResponse) -> ModelResponse:
                     'Content field missing from Gemini response', str(response)
                 )
         parts = response['candidates'][0]['content']['parts']
+        finish_reasons = [
+            finish_reason
+            for finish_reason in [response['candidates'][0].get('finish_reason')]
+            if finish_reason is not None
+        ]
+        vendor_id = response.get('vendor_id', None)
         usage = _metadata_as_usage(response)
         usage.requests = 1
-        return _process_response_from_parts(parts, response.get('model_version', self._model_name), usage)
+        return _process_response_from_parts(
+            parts,
+            response.get('model_version', self._model_name),
+            usage,
+            vendor_id=vendor_id,
+            finish_reasons=finish_reasons,
+        )
 
     async def _process_streamed_response(self, http_response: HTTPResponse) -> StreamedResponse:
         """Process a streamed response, and prepare a streaming response to return."""
@@ -597,7 +609,11 @@ def _function_call_part_from_call(tool: ToolCallPart) -> _GeminiFunctionCallPart
 
 
 def _process_response_from_parts(
-    parts: Sequence[_GeminiPartUnion], model_name: GeminiModelName, usage: usage.Usage
+    parts: Sequence[_GeminiPartUnion],
+    model_name: GeminiModelName,
+    usage: usage.Usage,
+    vendor_id: str | None,
+    finish_reasons: list[str],
 ) -> ModelResponse:
     items: list[ModelResponsePart] = []
     for part in parts:
@@ -609,7 +625,9 @@ def _process_response_from_parts(
             raise UnexpectedModelBehavior(
                 f'Unsupported response from Gemini, expected all parts to be function calls or text, got: {part!r}'
             )
-    return ModelResponse(parts=items, usage=usage, model_name=model_name)
+    return ModelResponse(
+        parts=items, usage=usage, model_name=model_name, vendor_id=vendor_id, finish_reasons=finish_reasons
+    )
 
 
 class _GeminiFunctionCall(TypedDict):
@@ -721,6 +739,7 @@ class _GeminiResponse(TypedDict):
     usage_metadata: NotRequired[Annotated[_GeminiUsageMetaData, pydantic.Field(alias='usageMetadata')]]
     prompt_feedback: NotRequired[Annotated[_GeminiPromptFeedback, pydantic.Field(alias='promptFeedback')]]
     model_version: NotRequired[Annotated[str, pydantic.Field(alias='modelVersion')]]
+    vendor_id: NotRequired[Annotated[str, pydantic.Field(alias='responseId')]]
 
 
 class _GeminiCandidates(TypedDict):

pydantic_ai_slim/pydantic_ai/models/groq.py

@@ -239,7 +239,9 @@ def _process_response(self, response: chat.ChatCompletion) -> ModelResponse:
         if choice.message.tool_calls is not None:
             for c in choice.message.tool_calls:
                 items.append(ToolCallPart(tool_name=c.function.name, args=c.function.arguments, tool_call_id=c.id))
-        return ModelResponse(items, usage=_map_usage(response), model_name=response.model, timestamp=timestamp)
+        return ModelResponse(
+            items, usage=_map_usage(response), model_name=response.model, timestamp=timestamp, vendor_id=response.id
+        )
 
     async def _process_streamed_response(self, response: AsyncStream[chat.ChatCompletionChunk]) -> GroqStreamedResponse:
         """Process a streamed response, and prepare a streaming response to return."""

pydantic_ai_slim/pydantic_ai/models/mistral.py

@@ -325,7 +325,9 @@ def _process_response(self, response: MistralChatCompletionResponse) -> ModelRes
                 tool = self._map_mistral_to_pydantic_tool_call(tool_call=tool_call)
                 parts.append(tool)
 
-        return ModelResponse(parts, usage=_map_usage(response), model_name=response.model, timestamp=timestamp)
+        return ModelResponse(
+            parts, usage=_map_usage(response), model_name=response.model, timestamp=timestamp, vendor_id=response.id
+        )
 
     async def _process_streamed_response(
         self,

pydantic_ai_slim/pydantic_ai/models/openai.py

@@ -104,6 +104,12 @@ class OpenAIModelSettings(ModelSettings, total=False):
     result in faster responses and fewer tokens used on reasoning in a response.
     """
 
+    openai_logprobs: bool
+    """Include log probabilities in the response."""
+
+    openai_top_logprobs: int
+    """Include log probabilities of the top n tokens in the response."""
+
     openai_user: str
     """A unique identifier representing the end-user, which can help OpenAI monitor and detect abuse.
 
@@ -287,6 +293,8 @@ async def _completions_create(
                 frequency_penalty=model_settings.get('frequency_penalty', NOT_GIVEN),
                 logit_bias=model_settings.get('logit_bias', NOT_GIVEN),
                 reasoning_effort=model_settings.get('openai_reasoning_effort', NOT_GIVEN),
+                logprobs=model_settings.get('openai_logprobs', NOT_GIVEN),
+                top_logprobs=model_settings.get('openai_top_logprobs', NOT_GIVEN),
                 user=model_settings.get('openai_user', NOT_GIVEN),
                 extra_headers=extra_headers,
                 extra_body=model_settings.get('extra_body'),
@@ -301,12 +309,38 @@ def _process_response(self, response: chat.ChatCompletion) -> ModelResponse:
         timestamp = datetime.fromtimestamp(response.created, tz=timezone.utc)
         choice = response.choices[0]
         items: list[ModelResponsePart] = []
+        vendor_details: dict[str, Any] | None = None
+
+        # Add logprobs to vendor_details if available
+        if choice.logprobs is not None and choice.logprobs.content:
+            # Convert logprobs to a serializable format
+            vendor_details = {
+                'logprobs': [
+                    {
+                        'token': lp.token,
+                        'bytes': lp.bytes,
+                        'logprob': lp.logprob,
+                        'top_logprobs': [
+                            {'token': tlp.token, 'bytes': tlp.bytes, 'logprob': tlp.logprob} for tlp in lp.top_logprobs
+                        ],
+                    }
+                    for lp in choice.logprobs.content
+                ],
+            }
+
         if choice.message.content is not None:
             items.append(TextPart(choice.message.content))
         if choice.message.tool_calls is not None:
             for c in choice.message.tool_calls:
                 items.append(ToolCallPart(c.function.name, c.function.arguments, tool_call_id=c.id))
-        return ModelResponse(items, usage=_map_usage(response), model_name=response.model, timestamp=timestamp)
+        return ModelResponse(
+            items,
+            usage=_map_usage(response),
+            model_name=response.model,
+            timestamp=timestamp,
+            vendor_details=vendor_details,
+            vendor_id=response.id,
+        )
 
     async def _process_streamed_response(self, response: AsyncStream[ChatCompletionChunk]) -> OpenAIStreamedResponse:
         """Process a streamed response, and prepare a streaming response to return."""

pydantic_graph/pydantic_graph/graph.py

@@ -10,7 +10,6 @@
 
 import logfire_api
 import typing_extensions
-from opentelemetry.trace import Span
 from typing_extensions import deprecated
 from typing_inspection import typing_objects
 
@@ -212,7 +211,7 @@ async def iter(
         state: StateT = None,
         deps: DepsT = None,
         persistence: BaseStatePersistence[StateT, RunEndT] | None = None,
-        span: AbstractContextManager[Span] | None = None,
+        span: AbstractContextManager[AbstractSpan] | None = None,
         infer_name: bool = True,
     ) -> AsyncIterator[GraphRun[StateT, DepsT, RunEndT]]:
         """A contextmanager which can be used to iterate over the graph's nodes as they are executed.