feat: Add id and finish_reason to ModelResponse

pydantic_ai_slim/pydantic_ai/messages.py

@@ -789,8 +789,40 @@ class ModelResponse:
     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."""
+    id: str | None = None
+    """Unique identifier for the model response, e.g. as returned by the model provider (OpenAI, etc)."""
+
+    finish_reason: str | None = None
+    """The reason the model finished generating this response, e.g. 'stop', 'length', etc."""
+
+    @property
+    def vendor_id(self) -> str | None:
+        """Vendor ID as specified by the model provider. This can be used to track the specific request to the model.
+
+        This is deprecated, use `id` instead.
+        """
+        import warnings
+
+        warnings.warn('vendor_id is deprecated, use id instead', DeprecationWarning, stacklevel=2)
+        return self.id
+
+    @vendor_id.setter
+    def vendor_id(self, value: str | None) -> None:
+        """Set the vendor ID.
+
+        This is deprecated, use `id` instead.
+        """
+        import warnings
+
+        warnings.warn('vendor_id is deprecated, use id instead', DeprecationWarning, stacklevel=2)
+        self.id = value
+
+    def __post_init__(self) -> None:
+        """Ensure vendor_details contains finish_reason for backward compatibility."""
+        if self.finish_reason and self.vendor_details is None:
+            self.vendor_details = {}
+        if self.finish_reason and self.vendor_details is not None:
+            self.vendor_details['finish_reason'] = self.finish_reason
 
     def otel_events(self, settings: InstrumentationSettings) -> list[Event]:
         """Return OpenTelemetry events for the response."""

pydantic_ai_slim/pydantic_ai/models/anthropic.py

@@ -282,7 +282,7 @@ def _process_response(self, response: BetaMessage) -> ModelResponse:
                     )
                 )
 
-        return ModelResponse(items, usage=_map_usage(response), model_name=response.model, vendor_id=response.id)
+        return ModelResponse(items, usage=_map_usage(response), model_name=response.model, id=response.id)
 
     async def _process_streamed_response(self, response: AsyncStream[BetaRawMessageStreamEvent]) -> StreamedResponse:
         peekable_response = _utils.PeekableAsyncStream(response)

pydantic_ai_slim/pydantic_ai/models/bedrock.py

@@ -296,7 +296,7 @@ async def _process_response(self, response: ConverseResponseTypeDef) -> ModelRes
             total_tokens=response['usage']['totalTokens'],
         )
         vendor_id = response.get('ResponseMetadata', {}).get('RequestId', None)
-        return ModelResponse(items, usage=u, model_name=self.model_name, vendor_id=vendor_id)
+        return ModelResponse(items, usage=u, model_name=self.model_name, id=vendor_id)
 
     @overload
     async def _messages_create(

pydantic_ai_slim/pydantic_ai/models/function.py

@@ -137,6 +137,8 @@ async def request(
         if not response.usage.has_values():  # pragma: no branch
             response.usage = _estimate_usage(chain(messages, [response]))
             response.usage.requests = 1
+        response.id = getattr(response, 'id', None)
+        response.finish_reason = getattr(response, 'finish_reason', None)
         return response
 
     @asynccontextmanager

pydantic_ai_slim/pydantic_ai/models/gemini.py

@@ -273,15 +273,16 @@ def _process_response(self, response: _GeminiResponse) -> ModelResponse:
         parts = response['candidates'][0]['content']['parts']
         vendor_id = response.get('vendor_id', None)
         finish_reason = response['candidates'][0].get('finish_reason')
+        vendor_details = {}
         if finish_reason:
-            vendor_details = {'finish_reason': finish_reason}
+            vendor_details['finish_reason'] = finish_reason
         usage = _metadata_as_usage(response)
         usage.requests = 1
         return _process_response_from_parts(
             parts,
             response.get('model_version', self._model_name),
             usage,
-            vendor_id=vendor_id,
+            id=vendor_id,
             vendor_details=vendor_details,
         )
 
@@ -662,7 +663,7 @@ def _process_response_from_parts(
     parts: Sequence[_GeminiPartUnion],
     model_name: GeminiModelName,
     usage: usage.Usage,
-    vendor_id: str | None,
+    id: str | None,
     vendor_details: dict[str, Any] | None = None,
 ) -> ModelResponse:
     items: list[ModelResponsePart] = []
@@ -680,9 +681,7 @@ 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, vendor_id=vendor_id, vendor_details=vendor_details
-    )
+    return ModelResponse(parts=items, usage=usage, model_name=model_name, id=id, vendor_details=vendor_details)
 
 
 class _GeminiFunctionCall(TypedDict):

pydantic_ai_slim/pydantic_ai/models/google.py

@@ -322,7 +322,7 @@ def _process_response(self, response: GenerateContentResponse) -> ModelResponse:
         usage = _metadata_as_usage(response)
         usage.requests = 1
         return _process_response_from_parts(
-            parts, response.model_version or self._model_name, usage, vendor_id=vendor_id, vendor_details=vendor_details
+            parts, response.model_version or self._model_name, usage, id=vendor_id, vendor_details=vendor_details
         )
 
     async def _process_streamed_response(self, response: AsyncIterator[GenerateContentResponse]) -> StreamedResponse:
@@ -506,7 +506,7 @@ def _process_response_from_parts(
     parts: list[Part],
     model_name: GoogleModelName,
     usage: usage.Usage,
-    vendor_id: str | None,
+    id: str | None,
     vendor_details: dict[str, Any] | None = None,
 ) -> ModelResponse:
     items: list[ModelResponsePart] = []
@@ -526,9 +526,7 @@ 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, model_name=model_name, usage=usage, vendor_id=vendor_id, vendor_details=vendor_details
-    )
+    return ModelResponse(parts=items, model_name=model_name, usage=usage, id=id, vendor_details=vendor_details)
 
 
 def _function_declaration_from_tool(tool: ToolDefinition) -> FunctionDeclarationDict:

pydantic_ai_slim/pydantic_ai/models/groq.py

@@ -266,7 +266,7 @@ def _process_response(self, response: chat.ChatCompletion) -> ModelResponse:
             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, vendor_id=response.id
+            items, usage=_map_usage(response), model_name=response.model, timestamp=timestamp, id=response.id
         )
 
     async def _process_streamed_response(self, response: AsyncStream[chat.ChatCompletionChunk]) -> GroqStreamedResponse:

pydantic_ai_slim/pydantic_ai/models/huggingface.py

@@ -253,7 +253,7 @@ def _process_response(self, response: ChatCompletionOutput) -> ModelResponse:
             usage=_map_usage(response),
             model_name=response.model,
             timestamp=timestamp,
-            vendor_id=response.id,
+            id=response.id,
         )
 
     async def _process_streamed_response(self, response: AsyncIterable[ChatCompletionStreamOutput]) -> StreamedResponse:

pydantic_ai_slim/pydantic_ai/models/mistral.py

@@ -341,7 +341,7 @@ def _process_response(self, response: MistralChatCompletionResponse) -> ModelRes
                 parts.append(tool)
 
         return ModelResponse(
-            parts, usage=_map_usage(response), model_name=response.model, timestamp=timestamp, vendor_id=response.id
+            parts, usage=_map_usage(response), model_name=response.model, timestamp=timestamp, id=response.id
         )
 
     async def _process_streamed_response(

pydantic_ai_slim/pydantic_ai/models/openai.py

@@ -419,7 +419,8 @@ def _process_response(self, response: chat.ChatCompletion | str) -> ModelRespons
             model_name=response.model,
             timestamp=timestamp,
             vendor_details=vendor_details,
-            vendor_id=response.id,
+            id=response.id,
+            finish_reason=choice.finish_reason,
         )
 
     async def _process_streamed_response(self, response: AsyncStream[ChatCompletionChunk]) -> OpenAIStreamedResponse:
@@ -706,7 +707,7 @@ def _process_response(self, response: responses.Response) -> ModelResponse:
             items,
             usage=_map_usage(response),
             model_name=response.model,
-            vendor_id=response.id,
+            id=response.id,
             timestamp=timestamp,
         )
 

pydantic_ai_slim/pydantic_ai/models/test.py

@@ -227,23 +227,44 @@ def _request(
                                 output[part.tool_name] = part.content
                 if output:
                     return ModelResponse(
-                        parts=[TextPart(pydantic_core.to_json(output).decode())], model_name=self._model_name
+                        parts=[TextPart(pydantic_core.to_json(output).decode())],
+                        model_name=self._model_name,
+                        id=None,
+                        finish_reason=None,
                     )
                 else:
-                    return ModelResponse(parts=[TextPart('success (no tool calls)')], model_name=self._model_name)
+                    return ModelResponse(
+                        parts=[TextPart('success (no tool calls)')],
+                        model_name=self._model_name,
+                        id=None,
+                        finish_reason=None,
+                    )
             else:
-                return ModelResponse(parts=[TextPart(response_text)], model_name=self._model_name)
+                return ModelResponse(
+                    parts=[TextPart(response_text)],
+                    model_name=self._model_name,
+                    id=None,
+                    finish_reason=None,
+                )
         else:
             assert output_tools, 'No output tools provided'
             custom_output_args = output_wrapper.value
             output_tool = output_tools[self.seed % len(output_tools)]
             if custom_output_args is not None:
                 return ModelResponse(
-                    parts=[ToolCallPart(output_tool.name, custom_output_args)], model_name=self._model_name
+                    parts=[ToolCallPart(output_tool.name, custom_output_args)],
+                    model_name=self._model_name,
+                    id=None,
+                    finish_reason=None,
                 )
             else:
                 response_args = self.gen_tool_args(output_tool)
-                return ModelResponse(parts=[ToolCallPart(output_tool.name, response_args)], model_name=self._model_name)
+                return ModelResponse(
+                    parts=[ToolCallPart(output_tool.name, response_args)],
+                    model_name=self._model_name,
+                    id=None,
+                    finish_reason=None,
+                )
 
 
 @dataclass