Merge branch 'main' into toolsets

# Conflicts:
#	pydantic_ai_slim/pydantic_ai/_agent_graph.py
#	tests/test_streaming.py

Author: DouweM

docs/a2a.md

@@ -27,7 +27,7 @@ The library is designed to be used with any agentic framework, and is **not excl
 
 ### Design
 
-**FastA2A** is built on top of [Starlette](https://starlette.io), which means it's fully compatible with any ASGI server.
+**FastA2A** is built on top of [Starlette](https://www.starlette.io), which means it's fully compatible with any ASGI server.
 
 Given the nature of the A2A protocol, it's important to understand the design before using it, as a developer
 you'll need to provide some components:
@@ -66,7 +66,7 @@ pip/uv-add fasta2a
 
 The only dependencies are:
 
-- [starlette](https://starlette.io): to expose the A2A server as an [ASGI application](https://asgi.readthedocs.io/en/latest/)
+- [starlette](https://www.starlette.io): to expose the A2A server as an [ASGI application](https://asgi.readthedocs.io/en/latest/)
 - [pydantic](https://pydantic.dev): to validate the request/response messages
 - [opentelemetry-api](https://opentelemetry-python.readthedocs.io/en/latest): to provide tracing capabilities
 

docs/mcp/client.md

@@ -31,7 +31,7 @@ Examples of both are shown below; [mcp-run-python](run-python.md) is used as the
 !!! note
     [`MCPServerSSE`][pydantic_ai.mcp.MCPServerSSE] requires an MCP server to be running and accepting HTTP connections before calling [`agent.run_toolsets()`][pydantic_ai.Agent.run_toolsets]. Running the server is not managed by PydanticAI.
 
-The name "HTTP" is used since this implemented will be adapted in future to use the new
+The name "HTTP" is used since this implementation will be adapted in future to use the new
 [Streamable HTTP](https://github.com/modelcontextprotocol/specification/pull/206) currently in development.
 
 Before creating the SSE client, we need to run the server (docs [here](run-python.md)):
@@ -372,7 +372,7 @@ async def main():
 
 _(This example is complete, it can be run "as is" with Python 3.10+)_
 
-You can disallow sampling by settings [`allow_sampling=False`][pydantic_ai.mcp.MCPServerStdio.allow_sampling] when creating the server reference, e.g.:
+You can disallow sampling by setting [`allow_sampling=False`][pydantic_ai.mcp.MCPServerStdio.allow_sampling] when creating the server reference, e.g.:
 
 ```python {title="sampling_disallowed.py" hl_lines="6" py="3.10"}
 from pydantic_ai.mcp import MCPServerStdio

pydantic_ai_slim/pydantic_ai/_agent_graph.py

@@ -608,7 +608,7 @@ async def process_function_tools(  # noqa: C901
                     content='Output tool not used - a final result was already processed.',
                     tool_call_id=call.tool_call_id,
                 )
-                yield _messages.FunctionToolResultEvent(part, tool_call_id=call.tool_call_id)
+                yield _messages.FunctionToolResultEvent(part)
 
             parts.append(part)
         else:
@@ -621,7 +621,7 @@ async def process_function_tools(  # noqa: C901
                 ctx.state.increment_retries(ctx.deps.max_result_retries, e)
                 yield _messages.FunctionToolCallEvent(call)
                 parts.append(e.tool_retry)
-                yield _messages.FunctionToolResultEvent(e.tool_retry, tool_call_id=call.tool_call_id)
+                yield _messages.FunctionToolResultEvent(e.tool_retry)
             else:
                 part = _messages.ToolReturnPart(
                     tool_name=call.tool_name,
@@ -682,7 +682,7 @@ async def process_function_tools(  # noqa: C901
                 for task in done:
                     index = tasks.index(task)
                     tool_result = task.result()
-                    yield _messages.FunctionToolResultEvent(tool_result, tool_call_id=tool_result.tool_call_id)
+                    yield _messages.FunctionToolResultEvent(tool_result)
 
                     if isinstance(tool_result, _messages.RetryPromptPart):
                         results_by_index[index] = tool_result

pydantic_ai_slim/pydantic_ai/messages.py

@@ -11,7 +11,7 @@
 import pydantic
 import pydantic_core
 from opentelemetry._events import Event  # pyright: ignore[reportPrivateImportUsage]
-from typing_extensions import TypeAlias
+from typing_extensions import TypeAlias, deprecated
 
 from . import _utils
 from ._utils import (
@@ -501,7 +501,10 @@ class RetryPromptPart:
     def model_response(self) -> str:
         """Return a string message describing why the retry is requested."""
         if isinstance(self.content, str):
-            description = self.content
+            if self.tool_name is None:
+                description = f'Validation feedback:\n{self.content}'
+            else:
+                description = self.content
         else:
             json_errors = error_details_ta.dump_json(self.content, exclude={'__all__': {'ctx'}}, indent=2)
             description = f'{len(self.content)} validation errors: {json_errors.decode()}'
@@ -1009,10 +1012,16 @@ class FunctionToolCallEvent:
     """Event type identifier, used as a discriminator."""
 
     @property
-    def call_id(self) -> str:
-        """An ID used for matching details about the call to its result. If present, defaults to the part's tool_call_id."""
+    def tool_call_id(self) -> str:
+        """An ID used for matching details about the call to its result."""
         return self.part.tool_call_id
 
+    @property
+    @deprecated('`call_id` is deprecated, use `tool_call_id` instead.')
+    def call_id(self) -> str:
+        """An ID used for matching details about the call to its result."""
+        return self.part.tool_call_id  # pragma: no cover
+
     __repr__ = _utils.dataclasses_no_defaults_repr
 
 
@@ -1022,11 +1031,14 @@ class FunctionToolResultEvent:
 
     result: ToolReturnPart | RetryPromptPart
     """The result of the call to the function tool."""
-    tool_call_id: str
-    """An ID used to match the result to its original call."""
     event_kind: Literal['function_tool_result'] = 'function_tool_result'
     """Event type identifier, used as a discriminator."""
 
+    @property
+    def tool_call_id(self) -> str:
+        """An ID used to match the result to its original call."""
+        return self.result.tool_call_id
+
     __repr__ = _utils.dataclasses_no_defaults_repr
 
 

pydantic_ai_slim/pydantic_ai/models/openai.py

@@ -61,6 +61,7 @@
     from openai.types.chat.chat_completion_content_part_image_param import ImageURL
     from openai.types.chat.chat_completion_content_part_input_audio_param import InputAudio
     from openai.types.chat.chat_completion_content_part_param import File, FileFile
+    from openai.types.chat.chat_completion_prediction_content_param import ChatCompletionPredictionContentParam
     from openai.types.responses import ComputerToolParam, FileSearchToolParam, WebSearchToolParam
     from openai.types.responses.response_input_param import FunctionCallOutput, Message
     from openai.types.shared import ReasoningEffort
@@ -126,6 +127,12 @@ class OpenAIModelSettings(ModelSettings, total=False):
     For more information, see [OpenAI's service tiers documentation](https://platform.openai.com/docs/api-reference/chat/object#chat/object-service_tier).
     """
 
+    openai_prediction: ChatCompletionPredictionContentParam
+    """Enables [predictive outputs](https://platform.openai.com/docs/guides/predicted-outputs).
+
+    This feature is currently only supported for some OpenAI models.
+    """
+
 
 class OpenAIResponsesModelSettings(OpenAIModelSettings, total=False):
     """Settings used for an OpenAI Responses model request.
@@ -320,6 +327,7 @@ async def _completions_create(
                 reasoning_effort=model_settings.get('openai_reasoning_effort', NOT_GIVEN),
                 user=model_settings.get('openai_user', NOT_GIVEN),
                 service_tier=model_settings.get('openai_service_tier', NOT_GIVEN),
+                prediction=model_settings.get('openai_prediction', NOT_GIVEN),
                 temperature=sampling_settings.get('temperature', NOT_GIVEN),
                 top_p=sampling_settings.get('top_p', NOT_GIVEN),
                 presence_penalty=sampling_settings.get('presence_penalty', NOT_GIVEN),

tests/models/test_bedrock.py

@@ -413,8 +413,7 @@ async def get_temperature(city: str) -> str:
                     content='30°C',
                     tool_call_id='tooluse_lAG_zP8QRHmSYOwZzzaCqA',
                     timestamp=IsDatetime(),
-                ),
-                tool_call_id='tooluse_lAG_zP8QRHmSYOwZzzaCqA',
+                )
             ),
             PartStartEvent(index=0, part=TextPart(content='The')),
             FinalResultEvent(tool_name=None, tool_call_id=None),

tests/models/test_google.py

@@ -375,8 +375,7 @@ async def get_temperature(city: str) -> str:
             FunctionToolResultEvent(
                 result=ToolReturnPart(
                     tool_name='get_capital', content='Paris', tool_call_id=IsStr(), timestamp=IsDatetime()
-                ),
-                tool_call_id=IsStr(),
+                )
             ),
             PartStartEvent(
                 index=0,
@@ -386,8 +385,7 @@ async def get_temperature(city: str) -> str:
             FunctionToolResultEvent(
                 result=ToolReturnPart(
                     tool_name='get_temperature', content='30°C', tool_call_id=IsStr(), timestamp=IsDatetime()
-                ),
-                tool_call_id=IsStr(),
+                )
             ),
             PartStartEvent(index=0, part=TextPart(content='The temperature in Paris')),
             FinalResultEvent(tool_name=None, tool_call_id=None),

tests/models/test_instrumented.py

@@ -240,6 +240,7 @@ async def test_instrumented_model(capfire: CaptureLogfire):
             {
                 'body': {
                     'content': """\
+Validation feedback:
 retry_prompt2
 
 Fix the errors and try again.\
@@ -619,6 +620,7 @@ async def test_instrumented_model_attributes_mode(capfire: CaptureLogfire):
                                 {
                                     'event.name': 'gen_ai.user.message',
                                     'content': """\
+Validation feedback:
 retry_prompt2
 
 Fix the errors and try again.\

tests/test_agent.py

@@ -303,7 +303,7 @@ def return_tuple(_: list[ModelMessage], info: AgentInfo) -> ModelResponse:
                 parts=[
                     ToolCallPart(tool_name='final_result', args='{"response": ["foo", "bar"]}', tool_call_id=IsStr())
                 ],
-                usage=Usage(requests=1, request_tokens=72, response_tokens=8, total_tokens=80),
+                usage=Usage(requests=1, request_tokens=74, response_tokens=8, total_tokens=82),
                 model_name='function:return_tuple:',
                 timestamp=IsNow(tz=timezone.utc),
             ),
@@ -942,7 +942,7 @@ def call_tool(messages: list[ModelMessage], info: AgentInfo) -> ModelResponse:
             ),
             ModelResponse(
                 parts=[TextPart(content='Mexico City')],
-                usage=Usage(requests=1, request_tokens=68, response_tokens=5, total_tokens=73),
+                usage=Usage(requests=1, request_tokens=70, response_tokens=5, total_tokens=75),
                 model_name='function:call_tool:',
                 timestamp=IsDatetime(),
             ),
@@ -1602,7 +1602,7 @@ def call_tool(messages: list[ModelMessage], info: AgentInfo) -> ModelResponse:
             ),
             ModelResponse(
                 parts=[TextPart(content='{"city": "Mexico City"}')],
-                usage=Usage(requests=1, request_tokens=68, response_tokens=11, total_tokens=79),
+                usage=Usage(requests=1, request_tokens=70, response_tokens=11, total_tokens=81),
                 model_name='function:call_tool:',
                 timestamp=IsDatetime(),
             ),

tests/test_streaming.py

@@ -997,7 +997,6 @@ def known_tool(x: int) -> int:
                     tool_call_id=IsStr(),
                     timestamp=IsNow(tz=timezone.utc),
                 ),
-                tool_call_id=IsStr(),
             ),
             FunctionToolResultEvent(
                 result=ToolReturnPart(
@@ -1006,7 +1005,6 @@ def known_tool(x: int) -> int:
                     tool_call_id=IsStr(),
                     timestamp=IsNow(tz=timezone.utc),
                 ),
-                tool_call_id=IsStr(),
             ),
         ]
     )
@@ -1057,8 +1055,7 @@ def call_final_result_with_bad_data(messages: list[ModelMessage], info: AgentInf
                     tool_name='final_result',
                     tool_call_id=IsStr(),
                     timestamp=IsNow(tz=timezone.utc),
-                ),
-                tool_call_id=IsStr(),
+                )
             ),
         ]
     )
@@ -1087,6 +1084,23 @@ class CityLocation(BaseModel):
         assert result.is_complete
 
 
+def test_function_tool_event_tool_call_id_properties():
+    """Ensure that the `tool_call_id` property on function tool events mirrors the underlying part's ID."""
+    # Prepare a ToolCallPart with a fixed ID
+    call_part = ToolCallPart(tool_name='sample_tool', args={'a': 1}, tool_call_id='call_id_123')
+    call_event = FunctionToolCallEvent(part=call_part)
+
+    # The event should expose the same `tool_call_id` as the part
+    assert call_event.tool_call_id == call_part.tool_call_id == 'call_id_123'
+
+    # Prepare a ToolReturnPart with a fixed ID
+    return_part = ToolReturnPart(tool_name='sample_tool', content='ok', tool_call_id='return_id_456')
+    result_event = FunctionToolResultEvent(result=return_part)
+
+    # The event should expose the same `tool_call_id` as the result part
+    assert result_event.tool_call_id == return_part.tool_call_id == 'return_id_456'
+
+
 async def test_deferred_tool():
     agent = Agent(TestModel(), output_type=[str, DeferredToolCalls])