-
Notifications
You must be signed in to change notification settings - Fork 565
feat(integrations): pydantic-ai integration #4906
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
Open
constantinius
wants to merge
18
commits into
master
Choose a base branch
from
constantinius/feat/integration/pydantic-ai
base: master
Could not load branches
Branch not found: {{ refName }}
Loading
Could not load tags
Nothing to show
Loading
Are you sure you want to change the base?
Some commits from the old base branch may be removed from the timeline,
and old review comments may become outdated.
+1,935
−0
Open
Changes from all commits
Commits
Show all changes
18 commits
Select commit
Hold shift + click to select a range
f667f4b
feat(integrations): added initial Pydantic AI integration implementation
constantinius 8c26119
fix: model name lookup and message history preservation
constantinius 742d77f
feat: add support for run_stream
constantinius 76a7a67
fix: deduping code
constantinius 9d166ba
fix(integrations): add pydantic-ai as an optional dependency
constantinius 3337c19
fix(integrations): fixing span description -> name
constantinius bf3ce00
feat: add include_prompts for Pydantic AI integration
constantinius 2290901
Merge branch 'master' into constantinius/feat/integration/pydantic-ai
sentrivana c7e7ec2
Add pydantic ai to ci
sentrivana 1a2cb97
add pytest-asyncio
sentrivana b9f3357
Merge branch 'master' into constantinius/feat/integration/pydantic-ai
sentrivana 48af290
fix: mypy issues
constantinius 7263a77
fix: working in feedback
constantinius a94870e
Update sentry_sdk/integrations/pydantic_ai/spans/invoke_agent.py
constantinius 7f335f9
fix(integrations): cleanups and working in feedback
constantinius 0929e58
fix(integrations): type checking import missing
constantinius 4b40d2d
fix: significantly simplifying instrumentation
constantinius 87ebafc
feat: add support for MCP Tool calls as well
constantinius File filter
Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
There are no files selected for viewing
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
| Original file line number | Diff line number | Diff line change |
|---|---|---|
|
|
@@ -81,6 +81,7 @@ | |
| "openai-base", | ||
| "openai-notiktoken", | ||
| "openai_agents", | ||
| "pydantic_ai", | ||
| ], | ||
| "Cloud": [ | ||
| "aws_lambda", | ||
|
|
||
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
| Original file line number | Diff line number | Diff line change |
|---|---|---|
| @@ -0,0 +1,45 @@ | ||
| from sentry_sdk.integrations import DidNotEnable, Integration | ||
|
|
||
| from .patches import ( | ||
| _patch_agent_run, | ||
| _patch_graph_nodes, | ||
| _patch_model_request, | ||
| _patch_tool_execution, | ||
| ) | ||
|
|
||
| try: | ||
| import pydantic_ai | ||
| except ImportError: | ||
| raise DidNotEnable("pydantic-ai not installed") | ||
|
|
||
|
|
||
| class PydanticAIIntegration(Integration): | ||
| identifier = "pydantic_ai" | ||
| origin = f"auto.ai.{identifier}" | ||
|
|
||
| def __init__(self, include_prompts=True): | ||
| # type: (bool) -> None | ||
| """ | ||
| Initialize the Pydantic AI integration. | ||
|
|
||
| Args: | ||
| include_prompts: Whether to include prompts and messages in span data. | ||
| Requires send_default_pii=True. Defaults to True. | ||
| """ | ||
| self.include_prompts = include_prompts | ||
|
|
||
| @staticmethod | ||
| def setup_once(): | ||
| # type: () -> None | ||
| """ | ||
| Set up the pydantic-ai integration. | ||
|
|
||
| This patches the key methods in pydantic-ai to create Sentry spans for: | ||
| - Agent invocations (Agent.run methods) | ||
| - Model requests (AI client calls) | ||
| - Tool executions | ||
| """ | ||
| _patch_agent_run() | ||
| _patch_graph_nodes() | ||
| _patch_model_request() | ||
| _patch_tool_execution() |
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
| Original file line number | Diff line number | Diff line change |
|---|---|---|
| @@ -0,0 +1 @@ | ||
| SPAN_ORIGIN = "auto.ai.pydantic_ai" |
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
| Original file line number | Diff line number | Diff line change |
|---|---|---|
| @@ -0,0 +1,4 @@ | ||
| from .agent_run import _patch_agent_run # noqa: F401 | ||
| from .graph_nodes import _patch_graph_nodes # noqa: F401 | ||
| from .model_request import _patch_model_request # noqa: F401 | ||
| from .tools import _patch_tool_execution # noqa: F401 |
238 changes: 238 additions & 0 deletions
238
sentry_sdk/integrations/pydantic_ai/patches/agent_run.py
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
| Original file line number | Diff line number | Diff line change |
|---|---|---|
| @@ -0,0 +1,238 @@ | ||
| from functools import wraps | ||
|
|
||
| import sentry_sdk | ||
| from sentry_sdk.tracing_utils import set_span_errored | ||
| from sentry_sdk.utils import event_from_exception | ||
|
|
||
| from ..spans import invoke_agent_span, update_invoke_agent_span | ||
|
|
||
| from typing import TYPE_CHECKING | ||
| from pydantic_ai.agent import Agent | ||
|
|
||
| if TYPE_CHECKING: | ||
| from typing import Any, Callable, Optional | ||
|
|
||
|
|
||
| def _capture_exception(exc): | ||
| # type: (Any) -> None | ||
| set_span_errored() | ||
|
|
||
| event, hint = event_from_exception( | ||
| exc, | ||
| client_options=sentry_sdk.get_client().options, | ||
| mechanism={"type": "pydantic_ai", "handled": False}, | ||
| ) | ||
| sentry_sdk.capture_event(event, hint=hint) | ||
|
|
||
|
|
||
| class _StreamingContextManagerWrapper: | ||
| """Wrapper for streaming methods that return async context managers.""" | ||
|
|
||
| def __init__( | ||
| self, | ||
| agent, | ||
| original_ctx_manager, | ||
| user_prompt, | ||
| model, | ||
| model_settings, | ||
| is_streaming=True, | ||
| ): | ||
| # type: (Any, Any, Any, Any, Any, bool) -> None | ||
| self.agent = agent | ||
| self.original_ctx_manager = original_ctx_manager | ||
| self.user_prompt = user_prompt | ||
| self.model = model | ||
| self.model_settings = model_settings | ||
| self.is_streaming = is_streaming | ||
| self._isolation_scope = None # type: Any | ||
| self._span = None # type: Optional[sentry_sdk.tracing.Span] | ||
| self._result = None # type: Any | ||
|
|
||
| async def __aenter__(self): | ||
| # type: () -> Any | ||
| # Set up isolation scope and invoke_agent span | ||
| self._isolation_scope = sentry_sdk.isolation_scope() | ||
| self._isolation_scope.__enter__() | ||
|
|
||
| # Store agent reference and streaming flag | ||
| sentry_sdk.get_current_scope().set_context( | ||
| "pydantic_ai_agent", {"_agent": self.agent, "_streaming": self.is_streaming} | ||
| ) | ||
|
|
||
| # Create invoke_agent span (will be closed in __aexit__) | ||
| self._span = invoke_agent_span( | ||
| self.user_prompt, self.agent, self.model, self.model_settings | ||
| ) | ||
| self._span.__enter__() | ||
|
|
||
| # Enter the original context manager | ||
| result = await self.original_ctx_manager.__aenter__() | ||
| self._result = result | ||
| return result | ||
|
|
||
| async def __aexit__(self, exc_type, exc_val, exc_tb): | ||
| # type: (Any, Any, Any) -> None | ||
| try: | ||
| # Exit the original context manager first | ||
| await self.original_ctx_manager.__aexit__(exc_type, exc_val, exc_tb) | ||
|
|
||
| # Update span with output if successful | ||
| if exc_type is None and self._result and hasattr(self._result, "output"): | ||
| output = ( | ||
| self._result.output if hasattr(self._result, "output") else None | ||
| ) | ||
| update_invoke_agent_span(self._span, output) | ||
| finally: | ||
| # Clean up invoke span | ||
| if self._span: | ||
| self._span.__exit__(exc_type, exc_val, exc_tb) | ||
|
|
||
| # Clean up isolation scope | ||
| if self._isolation_scope: | ||
| self._isolation_scope.__exit__(exc_type, exc_val, exc_tb) | ||
|
|
||
|
|
||
| def _create_run_wrapper(original_func, is_streaming=False): | ||
| # type: (Callable[..., Any], bool) -> Callable[..., Any] | ||
| """ | ||
| Wraps the Agent.run method to create an invoke_agent span. | ||
|
|
||
| Args: | ||
| original_func: The original run method | ||
| is_streaming: Whether this is a streaming method (for future use) | ||
| """ | ||
|
|
||
| @wraps(original_func) | ||
| async def wrapper(self, *args, **kwargs): | ||
| # type: (Any, *Any, **Any) -> Any | ||
| # Isolate each workflow so that when agents are run in asyncio tasks they | ||
| # don't touch each other's scopes | ||
| with sentry_sdk.isolation_scope(): | ||
| # Store agent reference and streaming flag in Sentry scope for access in nested spans | ||
| # We store the full agent to allow access to tools and system prompts | ||
| sentry_sdk.get_current_scope().set_context( | ||
| "pydantic_ai_agent", {"_agent": self, "_streaming": is_streaming} | ||
| ) | ||
|
|
||
| # Extract parameters for the span | ||
| user_prompt = kwargs.get("user_prompt") or (args[0] if args else None) | ||
| model = kwargs.get("model") | ||
| model_settings = kwargs.get("model_settings") | ||
|
|
||
| # Create invoke_agent span | ||
| with invoke_agent_span(user_prompt, self, model, model_settings) as span: | ||
| try: | ||
| result = await original_func(self, *args, **kwargs) | ||
|
|
||
| # Update span with output | ||
| output = result.output if hasattr(result, "output") else None | ||
| update_invoke_agent_span(span, output) | ||
|
|
||
| return result | ||
| except Exception as exc: | ||
| _capture_exception(exc) | ||
| raise exc from None | ||
|
|
||
| return wrapper | ||
|
|
||
|
|
||
| def _create_run_sync_wrapper(original_func): | ||
| # type: (Callable[..., Any]) -> Callable[..., Any] | ||
| """ | ||
| Wraps the Agent.run_sync method - no span needed as it delegates to run(). | ||
|
|
||
| Note: run_sync just calls self.run() via run_until_complete, so the | ||
| invoke_agent span will be created by the run() wrapper. | ||
| """ | ||
|
|
||
| @wraps(original_func) | ||
| def wrapper(self, *args, **kwargs): | ||
| # type: (Any, *Any, **Any) -> Any | ||
| # Just call the original function - it will call run() which has the instrumentation | ||
| try: | ||
| result = original_func(self, *args, **kwargs) | ||
| return result | ||
| except Exception as exc: | ||
| _capture_exception(exc) | ||
| raise exc from None | ||
|
|
||
| return wrapper | ||
|
|
||
|
|
||
| def _create_streaming_wrapper(original_func): | ||
| # type: (Callable[..., Any]) -> Callable[..., Any] | ||
| """ | ||
| Wraps run_stream method that returns an async context manager. | ||
| """ | ||
|
|
||
| @wraps(original_func) | ||
| def wrapper(self, *args, **kwargs): | ||
| # type: (Any, *Any, **Any) -> Any | ||
| # Extract parameters for the span | ||
| user_prompt = kwargs.get("user_prompt") or (args[0] if args else None) | ||
| model = kwargs.get("model") | ||
| model_settings = kwargs.get("model_settings") | ||
|
|
||
| # Call original function to get the context manager | ||
| original_ctx_manager = original_func(self, *args, **kwargs) | ||
|
|
||
| # Wrap it with our instrumentation | ||
| return _StreamingContextManagerWrapper( | ||
| agent=self, | ||
| original_ctx_manager=original_ctx_manager, | ||
| user_prompt=user_prompt, | ||
| model=model, | ||
| model_settings=model_settings, | ||
| is_streaming=True, | ||
| ) | ||
|
|
||
| return wrapper | ||
|
|
||
|
|
||
| def _create_streaming_events_wrapper(original_func): | ||
| # type: (Callable[..., Any]) -> Callable[..., Any] | ||
| """ | ||
| Wraps run_stream_events method - no span needed as it delegates to run(). | ||
|
|
||
| Note: run_stream_events internally calls self.run() with an event_stream_handler, | ||
| so the invoke_agent span will be created by the run() wrapper. | ||
| """ | ||
|
|
||
| @wraps(original_func) | ||
| async def wrapper(self, *args, **kwargs): | ||
| # type: (Any, *Any, **Any) -> Any | ||
| # Just call the original generator - it will call run() which has the instrumentation | ||
| try: | ||
| async for event in original_func(self, *args, **kwargs): | ||
| yield event | ||
| except Exception as exc: | ||
| _capture_exception(exc) | ||
| raise exc from None | ||
|
|
||
| return wrapper | ||
|
|
||
|
|
||
| def _patch_agent_run(): | ||
| # type: () -> None | ||
| """ | ||
| Patches the Agent run methods to create spans for agent execution. | ||
|
|
||
| This patches both non-streaming (run, run_sync) and streaming | ||
| (run_stream, run_stream_events) methods. | ||
| """ | ||
|
|
||
| # Store original methods | ||
| original_run = Agent.run | ||
| original_run_sync = Agent.run_sync | ||
| original_run_stream = Agent.run_stream | ||
| original_run_stream_events = Agent.run_stream_events | ||
|
|
||
| # Wrap and apply patches for non-streaming methods | ||
| Agent.run = _create_run_wrapper(original_run, is_streaming=False) # type: ignore | ||
| Agent.run_sync = _create_run_sync_wrapper(original_run_sync) # type: ignore | ||
|
|
||
| # Wrap and apply patches for streaming methods | ||
| Agent.run_stream = _create_streaming_wrapper(original_run_stream) # type: ignore | ||
| Agent.run_stream_events = _create_streaming_events_wrapper( # type: ignore[method-assign] | ||
| original_run_stream_events | ||
| ) |
Oops, something went wrong.
Oops, something went wrong.
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Bug: AI Integration Auto-Enablement Issue
The
PydanticAIIntegrationwas added to_MIN_VERSIONSbut is missing from_AUTO_ENABLING_INTEGRATIONS. This prevents it from being automatically enabled whenpydantic_aiis present, which is inconsistent with other AI integrations and itsauto.ai.pydantic_aiorigin.