Fixes#5246
When a custom @executor transforms agent output and sends a plain str,
the downstream AgentExecutor.from_str handler loses the full conversation
context. This adds a with_text() helper that creates a new
AgentExecutorResponse with replaced text while preserving the prior
conversation chain, so AgentExecutor.from_response is invoked instead.
- Add with_text(text) method to AgentExecutorResponse dataclass
- Add 3 regression tests in test_full_conversation.py
Co-authored-by: Evan Mattson <35585003+moonbox3@users.noreply.github.com>
The local MCP server can't be used for hosted tools tests because
Anthropic's backend needs to reach the MCP URL from their infrastructure
(not localhost on the CI runner). Revert to learn.microsoft.com/api/mcp
but catch BadRequestError, InternalServerError, APIConnectionError, and
APITimeoutError and pytest.skip so upstream outages don't block the
merge queue.
* Python: use local MCP server for hosted tools test and broaden image assertion
The hosted tools integration test was hitting rate limits on the external
learn.microsoft.com MCP server, causing persistent failures that retries
couldn't recover from. Switch to the local MCP server already spun up in
CI via LOCAL_MCP_URL, skipping when the env var isn't set.
Also broaden the image description assertion to accept common synonyms
(cottage, mansion, villa, etc.) instead of just "house", since the model
legitimately uses varied vocabulary for the same image.
* Address review feedback: validate LOCAL_MCP_URL scheme and use word boundaries
- Skip hosted tools test when LOCAL_MCP_URL lacks http/https scheme,
matching the pattern used in test_mcp.py.
- Use regex word boundaries for image assertion to avoid false matches
like "villain" matching "villa".
* Add agent-framework-gemini package
* Add AGENTS.md documentation
* Add LICENSE file
* Add README.md for agent-framework-gemini package
* Add Google Gemini API keys to .env.example
* Add Google Gemini chat client implementation
* Add tests for GeminiChatClient
* Add Google Gemini agent examples
* Fix client inheritence order
* Update Gemini agent examples
* Update documentation
* Update AGENTS.md
* Add tests for JSON string handling in GeminiChatClient
* Add final response assembly test in GeminiChatClient
* Add tests for handling empty candidates in GeminiChatClient
* Improve Pydantic response handling in GeminiChatClient
* Add tests for function result resolution and callable tool normalization
* Add test for function result resolution when call_id is generated
* Refactor GeminiChatClient to correct inheritance order
Also updates constructor parameter order for environment file handling
* Enhance documentation and clarify Gemini-specific fields
* Update ThinkingConfig with new attributes and type
* Add tests for GoogleSearch and GoogleMaps configs
* Suppress valid-type mypy error on GeminiChatOptionsT
* Move service_url method near overrides
* Order _prepare_config kwargs by base then Gemini-specific
* Use FunctionCallingConfigMode for clarity and type safety
* Fix code_execution doc
* Add agent-framework-gemini to project dependencies
* Remove package from core dependencies
Initial release will be done without agent-framework-gemini in
core[all].
* Move integration tests into one file
* Remove __init__.py file from gemini tests directory
* Introduce RawGeminiChatClient as lightweight chat client
Updated GeminiChatClient to inherit from RawGeminiChatClient, maintaining full functionality with added features.
* Updated variable names from `model_id` to `model`
Across the codebase, including environment variables and client initialization. Adjusted related tests and sample scripts to reflect this change, ensuring consistency in the usage of the Gemini model identifier.
* Update AGENTS.md
* Update Gemini package to alpha status
* Fix docstrings in Gemini tests
* Change 'model_id' to 'model' in response handling
* Fix model property change in response handling
* Add built-in tool factory methods to Gemini client
Replaces boolean tool options (code_execution, google_search_grounding,
google_maps_grounding) with static factory methods that return types.Tool
objects: get_code_interpreter_tool, get_web_search_tool, get_mcp_tool,
get_file_search_tool, and get_maps_grounding_tool.
Simplifies _prepare_tools to a single translation boundary between
FunctionTool (framework) and FunctionDeclaration (Gemini API), with
types.Tool objects passed through unchanged.
* Surface code execution parts
_parse_parts now maps executable_code and code_execution_result
parts to text Content objects so callers can see the code run
and its output. Unknown part types log at debug level rather than
being silently dropped.
* Update Gemini client documentation
* Unify Gemini model name
Co-authored-by: Eduard van Valkenburg <eavanvalkenburg@users.noreply.github.com>
* Update Agent Framework core version
Co-authored-by: Eduard van Valkenburg <eavanvalkenburg@users.noreply.github.com>
* Add Python 3.14 in classifiers
* Replace kwargs with parameters in tool factories
* Refactor chat options handling in Gemini client
* Add tests for handling unknown and consumed keys
* Update Gemini documentation
Now reflects new options and built-in tool factory methods
* Change build system to flit
Co-authored-by: Eduard van Valkenburg <eavanvalkenburg@users.noreply.github.com>
* Fix build system in pyproject.toml
* Fix type checking for generate_content_stream
---------
Co-authored-by: Eduard van Valkenburg <eavanvalkenburg@users.noreply.github.com>
* Python: Skip get_final_response in OTel _finalize_stream when stream errored
When a streaming error occurs, _finalize_stream (a cleanup hook registered by
AgentTelemetryLayer) was unconditionally calling get_final_response(), which
triggers all registered result hooks including after_run context providers.
This caused providers to fire incorrectly on error paths.
Guard against this by checking result_stream._consumed: True only after
StopAsyncIteration (normal completion), False when an exception was raised.
The fix applies to both the chat client and agent telemetry layers.
Closes#5231
* Python: Expose consumed/stream_error on ResponseStream and capture error in OTel span
Address Copilot review feedback on #5232:
- Add `_stream_error: Exception | None` to ResponseStream, set in __anext__'s
except branch so cleanup hooks can inspect the failure.
- Expose public `consumed` and `stream_error` properties to avoid coupling
observability.py to private stream internals.
- Update both _finalize_stream closures (chat and agent layers) to use the
public properties and call capture_exception() with the stream error before
returning early, ensuring the OTel span records the failure rather than
closing silently.
* Python: Address Copilot review feedback on stream error handling
- Use stream_error is not None as the guard in _finalize_stream instead of
not consumed, so the early-return path is keyed precisely to actual errors
rather than any non-normal completion state.
- Clear _stream_error after _run_cleanup_hooks() completes to avoid retaining
the exception traceback (and any large object graphs it references) on the
stream instance beyond the cleanup phase.
* Python: Remove consumed/stream_error properties, use private attrs directly
Per review feedback: since observability.py and _types.py are in the same
package, accessing _stream_error directly is fine and the public properties
are unnecessary.
* Python: Fix Pyright reportPrivateUsage via inline ignore comments
Keep _stream_error private (consistent with rest of ResponseStream), and
suppress reportPrivateUsage at the call sites in observability.py with
inline pyright: ignore comments — access is intentional within the package.
* AG-UI deterministic state updates from tool results
* fix(ag-ui): address PR #5201 review comments
1. Add missing AGUIEventConverter, AGUIHttpService, __version__ to
_IMPORTS in core ag_ui lazy-export list to match the .pyi stub.
2. Coalesce predictive and deterministic state snapshots into a single
StateSnapshotEvent when both mechanisms are active on the same tool
result, reducing redundant snapshot traffic.
3. Update state_update() docstring to clarify that a predictive snapshot
may be emitted before the deterministic one when predict_state_config
is active.
Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
---------
Co-authored-by: Copilot <copilot@github.com>
Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
* Fix HandoffBuilder dropping function-level middleware when cloning agents (#5173)
_clone_chat_agent() was using agent.agent_middleware (agent-level only)
instead of agent.middleware (all types), which silently dropped any
function middleware registered on the original agent.
Changed to use agent.middleware to preserve all middleware types
(agent, function, and chat) during cloning.
Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
* Python: Fix HandoffBuilder dropping function-level middleware when cloning agents
Fixes#5173
* Fix false-positive middleware regression test (#5173)
The test used isinstance(m, FunctionMiddleware) which matched
_AutoHandoffMiddleware (always appended during build) instead of the
user's @function_middleware decorator. Assert directly that
tracking_middleware is present in the cloned agent's middleware list.
Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
* Address review feedback for #5173: Python: [Bug]: HandoffBuilder drops function-level middleware when cloning agents
---------
Co-authored-by: Copilot <copilot@github.com>
Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
* Python: Add allowed_checkpoint_types support to CosmosCheckpointStorage (#5200)
Add allowed_checkpoint_types parameter to CosmosCheckpointStorage for
parity with FileCheckpointStorage. This ensures both providers use the
same restricted pickle deserialization by default.
Changes:
- Accept allowed_checkpoint_types kwarg in __init__, stored as frozenset
- Convert _document_to_checkpoint from @staticmethod to instance method
- Forward allowed_types to decode_checkpoint_value on all load paths
- Update class docstring to describe the new parameter
- Add tests covering built-in safe types, app type opt-in/blocking,
and all load paths (load, list_checkpoints, get_latest)
- Add changelog entry noting the breaking behavior change
BREAKING CHANGE: CosmosCheckpointStorage now uses restricted pickle
deserialization by default. Checkpoints containing application-defined
types will require passing those types via allowed_checkpoint_types.
Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
* Python: Add `allowed_checkpoint_types` support to `CosmosCheckpointStorage` for parity with `FileCheckpointStorage`
Fixes#5200
* Address PR review: add pickle security warning and fix docstring examples
- Reintroduce explicit security warning about pickle deserialization risks
- Convert Example:: block to .. code-block:: python with imports for
consistency with other docstring examples
- Note: PR title should be updated to include [BREAKING] prefix per
changelog convention (comment #3, requires GitHub UI change)
Fixes#5200
Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
---------
Co-authored-by: Copilot <copilot@github.com>
Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
* Fix python-feature-lifecycle skill YAML frontmatter
Remove copyright comment that preceded the YAML frontmatter delimiter,
which prevented the skill from loading. The --- block must be the very
first line of SKILL.md.
Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
* fix: update broken eslint-react plugin links in devui README
The upstream eslint-react repo moved plugins from packages/plugins/
to the top-level plugins/ directory, causing 404 errors detected by
linkspector CI.
Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
---------
Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
* Bump Python version to 1.1.0 for a release
* Fix changelog
* 1.0.1 instead of 1.1.0
* Update CHANGELOG.md
* update version and changelog
* Bump lower bounds
* Python: Migrate GitHub Copilot package to SDK 0.2.x
Replace all imports from the non-existent copilot.types module with
correct SDK 0.2.x module paths (copilot.session, copilot.client,
copilot.tools, copilot.generated.session_events). Fix PermissionRequest
attribute access from dict-style .get() to dataclass attribute access.
Add OTel telemetry support to Copilot samples via configure_otel_providers
and document new telemetry environment variables in samples README.
* Python: Fix remaining copilot.types import in sample validation script
* Python: Include model in default_options for telemetry span attributes
* Python: Address review feedback on log_level and session kwargs typing
* Python: Scope PR to SDK 0.2.x migration only, remove net-new OTel features
- Remove RawGitHubCopilotAgent split and AgentTelemetryLayer inheritance
- Remove TelemetryConfig plumbing and OTLP/file telemetry settings
- Remove configure_otel_providers() calls from samples
- Remove telemetry env var rows from samples README
- Retain only: import path fixes, PermissionRequest attribute access fix,
log_level default fix, session kwargs typed fix, dependency pin
* Python: Update tests for SDK 0.2.x API changes
- SubprocessConfig replaces CopilotClientOptions dict
- create_session and resume_session now use keyword args
- send and send_and_wait take plain string prompt instead of MessageOptions
- on_permission_request is always required; deny-all fallback replaces omission
* Python: Pin github-copilot-sdk to >=0.2.0,<=0.2.0
Tighten the upper bound from <0.3.0 to <=0.2.0 to avoid pulling in 0.2.1+
which has breaking API changes relative to 0.2.0. The lower bound stays at
>=0.2.0 since this migration requires the 0.2.x import paths; 0.1.x would
fail at import time.
* Python: Pin github-copilot-sdk to >=0.2.1,<=0.2.1
---------
Co-authored-by: Evan Mattson <35585003+moonbox3@users.noreply.github.com>
* Harden Python checkpoint persistence defaults
Add RestrictedUnpickler to _checkpoint_encoding.py that limits which
types may be instantiated during pickle deserialization. By default
FileCheckpointStorage now uses the restricted unpickler, allowing only:
- Built-in Python value types (primitives, datetime, uuid, decimal,
collections, etc.)
- All agent_framework.* internal types
- Additional types specified via the new allowed_checkpoint_types
parameter on FileCheckpointStorage
This narrows the default type surface area for persisted checkpoints
while keeping framework-owned scenarios working without extra
configuration. Developers can extend the allowed set by passing
"module:qualname" strings to allowed_checkpoint_types.
The decode_checkpoint_value function retains backward-compatible
unrestricted behavior when called without the new allowed_types kwarg.
Fixes#4894
Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
* fix: resolve mypy no-any-return error in checkpoint encoding
Add explicit type annotation for super().find_class() return value
to satisfy mypy's no-any-return check.
Fixes#4894
Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
* Simplify find_class return in _RestrictedUnpickler (#4894)
Remove unnecessary intermediate variable and apply # noqa: S301 # nosec
directly on the super().find_class() call, matching the established
pattern used on the pickle.loads() call in the same file.
Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
* Address review feedback for #4894: Python: Harden Python checkpoint persistence defaults
* Restore # noqa: S301 on line 102 of _checkpoint_encoding.py (#4894)
The review feedback correctly identified that removing the # noqa: S301
suppression from the find_class return statement would cause a ruff S301
lint failure, since the project enables bandit ("S") rules. This
restores consistency with lines 82 and 246 in the same file.
Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
* Address review feedback for #4894: Python: Harden Python checkpoint persistence defaults
* Address PR review comments on checkpoint encoding (#4894)
- Move module docstring to proper position after __future__ import
- Fix find_class return type annotation to type[Any]
- Add missing # noqa: S301 pragma on find_class return
- Improve error message to reference both allowed_types param and
FileCheckpointStorage.allowed_checkpoint_types
- Add -> None return annotation to FileCheckpointStorage.__init__
- Replace tempfile.mktemp with TemporaryDirectory in test
- Replace contextlib.suppress with pytest.raises for precise assertion
- Remove unused contextlib import
Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
* Address PR #4941 review comments: fix docstring position and return type
- Move module docstring before 'from __future__' import so it populates
__doc__ (comment #4)
- Change find_class return annotation from type[Any] to type to avoid
misleading callers about non-type returns like copyreg._reconstructor
(comment #2)
Comments #1, #3, #5, #6, #7, #8 were already addressed in the current code.
Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
* Address review feedback for #4894: review comment fixes
* fix: use pickle.UnpicklingError in RestrictedUnpickler and improve docstring (#4894)
- Change _RestrictedUnpickler.find_class to raise pickle.UnpicklingError
instead of WorkflowCheckpointException, since it is pickle-level concern
that gets wrapped by the caller in _base64_to_unpickle.
- Remove now-unnecessary WorkflowCheckpointException re-raise in
_base64_to_unpickle (pickle.UnpicklingError is caught by the generic
except Exception handler and wrapped).
- Expand decode_checkpoint_value docstring to show a concrete example of
the module:qualname format with a user-defined class.
- Add regression test verifying find_class raises pickle.UnpicklingError.
Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
* fix: address PR #4941 review comments for checkpoint encoding
- Comment 1 (line 103): Already resolved in prior commit — _RestrictedUnpickler
now raises pickle.UnpicklingError instead of WorkflowCheckpointException.
- Comment 2 (line 140): Add concrete usage examples to decode_checkpoint_value
docstring showing both direct allowed_types usage and FileCheckpointStorage
allowed_checkpoint_types usage. Rename 'SafeState' to 'MyState' across all
docstrings for consistency, making it clear this is a user-defined class name.
Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
* fix: replace deprecated 'builtin' repo with pre-commit-hooks in pre-commit config
pre-commit 4.x no longer supports 'repo: builtin'. Merge those hooks into
the existing pre-commit-hooks repo entry.
Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
* style: apply pyupgrade formatting to docstring example
Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
* fix: resolve pre-commit hook paths for monorepo git root
The poe-check and bandit hooks referenced paths relative to python/
but pre-commit runs hooks from the git root (monorepo root). Fix
poe-check entry to cd into python/ first, and update bandit config
path to python/pyproject.toml.
Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
* Fix pre-commit config paths for prek --cd python execution
Revert bandit config path from 'python/pyproject.toml' to 'pyproject.toml'
and poe-check entry from explicit 'cd python' wrapper to direct invocation,
since prek --cd python already sets the working directory to python/.
Also apply ruff formatting fixes to cosmos checkpoint storage files.
Fixes#4894
Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
* fix: add builtins:getattr to checkpoint deserialization allowlist
Pickle uses builtins:getattr to reconstruct enum members (e.g.,
WorkflowMessage.type which is a MessageType enum). Without it in the
allowlist, checkpoint roundtrip tests fail with
WorkflowCheckpointException.
Fixes#4894
Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
* Address review feedback for #4894: review comment fixes
---------
Co-authored-by: Copilot <copilot@github.com>
Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
* Fix reasoning text done events duplicating streamed delta content (#5157)
The OpenAI Responses API sends both reasoning_text.delta (incremental
chunks) and reasoning_text.done (full accumulated text) events. The
chat client was emitting Content for both, causing ag-ui to append the
full done text onto already-accumulated delta text, producing
duplicated reasoning output.
Stop emitting Content for reasoning_text.done and
reasoning_summary_text.done events, matching how output_text.done is
already handled (not emitted). The deltas contain all the content;
the done event is redundant.
Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
* fix(openai): emit reasoning done content as fallback when no deltas observed (#5157)
Address PR review feedback:
- Track item_ids that received reasoning deltas via seen_reasoning_delta_item_ids set
- Emit content from done events only when no deltas were received for the
item_id, preventing silent content loss on stream resumption
- Add comment documenting code_interpreter done event asymmetry
- Replace redundant ag-ui test with deduplication-focused test
- Add integration test for delta+done sequence in OpenAI chat client tests
- Add fallback path tests for done events without preceding deltas
Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
* Address review feedback for #5157: Python: [Bug]: "type": "response.reasoning_text.delta" and "response.reasoning_text.done" both get exposed as "text_reasoning"
* Fix AG-UI reasoning streaming to use proper Start/End pattern (#5157)
_emit_text_reasoning now follows the same streaming pattern as _emit_text:
- Emits ReasoningStartEvent/ReasoningMessageStartEvent only on the first
delta for a given message_id
- Emits only ReasoningMessageContentEvent for subsequent deltas
- Defers ReasoningMessageEndEvent/ReasoningEndEvent until
_close_reasoning_block is called (on content type switch or end-of-run)
This produces the correct protocol pattern:
ReasoningStartEvent
ReasoningMessageStartEvent
ReasoningMessageContentEvent(delta1)
ReasoningMessageContentEvent(delta2)
ReasoningMessageEndEvent
ReasoningEndEvent
Instead of wrapping every delta in a full Start→End sequence.
Backward compatibility is preserved: calling _emit_text_reasoning without
a flow argument still produces the full sequence per call.
Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
* Fix import ordering lint error in AG-UI test file (#5157)
Move inline import of TextMessageContentEvent to the top-level import
block and ensure alphabetical ordering to satisfy ruff I001 rule.
Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
* Fix mypy error: rename loop variable to avoid type conflict with WorkflowEvent
The 'event' variable was already typed as WorkflowEvent[Any] from the
async for loop at line 590. Reusing it in the _close_reasoning_block
loop (which returns list[BaseEvent]) caused an incompatible assignment
error. Renamed to 'reasoning_evt' to avoid the conflict.
Fixes#5162
Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
* Address review feedback for #5157: review comment fixes
* narrow test result reporting to explicit pytest JUnit XML
* Fix test args
* Fix pytest-results-action in merge workflow and remove committed test artifacts
Apply the same JUnit XML fix from python-tests.yml to python-merge-tests.yml:
add --junitxml=pytest.xml to all test commands and narrow the results action
path from ./python/**.xml to ./python/pytest.xml. Also remove accidentally
committed pytest.xml and python-coverage.xml and add them to .gitignore.
---------
Co-authored-by: Copilot <copilot@github.com>
Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
_prepare_options() now removes tools, tool_choice, and parallel_tool_calls
from run_options after injecting agent_reference. The Foundry API rejects
requests containing both fields. FunctionTools are still invoked client-side
by the function invocation layer.
Fixes#5087
Co-authored-by: Matt Van Horn <455140+mvanhorn@users.noreply.github.com>
Co-authored-by: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Co-authored-by: Evan Mattson <35585003+moonbox3@users.noreply.github.com>
* Guard against empty text in _parse_structured_response_value (#5145)
When using response_format with background=True (Responses API), polling
an in-progress response produces empty text. _parse_structured_response_value
unconditionally passed this to model_validate_json/json.loads, causing
ValidationError or JSONDecodeError.
Add an early return of None when text is empty, matching the existing
guard for response_format=None. This allows .value to safely return None
for in-progress background responses.
Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
* Python: Fix `response_format` crash on background polling with empty text
Fixes#5145
---------
Co-authored-by: Copilot <copilot@github.com>
Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
* Raise clear handler registration error for unresolved TypeVar (#4943)
Detect unresolved TypeVar in message parameter annotations during handler
registration in both _validate_handler_signature (Executor) and
_validate_function_signature (FunctionExecutor). Raises a ValueError with
an actionable message recommending @handler(input=..., output=...) or
@executor(input=..., output=...) instead of letting TypeVar leak through
to a confusing TypeCompatibilityError during workflow edge validation.
Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
* Address review feedback for #4943: reorder checks and harden function executor
- Move TypeVar check before validate_workflow_context_annotation in
_executor.py so users see the more actionable error first
- Wrap get_type_hints in try/except in _function_executor.py matching
the defensive pattern in _executor.py
- Repurpose duplicate test to cover bounded TypeVar rejection
- Add test_function_executor_allows_concrete_types for test symmetry
Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
* Narrow get_type_hints except clause and add missing tests (#4943)
- Narrow `except Exception` to `except (NameError, AttributeError, RecursionError)`
in both _executor.py and _function_executor.py so unexpected failures in
get_type_hints are not silently swallowed.
- Add test_handler_unresolvable_annotation_raises to test_function_executor_future.py
exercising the except branch of get_type_hints in the function executor path.
- Add test_function_executor_rejects_bounded_typevar_in_message_annotation to
test_function_executor.py for parity with the Executor bounded TypeVar test.
Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
* Add error ordering test for TypeVar vs WorkflowContext priority (#4943)
Add test_handler_typevar_error_takes_priority_over_context_error to verify
that when a handler has both a TypeVar message and an unannotated ctx, the
TypeVar error is raised first (the more actionable issue).
Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
* Python: Fix image content serialization sending null file_id to Foundry API
Omit file_id from input_image dict when not present instead of including
it as null, which Azure AI Foundry's stricter schema validation rejects.
* Python: Fix Foundry API rejecting rich content in function_call_output
Azure AI Foundry does not support list-format output in function_call_output
items. Add SUPPORTS_RICH_FUNCTION_OUTPUT flag (default True) to
RawOpenAIChatClient, set to False in RawFoundryChatClient so Foundry
falls back to string output for tool results with images/files.
Also omit file_id from input_image dicts when not set, since Foundry
rejects explicit nulls.
* Python: Surface rich tool content as user message when Foundry lacks support
When SUPPORTS_RICH_FUNCTION_OUTPUT is False, image/file items from tool
results are injected as a follow-up user message so the model can still
process the visual content via Foundry's supported user message format.
* Xfail Foundry image integration test for the meantime
---------
Co-authored-by: Copilot <copilot@github.com>
Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
* Fix GitHubCopilotAgent not calling context provider hooks (#3984)
GitHubCopilotAgent accepted context_providers in its constructor but
never called before_run()/after_run() on them in _run_impl() or
_stream_updates(), silently ignoring all context providers.
Add _run_before_providers() helper to create SessionContext and invoke
before_run on each provider. Both _run_impl() and _stream_updates() now
run the full provider lifecycle: before_run before sending the prompt
(with provider instructions prepended) and after_run after receiving the
response. This follows the same pattern used by A2AAgent.
Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
* Python: Fix GitHubCopilotAgent to invoke context provider before_run/after_run hooks
Fixes#3984
* fix(#3984): address review feedback for context provider integration
- Build prompt from session_context.get_messages(include_input=True) so
provider-injected context_messages are included in both non-streaming
and streaming paths (review comments #1, #2)
- Preserve timeout in opts (use get instead of pop) so providers can
observe it via context.options (review comment #3)
- Eliminate streaming double-buffer: move after_run invocation to a
ResponseStream result_hook (matching Agent class pattern) instead of
maintaining a separate updates list in the generator (review comment #4)
- Improve _run_before_providers docstring
Add tests for:
- Context messages included in prompt (non-streaming + streaming)
- Error path: after_run NOT called when send_and_wait/streaming raises
- Multiple providers: forward before_run, reverse after_run ordering
- BaseHistoryProvider with load_messages=False is skipped
- Streaming after_run response contains aggregated updates
- Streaming with no updates still sets empty response
- Timeout preserved in session context options for providers
Note: _run_before_providers remains on GitHubCopilotAgent for now. A
follow-up PR should extract it to BaseAgent so subclasses can reuse it
without duplicating the provider iteration logic.
Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
* Address review feedback for #3984: Python: [Bug]: GitHubCopilotAgent Memory Example
* refactor(#3984): promote _run_before_providers to BaseAgent
Move _run_before_providers from GitHubCopilotAgent into BaseAgent,
mirroring the existing _run_after_providers helper. Agent's
_prepare_session_and_messages now delegates to the shared base method,
eliminating the near-duplicate provider iteration logic that could
drift as the provider contract evolves.
Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
* Address review feedback for #3984: Python: [Bug]: GitHubCopilotAgent Memory Example
* revert: keep _run_before_providers in GitHubCopilotAgent only
Undo the promotion of _run_before_providers to BaseAgent. The method
stays in GitHubCopilotAgent where it is needed, and _agents.py
retains its original inline provider iteration in RawAgent.
Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
* fix: replace deprecated BaseContextProvider/BaseHistoryProvider with ContextProvider/HistoryProvider
Update imports and usages in GitHubCopilotAgent and its tests to use
the new non-deprecated class names from the core package.
Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
* fix: address review feedback - reorder providers before session, wrap streaming after_run in try/except, assert after_run on skipped HistoryProvider
- Move _run_before_providers before _get_or_create_session so provider
contributions can affect session configuration
- Wrap _run_after_providers in try/except in streaming _after_run_hook
to prevent provider errors from replacing successful responses
- Add after_run assertion to test_history_provider_skip_when_load_messages_false
Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
---------
Co-authored-by: Copilot <copilot@github.com>
Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
Add deduplication to `prepend_instructions_to_messages()` to skip
instructions that are already present as leading messages with the
same role and text. This prevents duplicate system messages when
instructions are injected by multiple layers (e.g. Agent + chat client).
Fixes#5049
* Refactor Anthropic model option and provider clients
Rename the Anthropic client model option from model_id to model, add provider-specific Anthropic wrappers for Foundry, Bedrock, and Vertex, and expose them through the Anthropic, Foundry, Amazon, and Google namespaces. Update core option handling, docs, samples, and tests accordingly.
Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
* Fix Anthropic skills sample typing
Cast the Anthropic beta client to Any in the skills sample so the pre-commit sample pyright check no longer fails on beta skills and files endpoints that are not exposed by the current SDK stubs.
Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
* undo sample mypy
* Retry CI after transient external failures
Retrigger PR validation after an unrelated Copilot review workflow SAML failure and a transient external tau2 git fetch failure in the Windows Python test setup.
Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
* Address review feedback on model option merging
Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
* Address Anthropic compatibility review feedback
Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
* moved all to `model`
* fixes for azure ai search
* Python: standardize remaining sample env var names
Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
* Python: fix foundry-local pyright compatibility
Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
* updated env vars in cicd
---------
Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
* Fix agent_with_hosted_mcp sample to use AzureOpenAIResponsesClient (#4861)
The agent_with_hosted_mcp sample used AzureOpenAIChatClient with an MCP tool
dict, but the Chat Completions API only supports 'function' and 'custom' tool
types, not 'mcp'. This caused a 400 error at runtime.
Switch the sample to AzureOpenAIResponsesClient which natively supports MCP
tools via the Responses API. Use get_mcp_tool() to construct the tool config.
Changes:
- main.py: Replace AzureOpenAIChatClient with AzureOpenAIResponsesClient
- requirements.txt: Update azure-ai-agentserver-agentframework to 1.0.0b16
and use agent-framework-azure-ai package
- agent.yaml: Use AZURE_OPENAI_RESPONSES_DEPLOYMENT_NAME env var
- Add regression test documenting chat client MCP tool passthrough behavior
Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
* Python: Fix agent_with_hosted_mcp sample to use Responses API client for MCP tools
Fixes#4861
* Remove REPRODUCTION_REPORT.md investigation artifact (#4861)
Remove the reproduction report markdown file from the test directory.
Investigation notes belong in the GitHub issue or PR description,
not as committed files in the source tree. The regression test in
test_openai_chat_client.py already provides automated verification.
Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
* Add MCP tool API rejection regression test (#4861)
Add test_mcp_tool_dict_causes_api_rejection to verify that MCP tool
dicts passed through to the Chat Completions API result in a clear
ChatClientException rather than being silently dropped. This completes
the regression test coverage requested in code review.
Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
* small fix
* Revert deletion of dotnet local.settings.json files
Restore the two local.settings.json files that were accidentally deleted in this PR.
Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
---------
Co-authored-by: Copilot <copilot@github.com>
Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
* Fix _add_text_reasoning_content dropping id during coalescing (#4852)
Preserve the id field (rs_* identifier) when coalescing text_reasoning
Content objects by passing id=self.id or other.id to the Content
constructor. This fixes the encrypted reasoning round-trip where the
missing id prevented _prepare_content_for_openai from including it in
the serialized reasoning item.
Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
* Python: Fix `_add_text_reasoning_content` to preserve `id` during coalescing
Fixes#4852
* Raise AdditionItemMismatch on conflicting text_reasoning ids (#4852)
Detect when both operands have different non-empty ids during
text_reasoning Content coalescing and raise AdditionItemMismatch
instead of silently keeping one. This prevents mis-associating
encrypted_content during round-trips.
Also adds tests for conflicting ids and the neither-has-id edge case.
Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
* Address review feedback for #4852: Python: [Bug]: Content._add_text_reasoning_content drops id during coalescing, breaking encrypted reasoning round-trip
* test fix
---------
Co-authored-by: Copilot <copilot@github.com>
Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
* Python: Add header_provider to MCPStreamableHTTPTool (#4808)
Add a header_provider callback parameter to MCPStreamableHTTPTool that
enables injecting dynamic per-request HTTP headers from runtime kwargs
(originating from FunctionInvocationContext.kwargs set in agent middleware).
The implementation uses contextvars and httpx event hooks to ensure headers
are task-local and safe for concurrent tool calls:
- header_provider receives the runtime kwargs dict and returns headers
- call_tool sets a ContextVar before delegating to MCPTool.call_tool
- An httpx request event hook reads from the ContextVar and injects headers
Example usage:
mcp_tool = MCPStreamableHTTPTool(
name="web-api",
url="https://api.example.com/mcp",
header_provider=lambda kwargs: {
"X-Auth-Token": kwargs.get("auth_token", ""),
},
)
Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
* Address review feedback for #4808: Python: [Bug]: Unable to pass AgentContext to MCPStreamableHTTPTool
* Add test for header_provider via FunctionTool.invoke with FunctionInvocationContext
Addresses PR review comment: exercises the full pipeline from
FunctionInvocationContext.kwargs through FunctionTool.invoke to
MCPStreamableHTTPTool.call_tool and header_provider, rather than
testing call_tool in isolation.
Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
* Address review feedback for #4808: review comment fixes
* Fix streamable MCP transport defaults
Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
* Fix Azure AI test client mocks
Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
* Fix MCP runtime kwarg regressions
Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
* Stabilize MCP tool runtime kwargs
Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
* Use context kwargs in MCP wrappers
Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
* updated mcp samples
* fix link
---------
Co-authored-by: Copilot <copilot@github.com>
Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
* Foundry Evals integration for Python
Merged and refactored eval module per Eduard's PR review:
- Merge _eval.py + _local_eval.py into single _evaluation.py
- Convert EvalItem from dataclass to regular class
- Rename to_dict() to to_eval_data()
- Convert _AgentEvalData to TypedDict
- Simplify check system: unified async pattern with isawaitable
- Parallelize checks and evaluators with asyncio.gather
- Add all/any mode to tool_called_check
- Fix bool(passed) truthy bug in _coerce_result
- Remove deprecated function_evaluator/async_function_evaluator aliases
- Remove _MinimalAgent, tighten evaluate_agent signature
- Set self.name in __init__ (LocalEvaluator, FoundryEvals)
- Limit FoundryEvals to AsyncOpenAI only
- Type project_client as AIProjectClient
- Remove NotImplementedError continuous eval code
- Add evaluation samples in 02-agents/ and 03-workflows/
- Update all imports and tests (167 passing)
Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
* fix: resolve mypy redundant-cast errors while keeping pyright happy
Use cast(list[Any], x) with type: ignore[redundant-cast] comments to
satisfy both mypy (which considers casting Any redundant) and pyright
strict mode (which needs explicit casts to narrow Unknown types).
Also fix evaluator decorator check_name type annotation to be
explicitly str, resolving mypy str|Any|None mismatch.
Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
* fix: CI failures — pyupgrade, evaluator overloads, sample API, reset attr
- Apply pyupgrade: Sequence from collections.abc, remove forward-ref quotes
- Add @overload signatures to evaluator() for proper @evaluator usage
- Fix evaluate_workflow sample to use WorkflowBuilder(start_executor=) API
- Fix _workflow.py executor.reset() to use getattr pattern for pyright
- Remove unused EvalResults forward-ref string in default_factory lambda
Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
* fix: skip gRPC-dependent observability test
The test_configure_otel_providers_with_env_file_and_vs_code_port test
triggers gRPC OTLP exporter creation, but the grpc dependency is
optional and not installed by default. Add skipif decorator matching
the pattern used by all other gRPC exporter tests in the same file.
Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
* fix: add nosec B101 for bandit assert check
Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
* style: align eval samples with repo conventions
- Move module docstrings before imports (after copyright header)
- Add -> None return type to all main() and helper functions
- Fix line-too-long in multiturn sample conversation data
- Add Workflow import for typed return in all_patterns_sample
Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
* Address PR review feedback: async fixes, sample bugs, deprecation warnings
- Simplify _ensure_async_result to direct await (async-only clients)
- Replace get_event_loop() with get_running_loop()
- Narrow _fetch_output_items exception handling to specific types
- Add warning log when _filter_tool_evaluators falls back to defaults
- Add DeprecationWarning to options alias in Agent.__init__
- Add DeprecationWarning to evaluate_response()
- Rename raw key to _raw_arguments in convert_message fallback
- Fix evaluate_agent_sample.py: replace evals.select() with FoundryEvals()
- Fix evaluate_multiturn_sample.py: use Message/Content/FunctionTool types
- Fix evaluate_workflow_sample.py: replace evals.select() with FoundryEvals()
- Update test mocks to use AsyncMock for awaited API calls
Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
* Add test coverage for review feedback items
- Add num_repetitions=2 positive test verifying 2×items and 4 agent calls
- Add _poll_eval_run tests: timeout, failed, and canceled paths
- Add evaluate_traces tests: validation error, response_ids path, trace_ids path
- Add evaluate_foundry_target happy-path test with target/query verification
Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
* Fix ruff ISC004 lint error and apply formatter
- Wrap implicit string concatenation in parens in evaluate_multiturn_sample.py
- Apply ruff formatter to 6 other files with minor formatting drift
Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
* Remove core type changes (extracted to fix/workflow-stale-session branch)
Reverts changes to _agents.py, _agent_executor.py, and _workflow.py
back to upstream/main. These fixes are now in a separate PR.
Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
* Address PR review round 2: bugs, tests, and architecture
Code fixes:
- Fix _normalize_queries inverted condition (single query now replicates
to match expected_count)
- Fix substring match bug: 'end' in 'backend' matched; use exact set
lookup for executor ID filtering
- Fix used_available_tools sample: tool_definitions→tools param, use
FunctionTool attribute access instead of dict .get()
- Add None-check in _resolve_openai_client for misconfigured project
- Add Returns section to evaluate_workflow docstring
- Cache inspect.signature in @evaluator wrapper (avoid per-item reflection)
Architecture:
- Extract _evaluate_via_responses as module-level helper; evaluate_traces
now calls it directly instead of creating a FoundryEvals instance
- Move Foundry-specific typed-content conversion out of core to_eval_data;
core now returns plain role/content dicts, FoundryEvals applies
AgentEvalConverter in _evaluate_via_dataset
Tests:
- evaluate_response() deprecation warning emission and delegation
- num_repetitions > 1 with expected_output and expected_tool_calls
- Mock output_items.list in test_evaluate_calls_evals_api
- Update to_eval_data assertions for plain-dict format
- Unknown param error now raised at @evaluator decoration time
Skipped (separate PR): executor reset loop, xfail removal, options alias
Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
* Fix CI: revert test_full_conversation, fix pyright errors
- Revert test_full_conversation.py to upstream/main (the session
preservation test was incorrectly changed to assert clearing)
- Fix pyright reportUnnecessaryComparison on get_openai_client() None
check by adding ignore comment
- Fix pyright reportPrivateUsage: add public EvalItem.split_messages()
method and use it in FoundryEvals._evaluate_via_dataset instead of
accessing private _split_conversation
Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
* Address PR review round 3: reliability, test gaps, cleanup
- Add try/except guard for non-numeric score in _coerce_result
- Add poll_interval minimum bound (0.1s) to prevent tight loops
- Add runtime async client check in _resolve_openai_client
- Remove _ensure_async_result wrapper (10 call sites → direct await)
- Better error message when queries provided without agent
- Import-time asserts for evaluator set consistency
- Remove 28 redundant @pytest.mark.asyncio decorators
- Add doc note about _raw_arguments sensitive data
- Tests: tool_called_check mode=any, _normalize_queries branches,
_extract_result_counts paths, _extract_per_evaluator, bare check
via evaluate_agent, output_items assertion, modulo wrapping,
async client check, queries-without-agent error
Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
* Fix CI: ruff S101 assert, pyright and mypy arg-type errors
- Replace module-level assert with if/raise for evaluator set
consistency checks (ruff S101 disallows bare assert)
- Add type: ignore[arg-type] and pyright: ignore[reportArgumentType]
on OpenAI SDK evals API calls that pass dicts where typed params
are expected (SDK accepts dicts at runtime)
Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
* Address PR review round 4: bugs, reliability, test fixes
- Fix all_passed ignoring parent result_counts when sub_results present
- Fix _extract_tool_calls: parse string arguments via json.loads before
falling back to None (real LLM responses use string arguments)
- Sanitize _raw_arguments to '[unparseable]' to avoid leaking sensitive
tool-call data to external evaluation services
- Add NOTE comment on to_eval_data message serialization dropping
non-text content (tool calls, results)
- Eliminate double conversation split in _evaluate_via_dataset: build
JSONL dicts directly from split_messages + AgentEvalConverter
- Raise poll_interval floor from 0.1s to 1.0s to prevent rate-limit
exhaustion
- Fix MagicMock(name=...) bug in test: sets display name not .name attr
- Fix mock_output_item.sample: use MagicMock object instead of dict so
_fetch_output_items exercises error/usage/input/output extraction
Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
* Address PR review round 5: reliability, docs, test coverage
Code fixes:
- Move import-time RuntimeError checks to unit tests (avoids breaking
imports for all users on developer set-drift mistake)
- _filter_tool_evaluators now raises ValueError when all evaluators
require tools but no items have tools (was silently substituting)
- Add poll_interval upper bound (60s) to prevent single-iteration sleep
- Log exc_info=True in _fetch_output_items for debugging API changes
- Fix evaluate() docstring: remove claim about Responses API optimization
- Validate target dict has 'type' key in evaluate_foundry_target
- Document to_eval_data() limitation: non-text content is omitted
Tests:
- TestEvaluatorSetConsistency: verify _AGENT/_TOOL subsets of _BUILTIN
- TestEvaluateTracesAgentId: agent_id-only path with lookback_hours
- TestFilterToolEvaluatorsRaises: ValueError on all-tool no-items
- TestEvaluateFoundryTargetValidation: target without 'type' key
- Assert items==[] on failed/canceled poll results
- Mock output_items.list in response_ids test for full flow
- TestAllPassedSubResults: result_counts=None + sub_results delegation
and parent failures override sub_results
- TestBuildOverallItemEmpty: empty workflow outputs returns None
Skipped r5-07 (_raw_arguments length hint): marginal debugging value,
could leak content size information.
Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
* Fix error message: evaluate_responses() → evaluate_traces(response_ids=...)
The referenced function doesn't exist; the correct API is
evaluate_traces(response_ids=...) from the azure-ai package.
Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
* Remove dead to_eval_data() method, fix docstring claims
- Remove to_eval_data() from EvalItem (dead code after r4-05 JSONL refactor)
- Migrate 15 tests from to_eval_data() to split_messages()
- Update sample to use split_messages() + Message properties
- Remove unimplemented Responses API optimization docstring claim
- Update split_messages() docstring to not reference removed method
Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
* Reduce default eval timeout from 600s to 180s (3 minutes)
Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
* Remove dead _evaluate_via_responses method from FoundryEvals
The method was never called — evaluate() uses _evaluate_via_dataset,
and evaluate_traces() calls _evaluate_via_responses_impl directly.
Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
* Revert unrelated formatting changes to get-started samples
Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
* Fix pyright: remove phantom FoundryMemoryProvider import, apply ruff format
- Remove import of non-existent _foundry_memory_provider module
(incorrectly kept during rebase conflict resolution)
- Apply ruff formatter to test_local_eval.py and get-started samples
Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
* Fix eval samples: use FoundryChatClient for Agent()
The upstream provider-leading client refactor (#4818) made client=
a required parameter on Agent(). Update the three getting-started
eval samples to use FoundryChatClient with FOUNDRY_PROJECT_ENDPOINT,
matching the standard pattern from 01-get-started samples.
Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
* Simplify self-reflection sample using FoundryEvals
Replace ~80 lines of manual OpenAI evals API code (create_eval,
run_eval, manual polling, raw JSONL params) with FoundryEvals:
- evaluate_groundedness() uses FoundryEvals.evaluate() with EvalItem
- Remove create_openai_client(), create_eval(), run_eval() functions
- Remove openai SDK type imports (DataSourceConfigCustom, etc.)
- run_self_reflection_batch creates FoundryEvals instance once,
reuses it for all iterations across all prompts
Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
* Update eval samples to FoundryChatClient and FOUNDRY_PROJECT_ENDPOINT
- Migrate all foundry_evals samples from AzureOpenAIResponsesClient to FoundryChatClient
- Update env var from AZURE_AI_PROJECT_ENDPOINT to FOUNDRY_PROJECT_ENDPOINT
- Use AzureCliCredential consistently across all samples
- Fix README.md: correct function names (evaluate_dataset -> FoundryEvals.evaluate, evaluate_responses -> evaluate_traces)
- Update self_reflection .env.example and README.md
Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
* Fix lint errors in eval samples (E501, ASYNC240, formatting)
Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
* Remove evaluate_all_patterns_sample.py (redundant with focused samples)
Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
* Fix async credential mismatch: use azure.identity.aio for async AIProjectClient
AIProjectClient from azure.ai.projects.aio requires an async credential.
Switch all foundry_evals samples from azure.identity.AzureCliCredential
to azure.identity.aio.AzureCliCredential. Also pass project_client to
FoundryChatClient instead of duplicating endpoint+credential.
Close credential in self_reflection sample to avoid resource leak.
Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
* Revert test_observability.py to upstream/main (not our test)
Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
* Address moonbox3 review: sphinx docstrings, pagination, isinstance check
- Convert all Example:: / Typical usage:: code blocks to .. code-block:: python
format matching codebase convention (both _evaluation.py and _foundry_evals.py)
- Add async pagination in _fetch_output_items via async for (handles large result sets)
- Replace hasattr(__aenter__) with isinstance(client, AsyncOpenAI) in _resolve_openai_client
- Move AsyncOpenAI import from TYPE_CHECKING to runtime (needed for isinstance)
Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
* Fix test failures and address remaining moonbox3 review comments
- Fix tests: use MagicMock(spec=AsyncOpenAI) for project_client mocks
(isinstance check now requires proper type, not duck-typing)
- Fix tests: replace mock_page.__iter__ with _AsyncPage helper for async for
- Fix evaluate_response: auto-extract queries from response messages when
query is not provided (previously always raised ValueError)
- Add debug logging when skipping internal _-prefixed executor IDs
Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
* Address Tao's PR review comments on Foundry Evals
- T1: Add comment explaining builtin.* pass-through in _resolve_evaluator
- T2: Add comment referencing OpenAI evals API for testing_criteria dict
- T3: Document Mustache-style {{item.*}} template placeholders
- T4: Document poll loop 60s sleep upper bound rationale
- T5: Narrow run type to RunRetrieveResponse, use typed field access
instead of vars()/getattr dance in _extract_result_counts and
_extract_per_evaluator; use run.error and run.report_url directly
- T6: Clarify openai_client docstring re: Azure Foundry endpoint
- T8: Remove misleading empty expected_tool_calls from sample
- Update tests to match real SDK PerTestingCriteriaResult shape
Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
* Remove unnecessary Any union from run type annotations
RunRetrieveResponse is the correct type — no backward compat needed
for a brand new feature.
Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
* Accept FoundryChatClient instead of raw AsyncOpenAI
FoundryEvals now takes client: FoundryChatClient as its primary
parameter instead of openai_client: AsyncOpenAI. The builtin.*
evaluators require a Foundry endpoint, so the type should reflect that.
- FoundryEvals.__init__: client: FoundryChatClient replaces openai_client
- evaluate_traces / evaluate_foundry_target: same change
- _resolve_openai_client: extracts .client from FoundryChatClient
- project_client fallback retained for standalone functions
- All samples updated to construct FoundryChatClient and pass as client=
- Tests updated (openai_client= → client=)
Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
* Remove implicit 60s upper bound on poll interval
If a developer sets a higher poll_interval, respect it. Only clamp
to remaining time and enforce a 1s minimum for rate-limit protection.
Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
* Remove 1s floor on poll interval — let the developer control it
Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
* Update python/samples/05-end-to-end/evaluation/foundry_evals/.env.example
Co-authored-by: Eduard van Valkenburg <eavanvalkenburg@users.noreply.github.com>
* Update python/samples/02-agents/evaluation/evaluate_agent.py
Co-authored-by: Eduard van Valkenburg <eavanvalkenburg@users.noreply.github.com>
* Address eavanvalkenburg review (round 2) on Python eval PR
- Rename model_deployment -> model across FoundryEvals and all samples
- Make model param optional, resolves from client.model
- Convert EvalResults from dataclass to regular class
- Remove deprecated evaluate_response() function
- Refactor splitters: BUILT_IN_SPLITTERS dict + standalone functions
- Change per_turn_items from classmethod to staticmethod
- Simplify EvalCheck type alias to use Awaitable[CheckResult]
- Remove errored property from EvalResults
- Remove default value from Evaluator protocol eval_name
- Rename assert_passed -> raise_for_status, add EvalNotPassedError
- Type agent param as SupportsAgentRun | None
- Fix Arguments docstring
- Update __init__.py exports
- Update all tests and samples
Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
* Move FoundryEvals to foundry package, split tool eval sample
- Move _foundry_evals.py from azure-ai to foundry package
- Move test_foundry_evals.py to foundry/tests/
- Update lazy re-exports in agent_framework.foundry namespace
- Update .pyi type stubs
- All samples now import from agent_framework.foundry
- Split tool-call evaluation into evaluate_tool_calls_sample.py
- Fix all_passed to check errored count from result_counts
- Fix raise_for_status to include errored item details
Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
* Auto-create FoundryChatClient from env vars when no client provided
FoundryEvals() now works zero-config when FOUNDRY_PROJECT_ENDPOINT and
FOUNDRY_MODEL environment variables are set. Auto-creates a FoundryChatClient
under the hood, matching the established env var pattern.
Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
* Fix pyright errors: remove dead _normalize_queries, suppress EvalAPIError check
- Remove unused _normalize_queries function and its tests
- Add pyright ignore for EvalAPIError None check (defensive guard)
Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
* Support multimodal image content in eval pipeline
Add image (data/uri) content handling to AgentEvalConverter.convert_message()
so that Content.from_data() and Content.from_uri() image payloads are
preserved as input_image parts in the Foundry evaluator format.
- Handle Content type='data' and type='uri' → emit input_image parts
- Add 6 unit tests for image content through convert_message/convert_messages
- Add integration test verifying images flow through EvalItem → JSONL path
- Add evaluate_multimodal.py sample demonstrating local image eval
Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
* Address remaining review comments
- Fix project_client docstring to say async-only (not sync/async)
- Add builtin evaluator name validation warning in _resolve_evaluator
- Replace getattr with typed attribute access in _poll_eval_run,
_extract_result_counts, _extract_per_evaluator, _fetch_output_items
- Remove cast import from _foundry_evals (no longer needed)
- Tighten _coerce_result: honour explicit 'passed' when both 'score'
and 'passed' are present; remove performative cast
- Fix self_reflection sample: add env file existence check
- Fix traces sample: correct Pattern 2 section label
- Update all Foundry eval samples to FoundryChatClient + FOUNDRY_MODEL
(remove AIProjectClient + AZURE_AI_MODEL_DEPLOYMENT_NAME pattern)
- Add eval_name and OpenAI client docs to FoundryEvals docstring
- Update test mocks to match typed SDK objects (_MockResultCounts)
Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
* Fix ruff lint errors (E501, SIM108, SIM102)
Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
* Fix pyright errors: type-narrow dict to dict[str, Any], add ignore comments
Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
* Replace ConversationSplitter type alias with Protocol
ConversationSplitter is now a runtime-checkable Protocol with a named
'conversation' parameter, making the expected signature self-documenting.
ConversationSplit enum members gain a __call__ method so they satisfy
the protocol directly -- ConversationSplit.LAST_TURN(conversation) works.
This simplifies _split_conversation from an isinstance dispatch to a
single split(conversation) call.
Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
* Standardize on AZURE_AI_MODEL_DEPLOYMENT_NAME and fix Unicode in samples
- Replace FOUNDRY_MODEL with AZURE_AI_MODEL_DEPLOYMENT_NAME in all
eval samples to match repo convention
- Replace Unicode symbols with ASCII equivalents in all eval sample
print statements to avoid cp1252 encoding errors on Windows
Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
* Update python/samples/03-workflows/evaluation/evaluate_workflow.py
Co-authored-by: Eduard van Valkenburg <eavanvalkenburg@users.noreply.github.com>
* Apply suggestions from code review
Co-authored-by: Eduard van Valkenburg <eavanvalkenburg@users.noreply.github.com>
* Rename ADR 0020 to 0023 (foundry evals integration)
Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
---------
Co-authored-by: alliscode <bentho@microsoft.com>
Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
Co-authored-by: Eduard van Valkenburg <eavanvalkenburg@users.noreply.github.com>
Tao ChenandGitHub
Dineshsuriya DandGitHub
hashwnathandGitHub
