Python: Fix Pydantic error when using Literal type for tool params

[moonbox3]

Motivation and Context

When using Literal type annotations for tool parameters in @ai_function decorated methods, Pydantic throws:

PydanticUserError: search_flows_input is not fully defined; you should define Data, then call search_flows_input.model_rebuild().

The _parse_annotation function incorrectly processed Literal types. For Literal["Data", "Security"], the args are ("Data", "Security"). The code checked if args[1] is a string (which "Security" is), then mistakenly created Annotated[ForwardRef('Data'), Field(description='Security')], treating literal values as type references.

The fix: added early return in _parse_annotation when get_origin(annotation) is Literal to preserve Literal types unchanged.

Testing:

Added 9 tests covering:

• Literal string/int/bool types in standalone functions and class methods
• Literal combined with Annotated and Field
• Unit tests for _parse_annotation function

Description

• Fixes Python: Pydantic error when attempting to use Literal type for tool parameter #2891

Contribution Checklist

• The code builds clean without any errors or warnings
• The PR follows the Contribution Guidelines
• All unit tests pass, and I have added new tests where possible
• Is this a breaking change? If yes, add "[BREAKING]" prefix to the title of the PR.

[markwallace-microsoft]

Python Test Coverage Report •

File	Stmts	Miss	Cover	Missing
packages/core/agent_framework
_tools.py	687	65	90%	224, 270, 321, 323, 486, 518–519, 620, 622, 642, 660, 674, 686, 691, 693, 700, 733, 787–789, 827, 850, 852–861, 870–876, 912, 922, 1106, 1479–1483, 1583, 1652, 1753, 1759, 1802–1803, 1938, 1979–1980, 2008–2010, 2045–2046, 2111–2112, 2119–2120
TOTAL	16604	2573	84%

Python Unit Test Overview

Tests	Skipped	Failures	Errors	Time
2406	130 💤	0 ❌	0 🔥	57.295s ⏱️

[Copilot]

Pull request overview

This PR fixes a Pydantic error that occurred when using Literal type annotations for tool parameters in @ai_function decorated methods. The issue was that _parse_annotation incorrectly processed Literal types, treating their literal values as type references and creating invalid Annotated types.

Key changes:

• Added early return for Literal types in _parse_annotation to preserve them unchanged
• Added comprehensive test coverage with 9 new tests covering various Literal use cases

Reviewed changes

Copilot reviewed 2 out of 2 changed files in this pull request and generated no comments.

File	Description
python/packages/core/agent_framework/_tools.py	Fixed _parse_annotation to return Literal types unchanged, preventing Pydantic errors
python/packages/core/tests/core/test_tools.py	Added 9 tests covering Literal types with strings/ints/bools in functions and class methods, plus unit tests for _parse_annotation