[BREAKING] Python: Update github-copilot-sdk integration to use ToolInvocation/ToolResult types

[moonbox3]

Motivation and Context

The upgrade of github-copilot-sdk to >=0.1.32 introduced breaking changes:

• Handler now receives a ToolInvocation dataclass instead of a plain dict.
• Handler now returns ToolResult with snake_case fields (result_type, text_result_for_llm) instead of camelCase (resultType, textResultForLlm).
• The SDK now requires Python >=3.11, dropping Python 3.10 support.

Fixes #4549

Description

Production code changes (_agent.py):

• Update _tool_to_copilot_tool handler to access invocation.arguments (attribute) instead of invocation.get("arguments") (dict).
• Update ToolResult construction to use snake_case fields (text_result_for_llm, result_type) instead of camelCase.
• Fix PermissionRequest import location (moved from copilot.types to copilot.generated.session_events in SDK 0.1.32).

Dependency & Python version changes:

• Bump github-copilot-sdk dependency from >=0.1.0 to >=0.1.32.
• Set requires-python = ">=3.11" for the agent-framework-github-copilot package, matching the SDK's requirement.
• Gate agent-framework-github-copilot in core's [all] extra with python_version >= '3.11' marker so uv sync still works on Python 3.10 for all other packages.

CI/CD changes:

• Update single-version workflows (python-code-quality, python-test-coverage, python-integration-tests, python-merge-tests) from Python 3.10 to 3.11.
• Add exclude-packages input to the python-setup action to exclude incompatible workspace members (removes them from workspace resolution via uv.workspace.exclude).
• On Python 3.10 matrix runs in python-tests and python-lab-tests, exclude agent-framework-github-copilot from install and skip its tests.

Sample fixes (github_copilot samples):

• Fix PermissionRequest usage: replace dict-style .get() calls with dataclass attribute access.

Test changes (test_github_copilot_agent.py):

• Update existing tool handler tests to pass ToolInvocation(...) and assert on ToolResult attributes.
• Add test_tool_handler_rejects_raw_dict_invocation to verify the old dict calling convention raises an error.
• Add test_tool_handler_with_empty_arguments to cover the empty-arguments edge case.

Contribution Checklist

• The code builds clean without any errors or warnings
• The PR follows the https://github.com/microsoft/agent-framework/blob/main/CONTRIBUTING.md
• All unit tests pass, and I have added new tests where possible
• Is this a breaking change? Yes — agent-framework-github-copilot now requires Python >=3.11 (matching the upstream github-copilot-sdk >=0.1.32 requirement). All other agent-framework packages continue to support Python 3.10.

[moonbox3]

Automated Code Review

Reviewers: 4 | Confidence: 92%

✓ Correctness

The diff updates two test call sites to pass a ToolInvocation object instead of a plain dictionary to copilot_tool.handler(), aligning the tests with a likely signature change in the handler that now expects a typed ToolInvocation rather than a raw dict. The import is correctly added and the construction ToolInvocation(arguments={"arg": "test"}) matches the previous dict shape. No correctness issues found.

✓ Security Reliability

This diff is a minor test-only change that replaces raw dictionary arguments with typed ToolInvocation objects when calling copilot tool handlers. This improves type safety in the tests and aligns them with an updated handler signature. There are no security or reliability concerns.

✓ Test Coverage

The diff updates two existing tests to pass a ToolInvocation object instead of a raw dict to copilot_tool.handler(), aligning tests with a signature change. The existing assertions (checking result_type for success and failure) remain meaningful. However, there are no new tests covering edge cases of the ToolInvocation type — e.g., missing arguments, extra fields, or None arguments — and no test verifies that passing a raw dict now raises an error (if that was the intent of the breaking change).

✓ Design Approach

The diff updates two test call sites to pass a proper ToolInvocation object instead of a raw dictionary to copilot_tool.handler(). This is a correct and straightforward change that aligns tests with a typed interface, replacing duck-typed dictionaries with the actual domain object. No design concerns.

Suggestions

• Consider adding a test that verifies passing a raw dict (the old calling convention) to copilot_tool.handler() now raises a TypeError or similar error, to lock in the new contract.
• Consider adding a test with ToolInvocation(arguments={}) (empty arguments) to cover the edge case where no arguments are provided.

---

Automated review by moonbox3's agents

[Copilot]

Pull request overview

This PR updates the GitHub Copilot Python integration to align with github-copilot-sdk>=0.1.32’s runtime types, and adds/adjusts tests so future regressions (back to dict-based tool invocation/results) are caught.

Changes:

• Update GitHubCopilotAgent tool handler to consume ToolInvocation and return ToolResult with snake_case fields.
• Update regression tests to pass ToolInvocation(...) and assert on ToolResult attributes.
• Bump agent-framework-github-copilot to Python >=3.11 and upgrade dependency to github-copilot-sdk>=0.1.32.

Reviewed changes

Copilot reviewed 3 out of 4 changed files in this pull request and generated 2 comments.

File	Description
python/packages/github_copilot/tests/test_github_copilot_agent.py	Updates tool handler tests to use ToolInvocation and validate ToolResult attributes.
python/packages/github_copilot/pyproject.toml	Raises minimum Python to 3.11, removes 3.10 classifier, updates mypy target, and bumps github-copilot-sdk dependency.
python/packages/github_copilot/agent_framework_github_copilot/_agent.py	Updates tool conversion handler to use invocation.arguments and construct ToolResult using snake_case fields; adjusts imports.

---

You can also share your feedback on Copilot code review. Take the survey.

[markwallace-microsoft]

Python Test Coverage Report •

File	Stmts	Miss	Cover	Missing
TOTAL	22676	2565	88%

report-only-changed-files is enabled. No files were changed during this commit :)

Python Unit Test Overview

Tests	Skipped	Failures	Errors	Time
4932	20 💤	0 ❌	0 🔥	1m 17s ⏱️