[Dineshsuriya D (droideronline)]

Motivation & Context

function_invocation_kwargs passed to a GitHubCopilotAgent (directly via agent.run(...), or from a workflow via workflow.run(..., function_invocation_kwargs=...)) never reached its tools. RawGitHubCopilotAgent.run accepted them only through **kwargs and dropped them, and the Copilot SDK tool handler called FunctionTool.invoke without a FunctionInvocationContext. Tools that declare a FunctionInvocationContext parameter therefore saw empty ctx.kwargs and no ctx.session, unlike the same tool on a chat-client-backed Agent.

Description & Review Guide

• What are the major changes?
  • RawGitHubCopilotAgent.run (all overloads) now takes function_invocation_kwargs as a named parameter and forwards it to _run_impl and _stream_updates.
  • The value, together with the current AgentSession, is threaded through _get_or_create_session, _create_session / _resume_session, _build_session_kwargs and _prepare_tools into _tool_to_copilot_tool.
  • The tool handler builds a FunctionInvocationContext(function, arguments, session, kwargs) and passes it to ai_func.invoke(...), along with the SDK's tool_call_id for telemetry.
• What is the impact of these changes?
  • Tools that declare a FunctionInvocationContext parameter now receive the run's function_invocation_kwargs and session, matching core Agent behavior. Tools without a context parameter are unaffected, and the context parameter is not exposed in the tool schema sent to the model.
  • Tools are rebuilt on every session create/resume, so each run's handlers capture that run's kwargs.
• What do you want reviewers to focus on?
  • Passing the AgentSession to tools via ctx.session, and the new keyword-only parameters on the private session and tool helpers.

Tests added for non-streaming and streaming runs, resumed sessions, and tools without a context parameter.

Related Issue

Fixes #8865

Contribution Checklist

• The code builds clean without any errors or warnings
• All unit tests pass, and I have added new tests where possible
• The PR follows the Contribution Guidelines
• This PR is linked to an issue and there is no other open PR for this issue (see Related Issue above).
• This is not a breaking change. If it is a breaking change, add the breaking change label (or add "[BREAKING]" to the title prefix, before or after any language prefix) — a workflow keeps the label and title prefix in sync automatically.

[Copilot]

Copilot was unable to review this pull request because the user who requested the review has reached their quota limit.

[Dineshsuriya D (droideronline)]

Fixed the failing "Test Typing Checks" job in 0eb1e19: the overlapping-run tests now use asyncio.ensure_future for agent.run(...) (typed as Awaitable, not a coroutine) and a named mock_on instead of the append(...) or lambda. No library code changed. Also merged latest main.