Include all assistant tool iterations in summaries

[pragnyanramtha]

Summary

• Accumulate AssistantAgent tool call results across max_tool_iterations for the final no-reflection summary.
• Preserve per-iteration ToolCallExecutionEvent contents and handoff handling scope.
• Add regression coverage for multi-iteration tool summaries.

Root Cause

AssistantAgent._process_model_result replaced executed_calls_and_results on each tool-call loop iteration. When max_tool_iterations was greater than 1 and reflect_on_tool_use was false, the final ToolCallSummaryMessage only included the last iteration.

Validation

• uv run pytest packages/autogen-agentchat/tests/test_assistant_agent.py -k "tool_call_loop_summary_includes_all_iterations or tool_call_loop_max_iterations or tool_call_loop_with_handoff"
• uv run pytest packages/autogen-agentchat/tests/test_assistant_agent.py
• uv run ruff check packages/autogen-agentchat/src/autogen_agentchat/agents/_assistant_agent.py packages/autogen-agentchat/tests/test_assistant_agent.py
• uv run ruff format --check packages/autogen-agentchat/src/autogen_agentchat/agents/_assistant_agent.py packages/autogen-agentchat/tests/test_assistant_agent.py
• uv run pyright src/autogen_agentchat/agents/_assistant_agent.py tests/test_assistant_agent.py
• git diff --check

[Copilot]

Copilot encountered an error and was unable to review this pull request. You can try again by re-requesting a review.