[HamzaKhanBUIC]

Summary

Fixes #62.

Resolves an issue where Google Gemini parallel function calls caused HTTP 400 Bad Request errors on subsequent model turns.

Problem

1. When Gemini emitted multiple parallel function calls in a single candidate turn, GeminiClient._convert_messages mapped each tool result into its own ypes.Content(role=user) container. The Gemini API requires all parallel function responses to be bundled together into a single role=user turn with one unction_response part per function call.
2. Inbound FunctionCall.id was discarded in favor of a randomly generated uuid.uuid4(), and outbound unction_response.id was omitted, causing call-result join failure.

Solution

• Grouping Parallel Tool Responses: Updated GeminiClient._convert_messages so that consecutive Role.TOOL messages following an assistant turn are grouped into a single ypes.Content(role=user) turn.
• Tool ID Preservation:
  • Outbound: Sets ool_part.function_response.id = msg.tool_call_id.
  • Inbound: Preserves c.id when provided by the API as a string, with UUID fallback.
• Regression Tests: Added est_gemini_parallel_function_calls_grouped_and_preserve_ids in ests/unit/test_gemini_adapter.py.

Verification

• uv run pytest tests/unit/test_gemini_adapter.py -> 6/6 passed
• uv run ruff check . -> All checks passed
• uv run ruff format --check . -> 110 files already formatted
• uv run mypy agentkit -> Success: no issues found in 42 source files
• Structural AST lint & Semgrep security check -> Clean

[Copilot]

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