Skip to content

Harden A2A eval bug paths#105

Merged
pmbstyle merged 4 commits into
mainfrom
fix-a2a-eval-system-bugs
May 19, 2026
Merged

Harden A2A eval bug paths#105
pmbstyle merged 4 commits into
mainfrom
fix-a2a-eval-system-bugs

Conversation

@pmbstyle
Copy link
Copy Markdown
Owner

Summary

  • isolate inbound A2A turns by context and mark empty A2A replies as failed
  • retry provider tool calls with tool_choice=auto when thinking-mode routes reject required/object tool choice
  • expose worker launcher Docker readiness through gateway_status
  • avoid rewriting normal short final replies like noted/done as textual tool invocations

Tests

  • uv run pytest tests\test_a2a_gateway.py tests\test_litellm_provider_payload_hardening.py tests\test_gateway_status_tool.py
  • uv run pytest tests\test_heartbeat.py::test_detect_textual_tool_invocation tests\test_a2a_gateway.py tests\test_litellm_provider_payload_hardening.py tests\test_gateway_status_tool.py
  • Test instance smoke: A2A direct worker summary completed, ops status reported docker/image ready, isolated context did not leak marker, short noted. reply preserved

@pmbstyle pmbstyle self-assigned this May 19, 2026
@pmbstyle pmbstyle marked this pull request as ready for review May 19, 2026 14:35
@pmbstyle pmbstyle merged commit b0dcab2 into main May 19, 2026
4 checks passed
@pmbstyle pmbstyle deleted the fix-a2a-eval-system-bugs branch May 19, 2026 14:35
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant