Skip to content

v0.1.37

Choose a tag to compare

@Henry-811 Henry-811 released this 12 Jan 18:32
· 648 commits to main since this release
103dd52

πŸš€ Release Highlights β€” v0.1.37 (2026-01-12)

πŸ“œ Execution Traces

  • Full History Preservation: Human-readable execution_trace.md files saved alongside agent snapshots
  • Compression Recovery: Agents can read trace files to recover detailed history after context compression
  • Cross-Agent Access: Other agents can access execution traces in temp workspaces to understand approaches
  • No Truncation: Tool calls, results, and reasoning blocks preserved in full
  • Grep-Friendly: Searchable markdown format for debugging and analysis

🧠 Thinking Mode Improvements

  • Claude Code Thinking: Streaming buffer support for Claude Code reasoning content
  • Gemini Thinking Fixes: Proper reasoning content handling and streaming buffer integration
  • Vote Reasoning Traces: Full vote context captured in execution trace files

🏷️ Standardized Agent Labeling

  • Unified Format: Consistent agent identification across all backends
  • Improved Anonymization: Better workspace anonymization for cross-agent sharing
  • Coordination Context: Clearer agent references in multi-agent workflows

πŸ› Bug Fixes

  • Claude Code Backend: Fixed skills and tool handling issues
  • Config Builder: Fixed configuration generation edge cases
  • Round Timeout: Improved timeout behavior during coordination

πŸ“– Getting Started

# Run a multi-agent task and check execution traces
uv run massgen --config massgen/configs/basic/multi/three_agents_default.yaml \
  "Explain the benefits of functional programming"

# Test round timeout behavior
uv run massgen --config massgen/configs/debug/round_timeout_test.yaml \
  "Write a short story"

What's Changed

Full Changelog: v0.1.36...v0.1.37