v0.1.29

🚀 Release Highlights — v0.1.29 (2025-12-24)

🔀 Subagent System

• Parallel Execution: Spawn child MassGen processes via spawn_subagents tool for independent task execution
• Process Isolation: Each subagent runs in its own workspace with automatic result aggregation and token usage tracking
• Backend Inheritance: Subagents automatically inherit parent agent's backend configuration

📊 Tool Metrics Distribution

• Per-Call Averages: Enhanced get_tool_metrics_summary() with average input/output chars per call
• Distribution Statistics: Min/max/median stats for output size and execution time

⚙️ Per-Agent System Messages

• New Mode: massgen --quickstart now offers "Different message per agent" option
• Flexible Configuration: Choose between "Skip", "Same for all", or "Different per agent"

🐛 Bug Fixes

• OpenAI Responses API: Fixed duplicate item errors when using previous_response_id
• Function Call ID: Preserved 'id' field for proper reasoning item pairing

---

📖 Getting Started

• Quick Start Guide: Try the new features today
• Try These Examples:
  • test_subagent_orchestrator.yaml - Try parallel subagent execution for research tasks
  • test_subagent_orchestrator_code_mode.yaml - Try subagents with code-based tools and Docker execution
  • massgen --quickstart - Configure per-agent system messages in the quickstart wizard

What's Changed

• docs: update slides after v0.1.26 by @sonichi in #656
• fix: Fix openai responses api issues by @ncrispino in #685
• feat: Add subagents by @ncrispino in #690
• docs: docs for v0.1.29 by @Henry-811 in #696
• feat: v0.1.29 by @Henry-811 in #694

Full Changelog: v0.1.28...v0.1.29