Entroly v0.5 — Multi-Agent Context Orchestration + OpenClaw Integration

[juyterman1000]

We just shipped the biggest update to Entroly yet:

What's New

Multi-Agent Context Orchestration — When your AI spawns subagents (OpenClaw, Claude Code tasks, etc.), each agent now gets its own optimized context slice. Entroly uses game-theoretic budget allocation (Nash bargaining) to split your token budget fairly across all running agents.

OpenClaw Context Engine — Deep integration with OpenClaw. Every agent type gets specialized optimization:

• Main agent: full codebase visibility
• Heartbeat: only loads changes since last check (~90% savings)
• Subagents: inherited parent context with budget decay
• Cron jobs: minimal context for scheduled tasks
• Group chat: entropy-based message filtering

Hierarchical Context Compression (HCC) — Instead of binary include/exclude, every code fragment now gets one of 3 compression levels (Full / Skeleton / Reference). Rate-distortion optimization picks the best level per fragment.

AutoTune — Entroly now automatically calibrates its scoring weights per workspace. No manual tuning needed.

Rust Performance — 21 modules, 249 tests, all passing. The new NKBE allocator and Cognitive Bus are written in Rust for <1ms overhead.

---

Upgrade: pip install --upgrade entroly

Feedback welcome. Open an issue or reply here.