Changelog

[v0.1.0] — 2025-01-15 Initial Release

• GEP v0 (Geneclaw Evolution Protocol): Observe → Diagnose → Propose → Gate → Apply closed loop
• 5-layer Gatekeeper: path allowlist/denylist, diff size limit, secret scan, code pattern detection, dry-run pytest gate
• geneclaw doctor: read-only environment health check
• geneclaw evolve: dry-run evolution proposal generation with heuristic and LLM diagnosis modes
• geneclaw gate: standalone Gatekeeper runner for proposals
• geneclaw apply: safe application with git branch, pytest gate, and auto-rollback on test failure
• geneclaw apply --rollback: one-command rollback for any applied proposal
• geneclaw report: table, JSON, and Markdown reporting from the event store
• geneclaw autopilot: multi-cycle autonomous loop with configurable risk thresholds
• geneclaw benchmark: performance benchmarking with before/after comparison
• geneclaw dashboard: local read-only audit dashboard
• Append-only JSONL event store with automatic secret/PII redaction
• Structured GEP JSON proposal format with unified diff, risk score, and rollback plan
• geneclaw.toml configuration system
• nanobot integration for agent runtime and tool use

• nanobot (HKUDS/nanobot) — agent runtime
• Python 3.9+
• pytest (for gate layer 5)
• GitPython (for branch management)

• LLM diagnosis mode requires manual API key configuration; not bundled with a provider
• Dashboard is local-only; no web-accessible deployment in v0.1.0
• Multi-agent coordination not yet supported
• Autopilot mode limited to single-agent, single-repo scenarios