Evolver turns logs into evolution prompts
Evolver is the JavaScript core of EvoMap’s GEP stack: a Node.js 18+ engine that scans runtime logs and memory signals, selects a matching Gene or Capsule, and emits a structured evolution prompt for the next agent iteration. It stays offline by default, with optional network features for skill sharing and leaderboards through EvoMap.
Strong concept, and the repo makes the boundary clear: this is about turning messy agent feedback into a governed evolution loop, not giving the model free rein.
- –Uses a protocol-first framing: logs in, constrained GEP prompt out, with auditability as a first-class feature.
- –Offline operation matters here; the network layer is optional, so the core loop is not dependent on evomap.ai.
- –The “not a code patcher” stance is a real product distinction and reduces risk compared with autonomous-edit tools.
- –The repo reads like infrastructure for agent ops teams that want repeatability, rollback, and traceable evolution artifacts.
- –The surge in stars suggests strong early attention, but the real test is whether the protocol stays useful beyond the demo loop.
DISCOVERED
45d ago
2026-04-16
PUBLISHED
45d ago
2026-04-16
RELEVANCE