evolver  by EvoMap

Protocol-driven AI agent self-evolution SDK

Created 1 month ago
1,487 stars

Top 27.3% on SourcePulse

GitHubView on GitHub
Project Summary

Summary

The Capability Evolver is a protocol-constrained self-evolution engine for AI agents, transforming ad hoc prompt tweaks into auditable, reusable evolution assets. It solves the challenge of managing AI agent evolution at scale by providing a structured, auditable, and deterministic approach. Ideal for teams maintaining large-scale agent prompts and logs, users requiring traceable evolution, and environments demanding protocol-bound changes, it offers a robust framework for controlled AI adaptation.

How It Works

The engine inspects runtime history for signals, guiding the selection of reusable "Genes" or "Capsules." It then emits a strict Genome Evolution Protocol (GEP) prompt to direct safe evolution. This approach leverages protocol-constrained evolution, audit trails, and structured assets for traceable, governed changes. Its novelty lies in prioritizing auditable, protocol-bound modifications over free-form creative alterations, establishing a systematic method for AI agent evolution.

Quick Start & Requirements

  • Primary run command: node index.js
  • Prerequisites: Node.js environment.
  • No explicit quick-start or demo links are provided.

Highlighted Details

  • Auto-Log Analysis: Scans logs for errors and patterns to inform evolution.
  • GEP Protocol: Standardized protocol for auditable evolution using Genes, Capsules, and Events.
  • Configurable Strategies: Supports EVOLVE_STRATEGY presets (balanced, innovate, harden, repair-only).
  • Signal De-duplication: Prevents repetitive repair loops by detecting stagnation.
  • Protected Source Files: Core evolver code is immutable.
  • Gene Validation Safety: Rigorous checks on commands executed during Gene validation.
  • Containerized Testing: Includes a vibe testing framework for end-to-end validation.

Maintenance & Community

A detailed changelog indicates active development, with recent updates focusing on operational modules and configurable strategies. No specific community channels or sponsorship information are listed. A roadmap outlines plans for a demo workflow and comparison table.

Licensing & Compatibility

Released under the MIT license, which is permissive for commercial use and closed-source linking.

Limitations & Caveats

Not intended for one-off scripts, free-form creative changes, or systems sensitive to protocol overhead. It acts as a safety-focused evolution guide, not a live patcher, requiring review and validation for production. The evolver generates prompts to guide evolution, not direct code editing. Core infrastructure skills are designated "Forbidden Innovation Zones."

Health Check
Last Commit

1 day ago

Responsiveness

Inactive

Pull Requests (30d)
22
Issues (30d)
217
Star History
1,632 stars in the last 30 days

Explore Similar Projects

Starred by Bryan Helmig Bryan Helmig(Cofounder of Zapier) and Jared Palmer Jared Palmer(SVP at GitHub; Founder of Turborepo; Author of Formik, TSDX).

dspyground by karthikscale3

1.3%
300
Optimize AI agent prompts with DSPy GEPA
Created 5 months ago
Updated 1 month ago
Starred by Peter Norvig Peter Norvig(Author of "Artificial Intelligence: A Modern Approach"; Research Director at Google) and Yiran Wu Yiran Wu(Coauthor of AutoGen).

Self-Evolving-Agents by CharlesQ9

1.3%
941
Survey of self-evolving AI agents
Created 7 months ago
Updated 4 months ago
Feedback? Help us improve.