Prometheus  by EuniAI

AI agent for autonomous software development and code repair

Created 1 year ago
428 stars

Top 69.0% on SourcePulse

GitHubView on GitHub
Project Summary

Prometheus is a research-backed AI platform that addresses complex codebase management and software productivity challenges. It targets developers and researchers by providing autonomous code agents that reason over codebases to generate verifiable, affordable software, moving beyond unreliable prototyping.

How It Works

The platform employs a multi-agent system orchestrated by LangGraph state machines, integrated with a unified knowledge graph (Neo4j, Tree-sitter AST) for deep semantic understanding. Specialized agents collaborate to classify issues, reproduce bugs, retrieve context, and generate validated patches, enabling end-to-end automated software repair and feature implementation. This approach offers superior reasoning depth and repository-wide understanding compared to single-agent or less integrated systems.

Quick Start & Requirements

Setup requires Docker, Docker Compose, Python 3.11+, and API keys for OpenAI, Anthropic, or Google Gemini. PostgreSQL and Neo4j are needed for database services. Installation involves cloning the repo (https://github.com/EuniAI/Prometheus), configuring .env with API keys and a JWT secret, and starting services via docker-compose up --build. Access is provided via http://localhost:9002/v1.2 (API) and http://localhost:9002/docs (Interactive Docs). Further details are available in the Multi-Agent Architecture documentation and the Research Paper.

Highlighted Details

  • Automated Issue Resolution: Provides end-to-end bug fixing, including reproduction, patch generation, and multi-level validation.
  • Feature Implementation Pipeline: Handles context-aware feature analysis, planning, and code generation, with optional regression testing.
  • Intelligent Context Retrieval: Utilizes graph-based semantic search across codebase structure, AST, and documentation.
  • Knowledge Graph Integration: Employs a Neo4j-powered unified representation of code structure and semantics for deep reasoning.
  • Containerized Execution: Leverages Docker for isolated build, test, and validation environments.

Maintenance & Community

The project fosters community through a Discord Server, GitHub Issues, and detailed Contributing Guidelines. It acknowledges support from Delysium.

Licensing & Compatibility

Prometheus is dual-licensed: the Community Edition is under GNU GPLv3, requiring derivative works to also be open-sourced. A separate Commercial Edition is available for proprietary, closed-source, or commercial use cases, requiring direct contact with the EUNI.AI Team for licensing terms.

Limitations & Caveats

While not explicitly detailed, the platform's research-backed nature and reliance on advanced components like knowledge graphs and multi-agent systems suggest a potentially steep learning curve and significant setup requirements. The GPLv3 license for the community edition imposes copyleft restrictions on derivative works.

Health Check
Last Commit

5 days ago

Responsiveness

Inactive

Pull Requests (30d)
0
Issues (30d)
3
Star History
241 stars in the last 30 days

Explore Similar Projects

Feedback? Help us improve.