DevFlow

소개

DevFlow is an intelligent memory system that equips AI agents, particularly those supporting the Model Context Protocol, with a persistent and searchable knowledge graph. Unlike complex enterprise solutions or simplistic prototypes, DevFlow offers an out-of-the-box experience by storing its entire knowledge graph in a single SQLite file, requiring no external databases or complex configurations. It's built by developers for developers, understanding the intricacies of software projects, and designed to prevent AI assistants from repeatedly forgetting crucial project context. The system is battle-tested on real projects, ensuring reliability and effectiveness in managing information flow for AI-driven development.

주요 기능

• Temporal awareness with full version history, allowing point-in-time graph retrieval for entities and relations.
• Workflow-aware prompts that guide AI agents through cascading development cycles (planner, task creator, coder, reviewer).
• Advanced semantic search capabilities using vector embeddings, cosine similarity, and hybrid search.
• 0 GitHub stars
• Configurable confidence decay for relations, automatically adjusting relevance over time if not reinforced.
• Zero-setup, single-file SQLite database for all knowledge graph data, including vector embeddings.

사용 사례

• Maintaining persistent project context and history for AI assistants in software development.
• Guiding AI agents through structured development workflows using context-aware prompts.
• Facilitating semantic retrieval of relevant project information, dependencies, and design decisions across sessions.