Asterism – a local-first knowledge graph that grows from your Claude chats (github.com)

Asterism has introduced a groundbreaking local-first personal knowledge graph that functions like a living star map, dynamically evolving from user conversations with Claude. Each interaction creates a trace, allowing the graph to visually represent knowledge: concepts shine brighter with revisits and dim as they are forgotten. Utilizing Hebbian learning principles, connections between ideas strengthen with each traversal while concepts decay if not revisited for a set duration. This innovative design circumvents traditional cloud reliance, with users maintaining control over their data in local SQLite storage. The significance of Asterism lies in its unique approach to personal AI memory systems, emphasizing privacy and interactivity. By implementing session-based time-to-live (TTL) for concepts and utilizing a causal chain topology to organize ideas, Asterism enables a more organic development of knowledge that mimics human thought processes. With features like implicit LLM context injection and triple extraction from dialogues, Asterism could redefine how users engage with AI models, ensuring that responses are rooted in personal history while leaving no digital trace outside the user’s machine. Asterism's commitment to user privacy, lacking telemetry or analytics, makes it a pioneering addition to the AI/ML landscape.