Rudi

Rudi

Causal graph memory for LLM.Flat token cost across sessions.

5 followers

Rudi is a context engine for long-running AI sessions. Instead of resending the whole chat history every turn — expensive, hits limits, forgets early decisions — Rudi stores only the decisions that matter as a causal graph in SQLite. The AI reads the graph, not the transcript. Cost stays flat across 1000+ turns. Unlike memory tools that store facts, Rudi tracks consequences: change one decision and it flags exactly what breaks downstream. Open source, local-first.
Rudi gallery image
Free
Launch tags:GitHub
Launch Team / Built With