Most AI memory relies on Vector RAG, which lacks structural context. MemoryOS is a hybrid knowledge base that runs Vector Search and Graph Traversal in parallel. Using recursive SQL CTEs, it finds relational links that standard RAG misses—like discovering "HNSW" when you only query "indexing." It features a real-time Retrieval Inspector, UMAP semantic cluster maps, and an MCP server to connect your local knowledge memory directly to any AI assistant like Claude or ChatGPT.

Wispr Flow: Dictation That Works Everywhere — Stop typing. Start speaking. 4x faster.

Stop typing. Start speaking. 4x faster.

Hey hunters! 👋 I built MemoryOS because standard AI memory (RAG) often feels "flat." It understands what your notes sound like (semantic similarity), but it doesn't understand how they relate (structural connectivity). MemoryOS creates a high-fidelity knowledge graph of your data with typed, weighted relations. Every query walks the graph and the vector space simultaneously, using recursive logic to bridge the gap between "what I said" and "what I meant." Key Features: Dual-Path Retrieval: Parallel pgvector search + Recursive BFS graph traversal. Visual X-Ray: A "Retrieval Inspector" that shows you the exact path the AI took through your nodes. AI Bridge: Built-in MCP server so you can use your MemoryOS as the brain for your coding assistants. It's open-source and we'd love to hear your thoughts on the hybrid approach! 🧠🕸️

Wispr Flow: Dictation That Works Everywhere — Stop typing. Start speaking. 4x faster.

Stop typing. Start speaking. 4x faster.

MemoryOS

Hybrid AI Memory: Vector RAG meets Knowledge Graphs

Hybrid AI Memory: Vector RAG meets Knowledge Graphs