mnemosyne-ollama drops zero-config local code search
mnemosyne-ollama is an open-source bridge connecting Ollama to the Mnemosyne retrieval engine for zero-config local code search. It uses a 6-tiered hybrid retrieval system to provide models with accurate code citations and ranked chunks without external vector databases.
Mnemosyne's shift away from vector embeddings towards structural signals like AST-aware compression is a smart move for codebase RAG.
* Uses 6-tiered hybrid retrieval instead of just vector search.
* Zero-dependency bridge (210 lines of code) makes it extremely lightweight.
* MCP integration allows it to work with both local models (via Ollama) and frontier models (via Claude Code).
* AST-aware compression significantly reduces token usage while keeping code meaning intact.
DISCOVERED
7d ago
2026-04-04
PUBLISHED
7d ago
2026-04-04
RELEVANCE
AUTHOR
Enough_Leopard3524