Launched this week
Reference
Local semantic search for AI agents
51 followers
Local semantic search for AI agents
51 followers
Reference is local semantic search for your files and code, built for AI agents. No cloud, nothing leaves your machine. Ask it "how did I implement rate limiting here" and get your actual code back, cited down to the exact function, not generic advice. Live index that updates as you save, code-aware chunking (tree-sitter), and a built-in MCP server (/search, /explain, /find_similar, /check_doc_drift)so Claude Code gets precise cited results instead of burning tokens on grep loops.
Reference
Built this after burning tokens and context for every new Claude thread I open. An embedding model uses a fraction of the memory a local LLM does, and gives me back what I (or Claude) are looking for instantly. It's local, offline, and now Claude can just ask the index directly. Would love to hear what you think!
DataBlur
Local + cited-to-the-exact-function is the right combo. Which embedding model runs locally, and how large can an indexed codebase get before search latency starts to hurt?
Reference
@kosta_zanin26 Thanks! Default is all-MiniLM-L6-v2 via Candle, running on Metal. A couple other models are selectable in app if you want more accuracy over speed.
On latency, it's a full in-memory scan, no ANN index, scales at about 0.5ms per 1k rows. A ~5k file codebase lands around 20ms, which you basically can't feel next to the embedding step itself. Starts to matter past ~200k rows, haven't needed to solve for that yet.