Production memory system that gives Claude persistent recall across conversations using LanceDB with hybrid vector and BM25 search. Ships with four deployment paths: full Jina embedding plus reranker for best quality, a budget mix with free SiliconFlow reranking, OpenAI-only for simplicity, or fully local with Ollama if you want zero API calls. Smart Extraction uses an LLM to pull structured facts from conversation history, and there's Weibull decay so old memories fade naturally. The workflow is thorough, maybe overly so. It actually checks your API keys with curl before touching config, which saves you from the usual "why isn't this working" debug loop. If you need Claude to remember things between sessions and want control over the retrieval stack, this handles it.
npx skills add https://github.com/cortexreach/memory-lancedb-pro-skill --skill memory-lancedb-pro