Persistent Memory

A local memory system that lives in your workspace and survives conversations. It handles the full lifecycle: init a `.memory/` directory, search with hybrid lexical and semantic retrieval, add facts or preferences when the user says "remember," and sync to check database health. The workflow is built around a CLI (either the Python script or the `pmem` shortcut) and expects you to search at task start, add sparingly, and keep entries concise. It auto-falls back from sqlite-vec to Python cosine if the vector extension isn't available. Useful if you're working in a repo where context actually matters across sessions and you want retrieval that isn't just keyword matching.