Luamemo

luamemo is a persistent semantic memory store for AI agents and conversational applications, written in pure Lua and designed to run in any Lua 5.1+ or LuaJIT environment. It gives agents long‑term, searchable memory backed by PostgreSQL, combining vector and full‑text search to retrieve relevant context from large or noisy histories. The project is production‑oriented: it ships with a memo CLI for writing, searching, summarizing, and maintaining memories, plus health checks, migrations, and automatic host calibration. It supports multiple embedding and reranking providers (local and remote models) behind a simple configuration layer, and exposes a lightweight knowledge‑graph for “current facts” with temporal validity. A built‑in Model Context Protocol (MCP) server lets editors and tools connect directly to the memory store, so LLM‑powered workflows can use long‑term memory without custom glue code.

View on GitHub