Documentation
Start here. Pick the path that matches what you are trying to do.
Quickstart
Five minutes from pip install to your first memory.add() and memory.search() call.
Self-hosting
Production deployment guide. SQLite plus FAISS today, Docker image coming next.
Configuration
Providers, env vars, memory config, retrieval modes, YMYL, decay functions.
What widemem is
An open-source Python memory layer for AI agents. Local-first by default (SQLite plus FAISS, zero external services). Importance scoring, temporal decay, hierarchical memory, batch conflict resolution, YMYL prioritization, and confidence-aware retrieval. Apache 2.0.
What it is not
widemem is not a RAG pipeline, a vector database, or a managed service. It is a library you install into your Python agent. It handles the full memory lifecycle (extraction, scoring, storage, retrieval, conflict resolution, decay) so you do not have to.
Canonical sources
The full reference lives in the README. Release history is in the CHANGELOG. Security disclosure policy is at /security.
Need support?
Open-source users get GitHub Issues and a best-effort response. Teams running widemem in production (especially regulated industries) can get a dedicated support contract via the enterprise page. That funds the roadmap.