Lossless-memory: a personal AI memory that never summarizes
Summary
Lossless-memory is an open-source project that provides a personal AI with lossless long-term memory by storing every utterance verbatim with timestamps and avoiding summarization. It uses a Temporal Backbone for time-first search over JSONL logs, an exact search index with SQLite FTS5, and an LLL topic index injected into the model context to preserve context across sessions. The system runs locally on a single machine, is designed for one user, and emphasizes offline operation and explainable memory traces, with documentation and quickstart guidance available on GitHub.