Remindb

•
129
•

Manages an agent's long-term memory in a single SQLite file, dramatically reducing LLM token usage for repeated context access.

Remindb provides a persistent, structured memory solution for AI agents, allowing them to intelligently index, search, and manage information without incurring high token costs for repeated reads. By parsing diverse file types into a hierarchical tree, tracking node 'temperature,' and implementing features like incremental diffs and in-band summarization, Remindb transforms raw notes into a dynamic knowledge base that an agent can efficiently learn from and contribute to, all within a portable SQLite database.

主要功能

01Portable and self-contained memory in a single SQLite file, easily shared across agents and devices.
02Structured memory tree with typed nodes for efficient indexing and navigation.
03Temperature-based caching that ranks nodes as 'hot' or 'cold' for intelligent retrieval and summarization.
040 GitHub stars
05Git-style versioning and incremental diffs for efficient memory updates and history tracking.
06FTS5 full-text search with token budgeting, returning ranked anchors in milliseconds.

使用案例

01Providing long-term, indexed memory for AI coding agents to reduce token costs.
02Enabling agents to track and summarize evolving project documentation and notes.
03Facilitating collaborative agent workflows with shared, versioned memory databases.