Persist and retrieve knowledge across agent runs using semantic embeddings in SQLite.
npx skills add https://github.com/mofa-org/mofa-skills --skill mofa-memory
Please help me install this Agent Skill. Skill: mofa-memory Source: https://github.com/mofa-org/mofa-skills/tree/main/_unpublished/mofa-memory Command: npx skills add https://github.com/mofa-org/mofa-skills --skill mofa-memory
Persistent memory across agent runs using semantic embeddings stored in SQLite. This solves the problem of losing valuable insights once an agent session ends by enabling long-term recall. Data persists locally and can be queried by content, tags, or query similarity.
Store a piece of content with tags, then retrieve related memories by query.
High-intent search queries and answers about installing and using this skill.
You can persist and retrieve knowledge across agent runs by storing content with semantic embeddings in a local SQLite database. This allows agents to recall reports, summaries, and source excerpts during cross-session continuity using cosine similarity queries.
The best way to enable semantic search for an agent memory store is using OpenAI text-embedding-3-small embeddings saved in SQLite. This approach supports retrieving relevant memories by calculating cosine similarity against a natural language query.
Yes, you can manage the memory lifecycle in SQLite using specific operations to store, retrieve, list, clear by tags, or completely delete all memories. This ensures outdated or unwanted cross-session knowledge is easily removable.
Cross-session recall works locally without external vector databases by leveraging SQLite for persistence and OpenAI text-embedding-3-small for generating semantic embeddings. This setup enables efficient similarity searches directly on your local machine.
You need semantic memory retrieval for research workflows when you want to maintain long-term recall of knowledge bases and source excerpts across multiple sessions. It prevents the loss of valuable insights once an individual agent session ends.
Using SQLite with semantic embeddings provides cosine similarity matching rather than exact keyword matches, meaning you can retrieve conceptually related memories across sessions even if the exact phrasing differs from your query.
yoanbernabeuConfigure GrepAI to use OpenAI embeddings for semantic code search.
vineethsomaRetrieve items by comparing query embeddings using cosine similarity, Euclidean distance, or dot product.
ironsheepManage active work sessions in P2-Knowledge-Base with todo-mcp context.
DammianMillerManage a 4-layer memory system using SQLite and Qdrant for AI coding agents.
