Store and retrieve agent memories via embedded vectors and HNSW search.
npx skills add https://github.com/Moerae-AI/Moerae --skill moerae-ai-memory
Please help me install this Agent Skill. Skill: Moerae - AI Memory Source: https://github.com/Moerae-AI/Moerae/tree/main Command: npx skills add https://github.com/Moerae-AI/Moerae --skill moerae-ai-memory
Moerae - AI Memory reduces the effort of building and maintaining an external, manually indexed memory layer by letting you store content once and retrieve it later through a simple put-and-search workflow, while the system handles embedding, indexing, and lifecycle management.
Store a memory in a project by asking: “Moerae, put into project myproject the fact 'The API rate limit is 1000 req/min' and make it searchable later.”
High-intent search queries and answers about installing and using this skill.
You can build semantic memory for AI agents by using a put-and-search workflow that turns content into embedded vectors, stores them in SQLite, and retrieves relevant nodes via HNSW vector search. This handles indexing and lifecycle management automatically.
For large text ingestion, semantic memory systems enforce a max_indexable_tokens threshold and automatically switch to embedding metadata summaries only, while preserving the full content for later retrieval to avoid exceeding token limits.
Yes, you need a local GGUF model to generate embeddings for vector search. The system stores these embedded nodes and segment state locally in SQLite, ensuring data remains isolated without external API dependencies.
AI agent memory supports hierarchical scoping by isolating searches within a single conversation by default. You can promote specific memory nodes to search promoted content across the entire project scope.
Managing stored vector nodes requires applying segment lifecycle rules to handle promotion and eviction. This ensures your SQLite database maintains relevant semantic memory nodes without manual context-window management or cleanup.
heysuhasManage vector embeddings and metadata in a local-first database for semantic search and RAG applications.
whiteboardevClassify user queries to retrieve memories via recent, keyword, or semantic search.
FrekiManagarmStore, retrieve, and consolidate conversational patterns in AgentDB.
davpatel605-beepStore and retrieve vector embeddings for semantic search and document retrieval.
