maasv-memory
ascottbell 제작
Coding Agents & IDEs요약
Structured long-term memory for OpenClaw agents, powered by [maasv](https://github.com/ascottbell/maasv).
설명
이 스킬이 하는 일
A self-hosted memory backend for OpenClaw agents that uses three retrieval signals: semantic similarity, keyword matching, and a knowledge graph. Includes entity extraction, temporal versioning, and experiential learning. All data lives in a SQLite file on your local machine; nothing is sent to any maasv service.
Combines semantic, keyword, and graph-based retrieval in a single self-hosted system, so agents get richer memory recall without sending conversation data to a third-party service.
사용 시점
- Recalling project decisions made in past agent sessions
- Tracking entities and relationships across long-running workflows
- Running agent memory with no cloud data storage
- Searching past reasoning and outcomes when facing similar problems
- Persisting conversation context between separate agent invocations
예시 워크플로
User asks: 'Remember that we decided to use PostgreSQL for this project because of the JSON query requirements'
- 1
Calls memory_store with the decision summary
- 2
maasv server runs entity extraction, identifying 'PostgreSQL' and 'JSON query requirements' as entities
- 3
Entities and relationships are written to the local SQLite database
- 4
In a later session, autoRecall queries the memory store
- 5
Injects the relevant context into the agent before it responds
Memory stored locally and surfaced automatically in future sessions when the topic is relevant
요구 사항
- Anthropic API key or OpenAI API key for entity extraction (MAASV_LLM_PROVIDER + MAASV_ANTHROPIC_API_KEY or MAASV_OPENAI_API_KEY)
- Voyage API key, OpenAI API key, or Ollama running locally for embeddings (MAASV_EMBED_PROVIDER + MAASV_VOYAGE_API_KEY if using Voyage)
설치
clawhub install ascottbell/maasv-memoryCLI 명령을 실행하거나, 프롬프트를 Claude Code, Cursor, Codex 등 사용하는 코딩 에이전트에 붙여넣으세요.