Self-hosted alternatives to memory-management

M

memory-management

AgentDB memory system with HNSW vector search

memory-management is an agent skill from the ruflo project for an AgentDB memory system with HNSW vector search. It gives an assistant persistent storage for patterns, semantic search over what it has stored and a knowledge base that it can build over time. The skill describes when to use it: when a successful pattern should be stored, when a similar solution needs to be found, when past work needs a semantic lookup, when the assistant should learn from previous tasks, when knowledge should be shared between agents, and when a knowledge base is being built. It also says when to skip it: for work that needs no learning, for one-off tasks that will not recur, when an external data source already provides the answer, and for read-only exploration. The listing claims 150 to 12,500 times faster pattern retrieval, which is the skill's own figure and should be measured on your data. The skill is useful for teams whose agents repeat similar tasks and would benefit from remembering what worked. Because it stores what the agent learns, check what gets saved, especially if the agent handles private or sensitive material. The skill shows a warning from Gen Agent Trust Hub on skills.sh, and Socket and Snyk pass. It has about 1,200 installs. Install it with npx skills add https://github.com/ruvnet/ruflo --skill memory-management.