Engineering Notes
On agent memory: memory layers vs vector databases, air-gapped self-hosting, and the memory lifecycle. 中文社区版本将陆续补充。 | 工程笔记：Agent 记忆层、气隙自托管与记忆生命周期。
Why Your AI Agent Needs a Long-Term Memory Layer (Not Just a Vector Database)
A vector database stores and retrieves vectors. A memory layer manages what your agent remembers, what it forgets, and what it distills — inside your own boundary. The difference is not academic: it decides whether your agent gets smarter over time or just accumulates a bigger index.
Self-Hosting Your Agent's Memory, Air-Gapped in Five Minutes
One Docker command. Embedded BGE-M3 embeddings. Zero external API calls — pull the cable and it keeps working. A concrete walkthrough of running the VecminDB memory layer entirely inside your own network.
Giving a RAG Agent an Honest Memory — A Case Walkthrough
How decay, distillation and correction change a retrieval agent's behavior over weeks of use. In the meantime, see the RAG memory scenario guide.