Smlay_ero
GitHub 上目前公开的两个仓库。星标从 GitHub 读取。
The two public repositories on GitHub. Star counts are fetched live.
多 Agent 共用记忆时，密钥留在用户手里。每个 Agent 只能解密被授权的部分； 同步节点只看到密文；撤销授权后，新数据对该 Agent 不可读。分类和搜索在本地完成。
Shared memory across agents, with keys held by the user. Each agent decrypts only its grant. The sync node sees ciphertext. After a revoke, new data is unreadable to that agent. Classification and search stay local.
本地 LLM 训练编排。你提供基座模型和领域数据，用自然语言写目标，只跑一种模式 （CPT / SFT / DPO 等）。达标就导出，可选 vLLM；达不到就停，并说明做不到。不附带默认模型。
A local training orchestrator. You provide the base model and domain data, describe the goal, and run one mode (CPT / SFT / DPO, and others). Export if it passes, with optional vLLM. If it cannot, it stops and says so. No default model is bundled.