写入一条记忆短文本（事实/笔记/摘要），可选标签；返回 {"id": 记录ID}。 内容为空串或纯空白时返回 {"error": "INVALID_ARGUMENT", "message": 原因}。 project 指定项目子库（空=全局库）；client 仅用于审计归类（自动识别）； 记录主键由服务端生成。 English: Write a memory short text (fact/note/summary) with optional tags; returns {"id": record_id}. Returns {"error": "INVALID_ARGUMENT", "message": reason} when content is empty or blank. project selects the project sub-store (empty = global); client is audit-bucketing only (auto-detected). The record primary key is generated server-side.

混合检索记忆与知识（向量语义 + BM25 关键词，归一化融合）； 返回命中列表，每项含 id/content/score/type/source。 project 指定项目子库时，检索范围 = 该项目子库 + 全局库（项目命中加分）； 不传 project 时只检索全局库。 top_k 小于 1 时返回 {"error": "INVALID_ARGUMENT", "message": 原因}。 client：来源客户端（可选，缺省从 clientInfo 自动识别，仅审计）。 English: Hybrid retrieval over memories and knowledge (vector semantics + BM25 keywords, normalized fusion); returns a hit list, each item having id/content/score/type/source. With project, the scope is that project's sub-store plus the global store (project hits boosted); without project, only the global store is searched. Returns {"error": "INVALID_ARGUMENT", "message": reason} when top_k is less than 1. client: source client (optional; auto-detected from clientInfo when omitted; audit only).

按 ID 读取单条记忆完整内容；记录不存在返回 {"error": "NOT_FOUND"}。 project：先查该项目子库，再查全局库；不传时先查全局库再查各项目子库。 client：来源客户端（可选，缺省从 clientInfo 自动识别，仅审计）。 English: Read the full content of a single memory by ID; returns {"error": "NOT_FOUND"} when absent. project: search that project's sub-store first, then the global store; when omitted, global first and then each project store. client is audit-only.

按 ID 更新记忆内容（变更后自动重新嵌入）；记录不存在返回 {"error": "NOT_FOUND"}； 内容为空串或纯空白时返回 {"error": "INVALID_ARGUMENT", "message": 原因}。 project：先查该项目子库，再查全局库；不传时先查全局库再查各项目子库。 client：来源客户端（可选，缺省从 clientInfo 自动识别，仅审计）。 English: Update a memory's content by ID (auto re-embed on change); returns {"error": "NOT_FOUND"} when absent; empty content returns {"error": "INVALID_ARGUMENT", "message": reason}. project: locate in that project's sub-store, then the global store; when omitted, global first and then each project store. client is audit-only.

按 ID 删除一条记忆；成功返回 {"ok": true}，记录不存在返回 {"error": "NOT_FOUND"}。 project：先查该项目子库，再查全局库；不传时先查全局库再查各项目子库。 client：来源客户端（可选，缺省从 clientInfo 自动识别，仅审计）。 English: Delete a memory by ID; returns {"ok": true} on success and {"error": "NOT_FOUND"} when it does not exist. project: locate in that project's sub-store then the global store; when omitted, global first and then each project store. client is audit-only.

建立两条记忆的关联（related/supersedes/depends_on/part_of/contradicts）； 重复建边幂等（created=false）。仅同作用域的 memory 记录可关联，禁止自关联； 失败返回 {"error": , "message": 原因}（NOT_FOUND/SELF_LINK/...）。 project：项目子库（空=全局库）；client：仅审计。 English: Link two memories (related/supersedes/depends_on/part_of/contradicts); duplicate edges are idempotent (created=false). Only memory records in the same scope can be linked, and self-links are rejected; failures return {"error": , "message": reason}.

取消两条记忆的关联（任一方向）；relation 为空时删除该对全部关系。 返回 {"ok": true, "removed": 删除数}；记录不存在返回 {"error": "NOT_FOUND"}。 English: Remove links between two memories in either direction; when relation is empty all relations of the pair are removed. Returns {"ok": true, "removed": n}.

列某条记忆的一跳邻居，返回 {"record_id", "links"}；每项含 id/content/type/tags/relation/direction（out/in/both）。 记录不存在返回 {"error": "NOT_FOUND"}。 English: List one-hop neighbors of a memory as {"record_id", "links"}; each item has id/content/type/tags/relation/direction (out/in/both).

导入本地文档（PDF/DOCX/MD/TXT 及 Office 格式）切分入库； 返回 {"source": 文件名, "chunks": 块数}；文件不存在或格式不支持时 返回 {"error": "FILE_NOT_FOUND" | "UNSUPPORTED_FORMAT", "message": 原因}。 project：文档 chunk 落入该项目子库（空=全局库）；client：仅用于审计归类。 English: Import a local document (PDF/DOCX/MD/TXT and Office formats), split and ingest it; returns {"source": filename, "chunks": count}; when the file is missing or the format is unsupported returns {"error": "FILE_NOT_FOUND" | "UNSUPPORTED_FORMAT", "message": reason}. project: chunks go to that project's sub-store (empty = global); client is audit-only.

抓取网页正文并切分入库；返回 {"source": url, "chunks": 块数}； 抓取/正文提取失败时返回 {"error": "WEB_FETCH_FAILED", "message": 原因}。 project：web chunk 落入该项目子库（空=全局库）；client：仅用于审计归类。 English: Fetch a webpage body and ingest it after splitting; returns {"source": url, "chunks": count}; on fetch/body-extraction failure returns {"error": "WEB_FETCH_FAILED", "message": reason}. project: chunks go to that project's sub-store (empty = global); client is audit-only.

基于知识库的 RAG 问答（检索 → 上下文拼装 → 护栏生成），返回 answer 与 sources； project：检索该项目子库 + 全局库（空=仅全局库），并写 ask 存取审计； client：来源客户端（可选，缺省从 clientInfo 自动识别，仅审计）； LLM 不可用时返回 {"error": "LLM_DISABLED", "message": 配置指引}。 English: Knowledge-base RAG Q&A (retrieve → build context → guarded generation), returning answer and sources; project: search that project's sub-store plus the global store (empty = global only); an ask access-audit is emitted; client is audit-only; returns {"error": "LLM_DISABLED", "message": setup guidance} when the LLM is unavailable.