@modusensus/dsh-mneme 0.1.5 → 0.1.6

package/README.md

@@ -3,7 +3,7 @@
 [![npm version](https://img.shields.io/npm/v/@modusensus/dsh-mneme?color=blue&label=npm)](https://www.npmjs.com/package/@modusensus/dsh-mneme)
 [![license](https://img.shields.io/badge/license-MIT-green)](LICENSE)
 [![dsh-plugin](https://img.shields.io/badge/dsh-plugin-awesome-orange)](https://github.com/awesome-dsh-plugin/awesome-dsh-plugin)
-[![tests](https://img.shields.io/badge/tests-129%20passed-success)](https://github.com/modusensus/dsh-mneme)
+[![tests](https://img.shields.io/badge/tests-136%20passed-success)](https://github.com/modusensus/dsh-mneme)
 
 > 给 DeepSeek Harness 的跨会话记忆插件：让 Agent 记住你、记住项目、自动整理记忆。**Mneme**（Μνήμη）——希腊记忆女神 Mnemosyne 之名，掌管记忆与梦境，正如 autoDream 在后台巩固记忆。
 
@@ -22,7 +22,7 @@
 | 工具 | 功能 |
 |------|------|
 | `memory_save` | 记录一条记忆（自动按标题去重合并） |
-| `memory_search` | 全文搜索（中文子串友好） |
+| `memory_search` | 全文搜索（中文子串友好，可启用向量语义搜索） |
 | `memory_list` | 按类型分页列出 |
 | `memory_update` | 修改已有记忆 |
 | `memory_delete` | 删除记忆 |
@@ -45,11 +45,11 @@
 
 ### Web 记忆面板
 
-侧边栏"记忆"按钮 → 模态面板：按类型浏览、全文搜索、查看详情。
+官方设置面板 → 「记忆库设置」→「记忆」标签：按类型浏览、全文搜索；启用向量搜索后可用「语义」切换做向量召回。
 
 ### 用户设置（画像 / 规则）与自定义指令 ⚙️
 
-侧边栏"设置"按钮 → 设置面板：
+官方设置面板 → 「记忆库设置」标签：
 
 - **用户画像**：一段自由文本描述用户自己（角色、背景、偏好），**每轮注入**到系统提示，让 Agent 始终遵循
 - **规则**：Agent 必须遵守的行为规则列表（如"回答先给结论"），同样每轮注入
@@ -57,6 +57,24 @@
 
 > 画像与规则通过独立的 `[用户设置]` 注入区块（优先级高于记忆库），即使记忆为空也会注入。
 
+### 向量搜索（语义搜索）🔎
+
+可选能力：接入 OpenAI 兼容的 embeddings API，让搜索能命中**字面不同但语义相近**的记忆。
+
+**配置**：官方设置 → 「记忆库设置」→ 滚动到底部「向量搜索」区块：
+
+| 字段 | 说明 |
+|------|------|
+| `启用向量搜索` | 总开关；开启后记忆面板出现「语义」切换 |
+| `API 地址 (Base URL)` | OpenAI 兼容端点，如 `https://api.openai.com/v1`；也支持 SiliconFlow、智谱、本地 Ollama 等 |
+| `API Key` | 对应服务的密钥 |
+| `模型名` | embedding 模型，如 `text-embedding-3-small`、`text-embedding-v3`、`bge-m3` 等 |
+
+保存配置后点「重建索引」，为已有记忆批量补建向量（新写入的记忆会自动嵌入）。之后在记忆面板输入查询并点「语义」，即可用向量召回语义相关结果；向量服务不可用时自动回退全文搜索。
+
+> ⚠️ 密钥仅保存在本机 `~/.dsh/memory/memory.db` 的 `user_settings` 表，不会上传，也不会写入代码仓库。
+> 需要 embedding 而非 rerank 模型：如阿里云 `text-embedding-v3` 可用，`qwen3-vl-rerank` 是 rerank 模型（不走 `/embeddings`）。
+
 ## 📦 安装
 
 ### 前置条件
@@ -138,7 +156,7 @@ dsh web
 │  autoDream：阈值调度 → LLM 决策清单               │
 │            → 校验（fail-safe）→ 应用 → 摘要       │
 ├─────────────────────────────────────────────────┤
-│  Web 面板：侧边栏入口 + 浏览/搜索/详情            │
+│  Web 面板：设置面板内嵌 + 浏览/搜索（含向量）    │
 └─────────────────────────────────────────────────┘
 ```
 
@@ -155,12 +173,13 @@ src/
 ├── summarize.js      # 会话结束 LLM 摘要
 ├── dream.js          # autoDream 调度 + runDream（LLM 决策 + 摘要）
 ├── dream/decisions.js# 决策校验（fail-safe）+ 决策应用
+├── embedding.js      # OpenAI 兼容 embeddings 客户端 + 向量检索
 ├── api.js            # HTTP 路由（Web 面板数据通道）
 └── index.js          # 插件接线
 lib/
 ├── client.js         # Web 面板（手写 ModuleLoader bundle）
 └── *.js              # src 的同步分发产物
-test/                 # 129 个 node:test 测试
+test/                 # 136 个 node:test 测试
 ```
 
 ## 🧪 开发
@@ -168,7 +187,7 @@ test/                 # 129 个 node:test 测试
 ```bash
 cd dsh-mneme
 npm install        # 安装 peer 依赖（以 devDependencies 形式，用于本地测试）
-npm test           # 运行 129 个测试（--test-isolation=none 用于受限沙箱，禁止子进程 spawn）
+npm test           # 运行 136 个测试（--test-isolation=none 用于受限沙箱，禁止子进程 spawn）
 npm run sync       # 把 src/ 同步到 lib/（发布时由 prepack 钩子自动执行）
 ```
 

package/lib/index.js

@@ -91,7 +91,7 @@ export const apply = (ctx, config) => {
   });
 
   ctx.inject(["tools"], (toolsCtx) => {
-    disposers.push(createTools(toolsCtx, service, cfg));
+    disposers.push(createTools(toolsCtx, service, cfg, embedder));
   });
 
   const summarizer = createSummarizer(ctx, service, cfg);

package/lib/tools.js

@@ -21,7 +21,7 @@ const MEMORY_ITEM_SCHEMA = {
   }
 };
 
-export function createTools(ctx, service, config) {
+export function createTools(ctx, service, config, embedder) {
   const tools = [
     defineTool({
       name: "memory_save",
@@ -63,10 +63,12 @@ export function createTools(ctx, service, config) {
 
     defineTool({
       name: "memory_search",
-      description: "Full-text search the cross-session memory store. Use when you need past context: how a problem was solved, user preferences, project decisions. Returns matching entries with source and timestamps.",
+      description: "Search the cross-session memory store. Use when you need past context: how a problem was solved, user preferences, project decisions. Substring-matches title/content/tags, and optionally augments results with semantic (vector) recall when an embeddings provider is configured. Returns matching entries with source and timestamps.",
       parameters: {
         query: { type: "string", required: true, description: "Search text; substring match over title/content/tags" },
-        limit: { type: "integer", description: "Max results (default 20)" }
+        limit: { type: "integer", description: "Max results (default 20)" },
+        mode: { type: "string", enum: ["auto", "keyword", "vector"], description: "auto (default) = keyword hits first + vector fill when enabled; keyword = text only; vector = semantic recall first (falls back to keyword)" },
+        semantic: { type: "boolean", description: "Shorthand: enable semantic (vector) recall (same as mode=vector when true)" }
       },
       output: {
         schema: {
@@ -82,7 +84,25 @@ export function createTools(ctx, service, config) {
         render: (_args, value) => TEXT_OUTPUT(`Found ${value.items.length} memory entr${value.items.length === 1 ? "y" : "ies"}.`)
       },
       async execute(args) {
-        const rows = service.toApiList(service.search(args.query, { limit: args.limit ?? 20 }));
+        const limit = args.limit ?? 20;
+        const rows = service.toApiList(service.search(args.query, { limit }));
+        const wantVector = args.mode === "vector" || args.semantic === true || args.mode === "auto";
+        if (wantVector && embedder) {
+          try {
+            const vector = await embedder.embed(args.query);
+            if (vector) {
+              const scored = service.toApiList(service.searchVector(vector, { limit }));
+              const seen = new Set(rows.map((m) => m.id));
+              for (const m of scored) {
+                if (rows.length >= limit) break;
+                if (!seen.has(m.id)) {
+                  seen.add(m.id);
+                  rows.push(m);
+                }
+              }
+            }
+          } catch { /* vector unavailable: keep keyword results */ }
+        }
         return { items: rows };
       }
     }),

package/package.json

@@ -1,7 +1,7 @@
 {
   "name": "@modusensus/dsh-mneme",
-  "description": "Cross-session memory plugin for DeepSeek Harness with autoDream consolidation: SQLite store, Markdown mirrors, 6 model tools, automatic injection, session summarization, and a Web GUI panel",
-  "version": "0.1.5",
+  "description": "Cross-session memory plugin for DeepSeek Harness with autoDream consolidation: SQLite store, Markdown mirrors, 6 model tools, automatic injection, session summarization, user profile/rules, custom slash commands, vector (semantic) search, and a Web GUI panel",
+  "version": "0.1.6",
   "license": "MIT",
   "type": "module",
   "main": "lib/index.js",

package/src/index.js

@@ -91,7 +91,7 @@ export const apply = (ctx, config) => {
   });
 
   ctx.inject(["tools"], (toolsCtx) => {
-    disposers.push(createTools(toolsCtx, service, cfg));
+    disposers.push(createTools(toolsCtx, service, cfg, embedder));
   });
 
   const summarizer = createSummarizer(ctx, service, cfg);

package/src/tools.js

@@ -21,7 +21,7 @@ const MEMORY_ITEM_SCHEMA = {
   }
 };
 
-export function createTools(ctx, service, config) {
+export function createTools(ctx, service, config, embedder) {
   const tools = [
     defineTool({
       name: "memory_save",
@@ -63,10 +63,12 @@ export function createTools(ctx, service, config) {
 
     defineTool({
       name: "memory_search",
-      description: "Full-text search the cross-session memory store. Use when you need past context: how a problem was solved, user preferences, project decisions. Returns matching entries with source and timestamps.",
+      description: "Search the cross-session memory store. Use when you need past context: how a problem was solved, user preferences, project decisions. Substring-matches title/content/tags, and optionally augments results with semantic (vector) recall when an embeddings provider is configured. Returns matching entries with source and timestamps.",
       parameters: {
         query: { type: "string", required: true, description: "Search text; substring match over title/content/tags" },
-        limit: { type: "integer", description: "Max results (default 20)" }
+        limit: { type: "integer", description: "Max results (default 20)" },
+        mode: { type: "string", enum: ["auto", "keyword", "vector"], description: "auto (default) = keyword hits first + vector fill when enabled; keyword = text only; vector = semantic recall first (falls back to keyword)" },
+        semantic: { type: "boolean", description: "Shorthand: enable semantic (vector) recall (same as mode=vector when true)" }
       },
       output: {
         schema: {
@@ -82,7 +84,25 @@ export function createTools(ctx, service, config) {
         render: (_args, value) => TEXT_OUTPUT(`Found ${value.items.length} memory entr${value.items.length === 1 ? "y" : "ies"}.`)
       },
       async execute(args) {
-        const rows = service.toApiList(service.search(args.query, { limit: args.limit ?? 20 }));
+        const limit = args.limit ?? 20;
+        const rows = service.toApiList(service.search(args.query, { limit }));
+        const wantVector = args.mode === "vector" || args.semantic === true || args.mode === "auto";
+        if (wantVector && embedder) {
+          try {
+            const vector = await embedder.embed(args.query);
+            if (vector) {
+              const scored = service.toApiList(service.searchVector(vector, { limit }));
+              const seen = new Set(rows.map((m) => m.id));
+              for (const m of scored) {
+                if (rows.length >= limit) break;
+                if (!seen.has(m.id)) {
+                  seen.add(m.id);
+                  rows.push(m);
+                }
+              }
+            }
+          } catch { /* vector unavailable: keep keyword results */ }
+        }
         return { items: rows };
       }
     }),