@a9i5k4/dsh-auto-memory 0.1.32 → 0.1.34

package/README.md

@@ -220,6 +220,18 @@ Then edit `package.json` in that directory and append to the `dsh.profile.bundle
 
 Restart **dsh web** (the 「Memory」entry appears in the sidebar).
 
+### Semantic engine (optional but recommended)
+
+The built-in JS semantic tier (e5-small q8, ~130MB) needs the `@huggingface/transformers` inference library, installed automatically as an optional dependency of the main package. If your pnpm security policy blocked its native scripts (you see `ERR_PNPM_IGNORED_BUILDS` / `Ignored build scripts: onnxruntime-node, sharp`), approve and reinstall once:
+
+```bash
+# approve the onnxruntime-node / sharp native install scripts, then reinstall transformers
+pnpm approve-builds
+pnpm add @huggingface/transformers
+```
+
+Restart `dsh web` — the welcome tour's semantic-engine step auto-detects readiness (SHA256 verify + inference self-test). Lexical BM25 (0GB) always works as a fallback; skipping the engine only lowers recall precision.
+
 > No pnpm? `npm install @a9i5k4/dsh-auto-memory` works the same.
 > pnpm v11 blocks packages published <1 day ago: set `minimumReleaseAge: 0` in pnpm-workspace.yaml or pin an explicit version for same-day updates.
 

package/README.zh-CN.md

@@ -230,6 +230,18 @@ cd ~/.dsh/profiles/web
 pnpm add @a9i5k4/dsh-auto-memory
 ```
 
+### 语义引擎（可选但推荐）
+
+内置 JS 语义档（e5-small q8 ~130MB）需要推理库 `@huggingface/transformers`（随主包作为可选依赖自动安装）。若你的 pnpm 因安全策略拦截了它的原生脚本（提示 `ERR_PNPM_IGNORED_BUILDS` / `Ignored build scripts: onnxruntime-node, sharp`），执行一次批准后重装即可：
+
+```bash
+# 批准 onnxruntime-node / sharp 的原生安装脚本,再重装 transformers
+pnpm approve-builds
+pnpm add @huggingface/transformers
+```
+
+装完重启 `dsh web`，向导的语义引擎步会自动检测到就绪（SHA256 校验 + 推理自检）。词法检索 0GB 永远兜底，不装也能用（仅召回精度较低）。
+
 然后编辑同目录 `package.json`，在 `dsh.profile.bundles` 数组追加：
 
 ```json

package/lib/client.js

@@ -492,6 +492,9 @@ window.__ModuleLoader__.load({
         '修复:更新弹窗在小卡形态下内容被裁剪、底部按钮不可达导致无法关闭——新增右上 ✕ 关闭(同步已读版本),内容区自动滚动。',
         '改进:「唤起回顾」决策↔投递时间线——每条决策标注真实投递结果(✓投递×N/未投递/技能✓),判定队列汇总与政策提示;记忆固化(stale 门降版本容忍,语料升版不再整单压制召回)。',
         '补丁 0.1.31:发布物运行时 section 名 dsh:m6-reference-tail → 裸名(发布身份统一,不再带预览标识)。',
+        '补丁 0.1.32:「记忆唤起」开关与设置页双向联动(导览与设置走同一 semantic-emit 接口);语义引擎分级引导(无引擎自动装 JS 档,发烧友可选 Python)。',
+        '补丁 0.1.33:修复语义引擎下载 404(镜像 URL 重复前缀);transformers 推理库声明为可选依赖并随包分发。',
+        '补丁 0.1.34:修复推理库检测——pnpm 把 transformers 提升到 profile 根 node_modules,检测补该路径,新用户装完即用,无需手动操作。',
       ], en: [
         '★ MAJOR: Brand-new Welcome Tour — after first launch or upgrade, every feature is introduced step by step with per-feature switches right in the tour (association / snapshot / greeting / night unattended / reflection / summaries / external memory / skill promotion…); engine detection, download, self-test and live external-source scanning are all inline.',
         '★ New brand visuals: an Office/Fluent-style liquid-glass app icon family — each step gets its own colored glass squircle icon (cyan inject / amber greeting / green calendar / violet engine / sky radar / coral finish) with dedicated looping motion (bell sway, page flip, linked rings, prism spin, radar sweep, rising spark); the signature three-slab frosted-glass logo stays on the Core step.',
@@ -502,6 +505,9 @@ window.__ModuleLoader__.load({
         'Fix: The update dialog could become impossible to close when its content overflowed the compact card — added an ✕ in the top-right (marks version seen) and made the content area scrollable.',
         'Improved: Recall-review decision↔delivery timeline (delivered×N / not delivered / skill badges per decision), review-queue digest with policy hints, and consolidation stale-gate version tolerance so corpus version bumps no longer suppress recall.',
         'Patch 0.1.31: release runtime section name dsh:m6-reference-tail → bare name (unified release identity, no preview marker).',
+        'Patch 0.1.32: the "Memory recall" switch now syncs bidirectionally with Settings (tour and settings share the semantic-emit endpoint); tiered engine onboarding (auto-install JS tier, Python optional for enthusiasts).',
+        'Patch 0.1.33: fixed semantic-engine download 404 (duplicate prefix in mirror URL); transformers inference lib declared as an optional dependency and shipped with the package.',
+        'Patch 0.1.34: fixed inference-lib detection — pnpm hoists transformers to the profile-root node_modules; detection now covers that path, so fresh installs work out of the box.',
       ] },
       '0.1.29': { zh: [
         '修复:工作区总览一直显示「未发现带记忆的工作区」——DSH 新版会话持久化改为 session.jsonl.zstd 压缩帧,旧逻辑只认 .jsonl 扫描不到任何会话;现已用 zstd 解压读取首行提取 cwd,并让空的 summary 缓存短时失效,避免空结果被永久固化。',
@@ -2547,14 +2553,14 @@ window.__ModuleLoader__.load({
       }, [tourShowing])
       useEffect(function () {
         if (!tourShowing) return undefined
-        if (tourDlPhase !== 'downloading' && tourDlPhase !== 'verifying') return undefined
+        if (wizSt.ready) return undefined // 已就绪无需轮询
         var iv = setInterval(function () {
           fetch('/api/dsh-auto-memory/semantic-status').then(function (r) { return r.json() }).then(function (j) {
             Object.assign(wizSt, j); try { dlgTick() } catch (e10) {}
           }).catch(function () {})
         }, 1500)
         return function () { clearInterval(iv) }
-      }, [tourShowing, tourDlPhase])
+      }, [tourShowing, tourDlPhase, wizSt.ready])
       // 无语义引擎自动引导:进入引擎下载步且引擎未就绪/未下载/未出错时,自动开始下载内置 JS 档(C2 默认)。
       // 用户可稍后在设置页手动装 Python 进阶档;词法检索 0GB 永远兜底,不阻塞。
       useEffect(function () {
@@ -2728,7 +2734,9 @@ window.__ModuleLoader__.load({
           : dl.phase === 'verifying' ? t('dlVerifying')
           : dl.phase === 'downloading' ? (t('dlDownloading') + ' · ' + fmtMB(dl.bytesDone || 0) + ' / ' + fmtMB(dl.bytesTotal || totalBytes))
           : dl.phase === 'error' ? (t('dlError') + ': ' + String(dl.error || '').slice(0, 90))
-          : dl.phase === 'done' ? t('dlDone')
+          : dl.phase === 'done' ? (wizSt.assetPresent && !wizSt.peerPresent
+              ? (locale === 'zh' ? '模型已下载,推理库缺失——先 pnpm approve-builds(批准原生脚本),再 pnpm add @huggingface/transformers,重启生效' : 'Model downloaded, inference lib missing — run pnpm approve-builds, then pnpm add @huggingface/transformers, restart')
+              : t('dlDone'))
           : (locale === 'zh' ? '未下载 · 词法检索照常可用' : 'Not downloaded · lexical search keeps working')
         var startDl = function () {
           apiPost('/api/dsh-auto-memory/semantic-download', { action: 'start', mirror: 'auto' }).catch(function () {})
@@ -3269,7 +3277,7 @@ window.__ModuleLoader__.load({
             else if (dl.phase === 'error') { stateTxt = t('dlError'); stateBg = 'rgba(196,74,74,.22)' }
             else if (dl.phase === 'cancelled') { stateTxt = t('dlCancelled'); stateBg = 'rgba(196,138,42,.2)' }
             else if (isJs && dlActive) { stateTxt = dl.phase === 'verifying' ? t('dlVerifying') : t('dlDownloading'); stateBg = 'rgba(36,86,196,.2)' }
-            else if (isJs && sem.assetPresent && !sem.peerPresent) { stateTxt = locale === 'zh' ? '模型已存在，缺运行库' : 'Model present, runtime missing'; stateBg = 'rgba(196,138,42,.2)' }
+            else if (isJs && sem.assetPresent && !sem.peerPresent) { stateTxt = locale === 'zh' ? '模型已存在,缺推理库——pnpm approve-builds 后 pnpm add @huggingface/transformers' : 'Model present, runtime missing — pnpm approve-builds && pnpm add @huggingface/transformers'; stateBg = 'rgba(196,138,42,.2)' }
             else { stateTxt = locale === 'zh' ? '未下载' : 'Not downloaded'; stateBg = 'rgba(196,138,42,.2)' }
             var progress = dlActive || (isJs && dl.phase === 'done')
               ? Math.min(100, Math.round(((dl.bytesDone || 0) / Math.max(1, dl.bytesTotal || totalJs)) * 100))

package/lib/index.js

@@ -4255,6 +4255,8 @@ export function apply(ctx, config) {
       const modelOnnx = cands.find((c) => fsMod.existsSync(c)) || cands[0]
       const peerCands = [
         path.join(pluginDir, 'node_modules', '@huggingface', 'transformers'),
+        // pnpm 默认把 peer/optional 依赖提升到 profile 根 node_modules(pluginDir 上两级)
+        path.join(pluginDir, '..', '..', '..', 'node_modules', '@huggingface', 'transformers'),
         path.join(pluginDir, '..', 'artifacts', 'm7-live-pre', 'js-semantic-trial', 'node_modules', '@huggingface', 'transformers'),
       ]
       const peerDir = peerCands.find((c) => fsMod.existsSync(c)) || peerCands[0]

package/lib/semantic-js.js

@@ -76,6 +76,8 @@ function defaultModelsDirCandidates(pluginDir) {
 function defaultPeerDirCandidates(pluginDir) {
   return [
     path.join(pluginDir, 'node_modules', '@huggingface', 'transformers'),
+    // pnpm 默认把 peer/optional 依赖提升到 profile 根 node_modules(pluginDir 上两级)
+    path.join(pluginDir, '..', '..', '..', 'node_modules', '@huggingface', 'transformers'),
     path.join(pluginDir, '..', 'artifacts', 'm7-live-pre', 'js-semantic-trial', 'node_modules', '@huggingface', 'transformers'),
   ]
 }
@@ -282,7 +284,10 @@ export function createSemanticDownloaderPre(opts = {}) {
   function tmpDir() { return path.join(modelsRoot, '.tmp-dl') }
 
   async function fetchToFile(fileRec, base) {
-    const res = await doFetch(base + fileRelUrl(fileRec.rel), { redirect: 'follow' })
+    // 镜像 base 已含模型子目录(Xenova/multilingual-e5-small/resolve/main/),
+    // rel 带 multilingual-e5-small/ 前缀用于落位;URL 拼接时剥掉前缀避免路径重复 404。
+    const urlRel = fileRec.rel.indexOf(E5_MODELS_SUBDIR + '/') === 0 ? fileRec.rel.slice(E5_MODELS_SUBDIR.length + 1) : fileRec.rel
+    const res = await doFetch(base + fileRelUrl(urlRel), { redirect: 'follow' })
     if (!res.ok) throw new Error('http-' + res.status)
     if (!res.body) throw new Error('no-body')
     const total = Number(res.headers.get('content-length')) || fileRec.bytes

package/package.json

@@ -1,7 +1,7 @@
 {
   "name": "@a9i5k4/dsh-auto-memory",
   "description": "主动联想记忆插件:记忆不靠模型调用,自己被唤回——Host 观察情境,NLP+向量化双轨检索,固定边界注入不破坏前缀缓存。三层记忆自动沉淀、欢迎向导、唤起回顾、无人值守、AI 问候与反思、日历、跨工具记忆继承。Proactive associative memory for DSH: zero-call recall, three-layer auto-consolidation, welcome tour, unattended mode.",
-  "version": "0.1.32",
+  "version": "0.1.34",
   "type": "module",
   "main": "lib/index.js",
   "exports": {
@@ -42,5 +42,8 @@
   "peerDependencies": {
     "@deepseek-ai/cordis": "^4.0.1"
   },
+  "optionalDependencies": {
+    "@huggingface/transformers": "^3.7.6"
+  },
   "license": "BSD-3-Clause"
 }