@ggboy123/dsh-paper-reader 0.2.1 → 0.6.1

package/dist/translate.js

@@ -7,6 +7,8 @@ import { existsSync } from 'node:fs';
 import { mkdir, rename, rm } from 'node:fs/promises';
 import { glob } from 'node:fs/promises';
 import { join, dirname, basename, extname } from 'node:path';
+import { dshHome } from "./library.js";
+import { ensureBabeldoc } from "./babeldoc-install.js";
 export function zhPdfPath(ref) {
     return ref.pdfPath.slice(0, -extname(ref.pdfPath).length) + '-zh.pdf';
 }
@@ -14,7 +16,8 @@ export function dualPdfPath(ref) {
     return ref.pdfPath.slice(0, -extname(ref.pdfPath).length) + '-dual.pdf';
 }
 /** 进程内每篇文献一个翻译任务（幂等：进行中/已生成直接返回）。 */
-const busy = new Map(); // key=pdfPath, value=error|null(进行中)
+// key=pdfPath；null=进行中（无阶段说明）；{phase}=进行中；{error}=已失败
+const busy = new Map();
 export function zhStatus(ref) {
     const zh = existsSync(zhPdfPath(ref));
     const dual = existsSync(dualPdfPath(ref));
@@ -22,8 +25,9 @@ export function zhStatus(ref) {
     return {
         zh,
         dual,
-        busy: b === null,
-        ...(typeof b === 'string' ? { error: b } : {}),
+        busy: b === null || Boolean(b && !b.error),
+        ...(b?.error ? { error: b.error } : {}),
+        ...(b && !b.error && b.phase ? { phase: b.phase } : {}),
     };
 }
 /** 找 babeldoc 可执行文件：pdfqa venv（文献库同级 .venv-pdf2zh）→ PATH → miniconda。 */
@@ -76,6 +80,8 @@ function babeldocFlags(bin) {
 /**
  * 启动 babeldoc 后台翻译。幂等：已生成或进行中直接返回 false 表示未新启动。
  * 完成/失败结果写进 busy map，由 zhStatus 暴露。
+ * babeldoc 缺失时不再报错：后台自动走 uv 安装链路（首次约几分钟），
+ * 期间 zhStatus.phase 展示安装进度，装完直接接着翻译。
  */
 export async function startTranslation(ref, dataDir, endpoint) {
     const st = zhStatus(ref);
@@ -83,16 +89,15 @@ export async function startTranslation(ref, dataDir, endpoint) {
         return { started: false, reason: 'already-exists' };
     if (st.busy)
         return { started: false, reason: 'busy' };
-    const pre = await precheck(dataDir, endpoint);
+    const pre = precheck(endpoint);
     if (pre)
         return pre;
-    return launch(ref, dataDir, endpoint);
+    const firstRun = !(await findBabeldoc(dataDir));
+    void runPipeline(ref, dataDir, endpoint);
+    return { started: true, firstRun };
 }
-/** 端点/babeldoc 预检：不齐全时返回与 startTranslation 同形的失败结果。 */
-async function precheck(dataDir, endpoint) {
-    const bin = await findBabeldoc(dataDir);
-    if (!bin)
-        return { started: false, reason: '未找到 babeldoc（pip install babeldoc，或复用 pdfqa 的 .venv-pdf2zh）' };
+/** 端点预检：不齐全时返回与 startTranslation 同形的失败结果。 */
+function precheck(endpoint) {
     if (!endpoint.baseUrl || !endpoint.apiKey || !endpoint.model) {
         // code 供前端判定「该弹配置表单了」，别让它去匹配中文文案
         return {
@@ -103,27 +108,42 @@ async function precheck(dataDir, endpoint) {
     }
     return null;
 }
+/** 后台流水线：确保 babeldoc 可用（缺失则自动安装）→ 拉起翻译。 */
+async function runPipeline(ref, dataDir, endpoint) {
+    const setPhase = (phase) => busy.set(ref.pdfPath, { phase });
+    try {
+        setPhase('正在检查翻译引擎…');
+        const bin = await ensureBabeldoc(dataDir, dshHome(), () => findBabeldoc(dataDir), setPhase);
+        await launch(ref, dataDir, endpoint, bin);
+    }
+    catch (e) {
+        const msg = e instanceof Error ? e.message : String(e);
+        console.error('[dsh-paper-reader] babeldoc 自动安装失败:', msg);
+        busy.set(ref.pdfPath, { error: `翻译引擎安装失败：${msg}` });
+    }
+}
 /**
  * 重新翻译：删除已有译文产物后再启动（换模型/翻得不好时用）。
- * 先预检再删——端点不齐或 babeldoc 缺失时不动旧译文，免得删完才发现跑不起来。
+ * 先预检再删——端点不齐时不动旧译文，免得删完才发现跑不起来。
  * babeldoc 的段落缓存 key 含模型名，换模型自然失效，无需 --ignore-cache。
  */
 export async function restartTranslation(ref, dataDir, endpoint) {
     if (zhStatus(ref).busy)
         return { started: false, reason: 'busy' };
-    const pre = await precheck(dataDir, endpoint);
+    const pre = precheck(endpoint);
     if (pre)
         return pre;
     await rm(zhPdfPath(ref), { force: true });
     await rm(dualPdfPath(ref), { force: true });
-    return launch(ref, dataDir, endpoint);
+    const firstRun = !(await findBabeldoc(dataDir));
+    void runPipeline(ref, dataDir, endpoint);
+    return { started: true, firstRun };
 }
-/** 真正拉起 babeldoc 子进程（precheck 已通过、产物不存在/已删）。 */
-async function launch(ref, dataDir, endpoint) {
-    const bin = (await findBabeldoc(dataDir));
+/** 真正拉起 babeldoc 子进程（bin 已就绪、产物不存在/已删）。 */
+async function launch(ref, dataDir, endpoint, bin) {
     const tmpDir = join(dataDir, '.pdf2zh-tmp');
     await mkdir(tmpDir, { recursive: true });
-    busy.set(ref.pdfPath, null);
+    busy.set(ref.pdfPath, null); // 清掉安装阶段说明，回到「翻译进行中」
     const stem = basename(ref.pdfPath, extname(ref.pdfPath));
     const dstStem = ref.pdfPath.slice(0, -extname(ref.pdfPath).length);
     // 推理型模型（glm-5.x / k3 / deepseek-flash）默认每段都先跑隐藏推理，翻译慢好几倍还偶尔返回空。
@@ -155,7 +175,7 @@ async function launch(ref, dataDir, endpoint) {
             if (err) {
                 const tail = (stdout + '\n' + stderr).slice(-400);
                 console.error('[dsh-paper-reader] babeldoc failed:', tail);
-                busy.set(ref.pdfPath, `babeldoc 执行失败: ${err.message}`);
+                busy.set(ref.pdfPath, { error: `babeldoc 执行失败: ${err.message}` });
                 return;
             }
             // 产出从 scratch 挪回文献目录：<名>.zh-CN.mono.pdf → -zh.pdf，.dual.pdf → -dual.pdf
@@ -167,19 +187,18 @@ async function launch(ref, dataDir, endpoint) {
                     await rename(f, dstStem + '-dual.pdf');
                 }
                 if (!existsSync(zhPdfPath(ref)) && !existsSync(dualPdfPath(ref))) {
-                    busy.set(ref.pdfPath, 'babeldoc 未产出译文 PDF');
+                    busy.set(ref.pdfPath, { error: 'babeldoc 未产出译文 PDF' });
                     return;
                 }
                 busy.delete(ref.pdfPath);
                 console.log(`[dsh-paper-reader] babeldoc done: ${ref.name}`);
             }
             catch (e) {
-                busy.set(ref.pdfPath, `移动译文产物失败: ${e instanceof Error ? e.message : String(e)}`);
+                busy.set(ref.pdfPath, { error: `移动译文产物失败: ${e instanceof Error ? e.message : String(e)}` });
             }
         })();
     });
     child.unref?.();
-    return { started: true };
 }
 /** 解析 PDF 变体路径（原文/zh/dual）。 */
 export function pdfVariantPath(ref, variant) {

package/lib/client.js

@@ -54,6 +54,29 @@ window.__ModuleLoader__.load({
       window.dispatchEvent(new CustomEvent('dpr:paper', { detail: { topic, name, n } }))
     }
 
+    /* ---------- 会话消息里的阅读器跳转链接 ----------
+       agent 引用页码时给 [文字](http://<host>/paper-reader/?topic=..&name=..&page=N)
+       （dsh 渲染器对 http(s) 链接加 target=_blank，capture 拦截是应用内跳转的唯一窗口）。
+       同论文 → 阅读器 iframe 原地跳页：不重载文档，用户正在看中文版就落在中文版该页；
+       跨论文/iframe 未就绪 → setCurrentPaper 换论文，页码经 pendingJumpPage 挂上 src，
+       阅读器 bootstrap 直达目标页。 */
+    let readerIframeEl = null
+    let pendingJump = null // { page, q }：跨论文/iframe 未就绪时挂上 src，由阅读器 bootstrap 消费
+    function jumpToPaperPage(topic, name, page, q) {
+      // 阅读器页签若关着尽量展开（seat 未绑定时静默失败，跳转本身仍生效）
+      try { clientApi?.sidebarRight?.openTab('dpr-reader', { revealIfOpened: true }) } catch {}
+      const cur = JSON.parse(localStorage.getItem('dpr.last') || 'null')
+      const tp = topic || cur?.topic, nm = name || cur?.name
+      if (!tp || !nm) return
+      const win = readerIframeEl?.contentWindow
+      if (cur && tp === cur.topic && nm === cur.name && win?.dpr?.jumpPage) {
+        win.dpr.jumpPage(page, q || undefined)
+        return
+      }
+      pendingJump = { page, q }
+      setCurrentPaper(tp, nm, cur && tp === cur.topic && nm === cur.name ? cur.n : 0)
+    }
+
     async function api(path, opts) {
       const r = await fetch('/paper-reader/api' + path, opts)
       if (!r.ok) {
@@ -116,7 +139,12 @@ window.__ModuleLoader__.load({
       const [geo, setGeo] = useState({ visible: false, left: 0, top: 0, width: 0, height: 0 })
       const lastShownRef = React.useRef(0)
       useEffect(() => {
-        const on = (e) => setCur({ topic: e.detail.topic, name: e.detail.name })
+        const on = (e) => {
+          // 跳转链接挂的页码/引文（jumpToPaperPage 的跨论文路径）：随 src 传给阅读器 bootstrap
+          const j = pendingJump
+          pendingJump = null
+          setCur({ topic: e.detail.topic, name: e.detail.name, page: j?.page, q: j?.q })
+        }
         window.addEventListener('dpr:paper', on)
         return () => window.removeEventListener('dpr:paper', on)
       }, [])
@@ -185,9 +213,10 @@ window.__ModuleLoader__.load({
         return () => { alive = false; clearInterval(timer) }
       }, [])
       if (!cur) return null
-      const src = `/paper-reader/?topic=${encodeURIComponent(cur.topic)}&name=${encodeURIComponent(cur.name)}`
+      const src = `/paper-reader/?topic=${encodeURIComponent(cur.topic)}&name=${encodeURIComponent(cur.name)}${cur.page ? '&page=' + cur.page : ''}${cur.q ? '&q=' + encodeURIComponent(cur.q) : ''}`
       return createPortal(
         h('iframe', {
+          ref: (el) => { readerIframeEl = el },
           src, title: t('pdfTab'),
           'data-dsh-plugin': 'dsh-paper-reader', 'data-dsh-surface': 'reader-overlay',
           style: {
@@ -752,6 +781,23 @@ window.__ModuleLoader__.load({
 
           clientApi = { slots: ctx.slots, layout: ctx.layout, uiWorkspace: ctx.uiWorkspace, sidebarRight: ctx.sidebarRight }
 
+          // 阅读器跳转链接拦截（capture 阶段，一次绑定）：渲染器给 http(s) 链接加了
+          // target=_blank，不拦会新开浏览器标签页 —— 拦下后交给应用内跳转
+          if (!window.__dprJumpLinkBound) {
+            window.__dprJumpLinkBound = true
+            document.addEventListener('click', (e) => {
+              const a = e.target?.closest?.('a[href*="/paper-reader/"]')
+              if (!a) return
+              let url
+              try { url = new URL(a.href) } catch { return }
+              const page = +(url.searchParams.get('page') || 0)
+              if (!page) return
+              e.preventDefault()
+              e.stopImmediatePropagation()
+              jumpToPaperPage(url.searchParams.get('topic'), url.searchParams.get('name'), page, url.searchParams.get('q') || '')
+            }, true)
+          }
+
           // 常驻浮层（与模式开关无关）：阅读器 iframe + 对话头部历史/新建按钮组。
           // shell.overlay 是官方钦定的整页浮层自留地（list slot，additive，不遮蔽任何人）。
           // （对话头部按钮不走 slot —— session 作用域 slot 不渲染动态包条目，只能自渲染。）

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@ggboy123/dsh-paper-reader",
-  "version": "0.2.1",
+  "version": "0.6.1",
   "description": "DeepSeek Harness 插件：论文伴读工作台 — PDF 转录 / 检索 / 原生对话伴读 + 内置阅读器页面。",
   "license": "MIT",
   "type": "module",
@@ -55,7 +55,8 @@
     "reader"
   ],
   "dependencies": {
-    "@deepseek-ai/dsh-tools": "0.1.5-rc.2"
+    "@deepseek-ai/dsh-tools": "0.1.5-rc.2",
+    "pdfjs-dist": "^6.3.289"
   },
   "peerDependencies": {
     "@deepseek-ai/cordis": "^4.0.2"

package/presets/paper-reader/agent.cordis.yml

@@ -12,7 +12,9 @@
     complete: true
     includeRuntimeContext: false
     prefix: |
-      You are a paper-reading tutor (论文伴读老师) powered by the {{model}} model. You don't just answer questions about a paper — you TEACH it, the way a good advisor walks a student through a paper: with a plan, step-by-step guidance, and exams.
+      You are a paper-reading examiner and coach (论文伴读教练) powered by the {{model}} model. Your job is NOT to deliver the paper's content — the paper itself does that, and the student learns by reading it with their own eyes. Your job is to design good questions, send the student off to read independently, and then rigorously verify what they actually understood. You are more examiner than lecturer.
+
+      The cardinal rule: NEVER summarize, walk through, or pre-explain content the student hasn't read and answered questions on yet. Feeding conclusions before the student reads robs them of the learning — that is the failure mode of this role. Your default loop is: assign reading + questions → wait for the student's own answer → probe and grade → and ONLY if a real gap remains after they've genuinely tried, explain that specific missing point with a page reference.
 
       Tools at your disposal:
       - list_papers: list topics and papers in the library.
@@ -21,18 +23,25 @@
       - study_progress: read the student's learning record for a paper (plan + quiz history), stored in <论文>.study.json next to the PDF.
       - study_update: update that record — set_plan (3~6 steps), complete_step (with your feedback note), record_quiz (score/total + weak-point summary).
 
+      Tool results carry a `阅读器链接基址` (readerUrl) line. EVERY page reference in your replies MUST be a clickable markdown link built from it: [链接文字](<readerUrl 原样照搬>&page=N) — e.g. [第 4 页](…&page=4). Clicking it jumps the student's built-in reader straight to that page, in whichever language version (原文/中英对照/纯中文) they are currently viewing — so never add any language/variant parameter yourself.
+
+      When pointing at a specific passage (a re-read assignment, the evidence for an answer), ALSO append a quote parameter so the reader highlights the exact passage: &q=<a verbatim 3~8 word phrase lifted from that passage, URL-encoded, e.g. &q=inexpensive%20commodity%20hardware>. Rules for q: copy it character-for-character from the search_paper/transcript text (never paraphrase), prefer a phrase at the START of the passage, no line breaks inside it, always percent-encode spaces (%20) and other non-ASCII characters. If you're unsure of the exact wording, omit q rather than guess — a wrong q silently degrades to a plain page jump. Bare "（第 N 页）" text citations are a formatting failure; always link. If the latest tool result has no readerUrl (null), fall back to plain text page numbers.
+
       Your session is bound to one paper: every paper tool auto-locates it when you omit topic/name, so NEVER ask the user which paper they mean in a paper session — just call the tool. Only pass topic/name explicitly when the user asks about a DIFFERENT paper.
 
       Two modes:
 
-      1) Quick Q&A. When the user asks a factual question (often a passage selected in the reader, with its page number), just answer it: search_paper first, answer with explicit page citations like （第 4 页）, match the user's language. No lesson, no plan, no quiz.
+      1) Quick Q&A. When the user asks a factual question (often a passage selected in the reader, with its page number), just answer it: search_paper first, answer with explicit page links like [第 4 页](…&page=4), match the user's language. No lesson, no plan, no quiz.
 
-      2) Guided learning. When the user wants to LEARN the paper ("带我读", "我想学这篇", "继续", "检验我", or is working through a plan), act as the tutor:
+      2) Guided learning. When the user wants to LEARN the paper ("带我读", "我想学这篇", "继续", "检验我", or is working through a plan), act as the examiner-coach:
       - Always start by calling study_progress to see where the student is.
-      - No plan yet: skim the paper (transcribe/search the abstract, introduction, section headings), then design a 3~6 step reading plan. Each step names WHAT to read (with pages) and WHAT the student should be able to explain afterwards. Save it with set_plan, present it, and immediately start guiding step 1.
-      - Guiding a step: point to the exact pages/sections, explain the background the student needs, and end with one focused question the student answers in their own words. When they respond, give specific feedback (what was right, what was missing, with page references) and IN THE SAME TURN call complete_step with a short note — feedback without the complete_step call is a bookkeeping failure; the next step only begins after the record shows it. If you skipped earlier bookkeeping, read study_progress and bring it up to date before continuing.
-      - After the plan is done, or when the student asks to be tested: run a quiz — 3~5 questions grounded in the paper (mix of recall and "why is it designed this way"), one at a time or as a set; grade each answer specifically, and IN THE SAME TURN as the final grading call record_quiz with the score and a weak-point summary, then finish with targeted review advice (which pages to revisit).
-      - Be encouraging but honest. Praise what was genuinely understood; never wave through a wrong answer. Chinese students get Chinese, English students get English.
+      - No plan yet: skim the paper yourself (transcribe/search the abstract, introduction, section headings), then design a 3~6 step plan with set_plan. The FIRST step is always an independent full read (独立通读) verified by 2~4 simple check questions; the remaining steps are theme-by-theme 精读, each verified by questions. After saving the plan, present ONLY the step-1 assignment: which pages to read and the check questions. Do NOT preview the paper — no abstract recap, no "this paper is about...". The questions themselves are the orientation.
+      - Step 1 (独立通读): pose 2~4 simple questions (e.g. 这篇论文要解决什么问题？方法用一句话怎么概括？最关键的实验结果/结论是什么？), point at the pages that matter if the paper is long, then STOP and wait. The student reads on their own and comes back with answers in their own words. Grade each answer specifically — what's right, what's vague, what's wrong, with page links. If an answer is hollow, ask ONE sharper follow-up on the same point before deciding. complete_step (with a short feedback note) only when the answers demonstrate real grasp.
+      - 精读 steps (question-driven): for each step, FIRST give the question(s) and the exact pages to read (as links), then let the student go read and answer. Questions must force engagement: "为什么这里用 X 而不是 Y？", "第 N 页的图/公式说明了什么？", "去掉这个设计会怎样？" — never "请总结第 N 节".
+      - 严格师傅原则（答案不如意时的固定动作）: when an answer is vague, wrong, or smells pasted, your FIRST move is NOT to explain and NOT merely to follow up verbally — send the student back to the source: name the exact section and give a re-read link with q so the passage highlights on click ([重读 3.2 节（第 5 页）](…&page=5&q=…)), tell them what specific thing to look for, and have them re-answer the same question afterwards. Escalate up the hint ladder only across re-attempts: re-read assignment → a smaller sub-question → pointing at the exact paragraph. Explain directly ONLY after a genuine failed re-attempt — and then explain just that point, not the whole section.
+      - complete_step is earned by demonstrated understanding (a correct answer in the student's own words), never by "我看完了". Feedback without the complete_step call in the same turn is a bookkeeping failure; if you skipped earlier bookkeeping, read study_progress and bring it up to date first.
+      - After the plan is done, or when the student asks to be tested: run a quiz — 3~5 questions grounded in the paper (mix recall, "why is it designed this way", and "apply it to a new case"); grade each answer specifically, and IN THE SAME TURN as the final grading call record_quiz with the score and a weak-point summary, then finish with targeted review advice — each weak point gets a re-read link to the exact page.
+      - Be warm but demanding. Praise genuine understanding; never wave through a wrong answer. If an answer smells pasted from the abstract or another AI, ask them to close the paper and re-answer from memory in their own words. Chinese students get Chinese, English students get English.
 
       The conversation itself is durable (the harness persists it) — never offer to save, archive, or export the conversation, and never mention file paths or tool names in your reply.
 

package/presets/paper-reader/preset.yml

@@ -1,3 +1,3 @@
 name: 论文伴读
-description: 像老师一样带读论文的 Agent：制定分步阅读计划、逐节指导、出题检验并记录成绩（每篇论文一份 .study.json 学习档案）。
+description: 考官式论文伴读：你先独立阅读，AI 只负责出题、校验和追问薄弱点，不灌输内容，只在你真正卡住时才讲解（每篇论文一份 .study.json 学习档案）。
 order: 10

package/reader/index.html

@@ -4,7 +4,12 @@
 <meta charset="utf-8">
 <meta name="viewport" content="width=device-width, initial-scale=1">
 <title>论文伴读 · dsh-paper-reader</title>
-<script src="https://cdn.jsdelivr.net/npm/pdfjs-dist@3.11.174/build/pdf.min.js"></script>
+<!-- pdf.js 6.x 本地 vendor（不再走 CDN）：3.11 的字体解析对 babeldoc 嵌入的 CID 字体会回退成黑体 -->
+<script type="module">
+  import * as pdfjsLib from '/paper-reader/vendor/pdf.min.mjs';
+  window.pdfjsLib = pdfjsLib;
+  pdfjsLib.GlobalWorkerOptions.workerSrc = '/paper-reader/vendor/pdf.worker.min.mjs';
+</script>
 <style>
   :root {
     --bg: #f6f7f9; --panel: #ffffff; --border: #e2e5ea; --text: #24292f;
@@ -37,9 +42,33 @@
   #pdf-pages { display: flex; flex-direction: column; align-items: safe center; gap: 14px; }
   .page { position: relative; background: #fff; box-shadow: 0 1px 4px rgba(0,0,0,.18); }
   .page.flash { outline: 3px solid var(--hl); outline-offset: 2px; transition: outline .15s; }
-  .textLayer { position: absolute; inset: 0; overflow: hidden; line-height: 1; }
-  .textLayer span { position: absolute; white-space: pre; color: transparent; cursor: text; transform-origin: 0 0; }
+  /* textLayer CSS 必须是 pdf.js 6.x 官方规则的子集：TextLayer 只往 span 写 --font-height/--scale-x/--rotate
+     变量，font-size 和 transform 全靠这里的 calc 表达式撑开；缺了的话 span 会按 body 字号渲染，
+     盒子与 canvas 字形错位 —— 表现为选中高亮缺字符。--scale-factor 由 JS 在渲染时写入 */
+  .textLayer {
+    position: absolute; inset: 0; overflow: clip; line-height: 1; text-align: initial;
+    letter-spacing: normal; word-spacing: normal; text-size-adjust: none;
+    forced-color-adjust: none; transform-origin: 0 0; caret-color: CanvasText; z-index: 0;
+    --total-scale-factor: calc(var(--scale-factor, 1) * var(--user-unit, 1));
+    --text-scale-factor: calc(var(--total-scale-factor) * var(--min-font-size, 1));
+    --min-font-size-inv: calc(1 / var(--min-font-size, 1));
+    --scale-round-x: 1px; --scale-round-y: 1px;
+  }
+  .textLayer :is(span, br) { color: transparent; position: absolute; white-space: pre; cursor: text; transform-origin: 0 0; user-select: text; }
+  .textLayer > :not(.markedContent),
+  .textLayer .markedContent span:not(.markedContent) {
+    z-index: 1;
+    --font-height: 0;
+    font-size: calc(var(--text-scale-factor) * var(--font-height));
+    --scale-x: 1; --rotate: 0deg;
+    transform: rotate(var(--rotate)) scaleX(var(--scale-x)) scale(var(--min-font-size-inv));
+  }
+  .textLayer .markedContent { display: contents; }
   .textLayer ::selection { background: rgba(22,93,255,.3); }
+  .textLayer br::selection { background: transparent; }
+  /* 跳转链接的片段高亮（&q= 引文定位）：常驻不淡出，用户在阅读器里任意点一下即消
+     （同代码编辑器的选中高亮语义：消失由用户触发，不由计时器触发） */
+  .textLayer span.dpr-hl { background: rgba(255,196,0,.45); border-radius: 2px; }
   #empty { margin: 60px auto; color: var(--text-2); text-align: center; max-width: 420px; }
 
   /* ---------- 选中即问浮层 ---------- */
@@ -144,6 +173,9 @@ function esc(s) { return s.replace(/[&<>"]/g, (c) => ({ '&': '&amp;', '<': '&lt;
 let current = null;        // { topic, name }
 let zhState = { zh: false, dual: false, busy: false };
 let zhVariant = null;      // null=原文 'dual'=中英对照 'zh'=纯中文
+// 每篇论文记住上次用的语言版本：伴读会话的跳转链接原地跳页不重载，
+// 加上这里的恢复，「中文界面跳中文」在同篇跳转/重开/跨论文三条路径上都成立
+const variantKey = (topic, name) => `dpr.variant.${topic}/${name}`;
 
 /* ================= 打开文献 ================= */
 function pdfUrl(variant) {
@@ -155,8 +187,6 @@ async function openPaper(topic, name, opts = {}) {
   if (!name) return;
   current = { topic, name };
   localStorage.setItem('dpr.last', JSON.stringify(current));
-  zhVariant = null;
-  $('zh-toggle').classList.remove('active');
   $('status').textContent = '加载 PDF…';
   const q = `topic=${encodeURIComponent(topic)}&name=${encodeURIComponent(name)}`;
   const meta = await api('/paper?' + q);
@@ -164,8 +194,13 @@ async function openPaper(topic, name, opts = {}) {
     ? `${name} · ${meta.pageCount ?? '?'} 页`
     : `${name} · 未转录（提问时会自动转录）`;
   await refreshZhState();
-  await loadPdf(pdfUrl(null));
-  if (opts.page) jumpToPage(opts.page);
+  // 恢复该论文上次的语言版本（译文已不存在时回落原文）
+  const saved = localStorage.getItem(variantKey(topic, name));
+  zhVariant = (saved === 'zh' && zhState.zh) || (saved === 'dual' && zhState.dual) ? saved : null;
+  $('zh-toggle').classList.toggle('active', zhVariant !== null);
+  if (zhVariant) { fitMode = true; fitBtn.classList.add('active'); } // 译文版页宽不同，强制适宽
+  await loadPdf(pdfUrl(zhVariant));
+  if (opts.page) jumpTo(opts.page, opts.q); // 不 await：跳转内部自带「等页面渲出来」轮询
 }
 
 /* ================= 中英切换 ================= */
@@ -175,9 +210,11 @@ async function refreshZhState() {
   btn.style.display = '';
   btn.disabled = zhState.busy;
   btn.textContent = zhState.busy ? '译' : '中';
-  btn.title = zhState.busy ? `翻译进行中${zhState.model ? `（${zhState.model}）` : ''}…`
+  btn.title = zhState.busy ? (zhState.phase || `翻译进行中${zhState.model ? `（${zhState.model}）` : ''}…`)
     : (zhState.zh || zhState.dual) ? '切换中文译文'
     : '还没有中文版，点击生成（babeldoc，需要几分钟）';
+  // 安装/准备阶段在状态栏持续可见（轮询每 5s 刷新一次）
+  if (zhState.busy && zhState.phase) $('status').textContent = zhState.phase;
   // ↻ 只在「已有译文且空闲」时出现；状态变化时顺手拆掉两击确认的待确认态
   const regen = $('zh-regen');
   const canRegen = (zhState.zh || zhState.dual) && !zhState.busy;
@@ -190,6 +227,7 @@ $('zh-toggle').onclick = async () => {
   if (zhState.dual || zhState.zh) {
     // 两态切换：原文 ↔ 纯中文；只存在对照版时退到对照
     zhVariant = zhVariant === null ? (zhState.zh ? 'zh' : 'dual') : null;
+    localStorage.setItem(variantKey(current.topic, current.name), zhVariant ?? '');
     $('zh-toggle').classList.toggle('active', zhVariant !== null);
     // 切换变体后强制适宽：译文版页宽与原文不同（对照双栏≈2×），不重算会溢出裁切
     fitMode = true; fitBtn.classList.add('active');
@@ -202,12 +240,14 @@ $('zh-toggle').onclick = async () => {
 // 生成/重翻共用：force=true 时服务端先删旧译文再启动；成功后轮询到完成
 async function startGenerate(force) {
   try {
-    await api('/zh/generate', {
+    const r = await api('/zh/generate', {
       method: 'POST', headers: { 'content-type': 'application/json' },
       body: JSON.stringify(force ? { ...current, force: true } : current),
     });
     const m = zhState.model ? `（${zhState.model}）` : '';
-    toast(force ? `已删除旧译文，开始重新翻译${m}…` : `开始生成中文版${m}（约几分钟）…`, 5000);
+    toast(r.firstRun
+      ? '首次使用：正在自动安装翻译引擎（uv + Python + babeldoc，约几分钟），装完自动开始翻译…'
+      : (force ? `已删除旧译文，开始重新翻译${m}…` : `开始生成中文版${m}（约几分钟）…`), 6000);
   } catch (e) {
     // 没配端点 → 直接弹配置表单；填好保存后自动回来重试（记住是不是重翻）
     if (e.code === 'no-endpoint') { cfgRetryForce = force; openCfgPop(true); return; }
@@ -349,7 +389,7 @@ $('cfg-save').onclick = async () => {
 };
 
 /* ================= PDF 渲染(pdf.js:缩放 + devicePixelRatio 高清,移植自 pdfqa) ================= */
-pdfjsLib.GlobalWorkerOptions.workerSrc = 'https://cdn.jsdelivr.net/npm/pdfjs-dist@3.11.174/build/pdf.worker.min.js';
+/* pdfjsLib 与 workerSrc 由头部 module 脚本设置（本地 vendor，pdf.js 6.x） */
 
 let pdfDoc = null;
 let zoomScale = 1.5;
@@ -367,7 +407,7 @@ let renderToken = 0;
 })();
 
 async function loadPdf(url) {
-  pdfDoc = await pdfjsLib.getDocument(url).promise;
+  pdfDoc = await pdfjsLib.getDocument({ url }).promise; // v6 起不接受字符串参数，必须传对象
   $('empty').style.display = 'none';
   if (fitMode) { fitToWidth(); return; } // fitToWidth 内部 setZoom → renderAllPages
   await renderAllPages();
@@ -417,7 +457,8 @@ async function renderAllPages() {
     if (token !== renderToken) return;
     const tc = await page.getTextContent();
     if (token !== renderToken) return;
-    await pdfjsLib.renderTextLayer({ textContent: tc, container: tl, viewport, textDivs: [] }).promise;
+    // pdf.js 6.x: renderTextLayer() 移除，改 TextLayer 类
+    await new pdfjsLib.TextLayer({ textContentSource: tc, container: tl, viewport }).render();
   }
   const newRange = pane.scrollHeight - pane.clientHeight;
   pane.scrollTop = anchorRatio * Math.max(0, newRange);
@@ -448,6 +489,55 @@ function jumpToPage(n) {
   setTimeout(() => div.classList.remove('flash'), 1600);
 }
 
+/* 片段定位高亮（跳转链接的 &q= 引文）：在指定页 textLayer 里找 q、高亮并滚到它。
+   归一化时忽略空白/连字符/软连字符、大小写 —— 论文跨行断词（dis-tributed）也能命中。
+   找不到（如用户在中文界面而 q 是英文原文）返回 false，调用方退回仅跳页。 */
+function clearHighlights() {
+  document.querySelectorAll('.textLayer span.dpr-hl').forEach((sp) => sp.classList.remove('dpr-hl'));
+}
+// 高亮的消失交给用户：阅读器里任意点一下即清（mousedown 比 click 早，划选文字的那一下也算）
+document.addEventListener('mousedown', clearHighlights);
+
+function highlightSnippet(page, q) {
+  const spans = [...(document.querySelector(`.page[data-page="${page}"]`)?.querySelectorAll('.textLayer span:not(.markedContent)') ?? [])];
+  if (!spans.length || !q) return false;
+  let hay = '';
+  const startOf = [];
+  for (const sp of spans) { startOf.push(hay.length); hay += sp.textContent; }
+  // 归一化串 + 回原串的索引映射（归一化改变长度，不能直接用原索引）
+  const map = [];
+  let nh = '';
+  for (let i = 0; i < hay.length; i++) {
+    const c = hay[i].toLowerCase();
+    if (/\s/.test(c) || c === '-' || c === '­') continue;
+    nh += c; map.push(i);
+  }
+  const needle = q.toLowerCase().replace(/[\s\-­]+/g, '');
+  const at = needle ? nh.indexOf(needle) : -1;
+  if (at < 0) return false;
+  const from = map[at], to = map[at + needle.length - 1] + 1;
+  const hit = spans.filter((sp, i) => startOf[i] < to && startOf[i] + sp.textContent.length > from);
+  if (!hit.length) return false;
+  clearHighlights();
+  for (const sp of hit) sp.classList.add('dpr-hl');
+  hit[0].scrollIntoView({ behavior: 'smooth', block: 'center' });
+  return true;
+}
+
+/* 统一跳转入口：先等目标页渲染出来（fit 模式下 loadPdf 返回时页面可能还在渲），
+   有 q 就尝试片段高亮，否则整页跳转 + 边框闪烁 */
+async function jumpTo(page, q) {
+  const p0 = Math.max(1, (page | 0) || 1);
+  for (let i = 0; i < 100; i++) {
+    if (pdfDoc && document.querySelector(`.page[data-page="${Math.min(p0, pdfDoc.numPages)}"]`)) {
+      const p = Math.min(p0, pdfDoc.numPages);
+      if (!(q && highlightSnippet(p, q))) jumpToPage(p);
+      return;
+    }
+    await new Promise((r) => setTimeout(r, 100));
+  }
+}
+
 /* ================= 选中即问 ================= */
 const askPop = $('ask-pop');
 let askSelection = null;
@@ -500,6 +590,9 @@ $('ask-input').addEventListener('keydown', (e) => { if (e.key === 'Enter' && !e.
 async function bootstrap() {
   const params = new URLSearchParams(location.search);
   const pTopic = params.get('topic'), pName = params.get('name');
+  // 伴读会话里的跳转链接（[重读第 N 页](…?page=N)）经此直达目标页
+  const pPage = +(params.get('page') || 0) || undefined;
+  const pQ = params.get('q') || undefined;
   const lib = await api('/library');
   if (lib.topics.length === 0) {
     $('empty').style.display = 'block';
@@ -513,14 +606,20 @@ async function bootstrap() {
     if (last && lib.papers.some((p) => p.topic === last.topic && p.name === last.name)) target = last;
     else target = lib.papers[0];
   }
-  if (target) await openPaper(target.topic, target.name);
+  if (target) await openPaper(target.topic, target.name, { page: pPage, q: pQ });
 }
 bootstrap().catch((e) => {
   $('empty').style.display = 'block';
   $('empty').textContent = '文献库加载失败：' + e.message;
 });
 
-window.dpr = { openPaper: (topic, name, page) => openPaper(topic, name, { page }), jumpToPage };
+window.dpr = {
+  openPaper: (topic, name, page, q) => openPaper(topic, name, { page, q }),
+  jumpToPage,
+  // 伴读会话的跳转链接用：当前文档里原地跳转，不重载 —— 用户正在看中文版就落在中文版；
+  // 带 q（原文短引文）时定位并闪烁高亮对应片段，找不到则退回整页跳转
+  jumpPage: (n, q) => { jumpTo(n, q); },
+};
 </script>
 </body>
 </html>