dsh-plugin-litsearch 0.2.0 → 0.3.5

package/lib/host/router.js

@@ -6,19 +6,26 @@
  * `sec-fetch-site: cross-site`。写端点额外要求自定义头 + JSON content-type，
  * 两者一起让浏览器必须预检，堵掉跨站表单与简单请求。
  */
-import { MAX_LIMIT, ROUTE_PREFIX, WRITE_HEADER } from '../search/config.js';
+import { spawn } from 'node:child_process';
+import { mkdir, stat } from 'node:fs/promises';
+import { dirname, extname, isAbsolute, resolve } from 'node:path';
+import { MAX_LIMIT, ROUTE_PREFIX, SESSION_HEADER, WRITE_HEADER } from '../search/config.js';
 import { SearchError } from '../search/errors.js';
 import { asRecord } from '../search/normalize.js';
 import { isSourceId } from '../search/providers/index.js';
+import { TASK_LANGUAGES } from '../search/types.js';
+import { resolvePdfRoot } from './pdf/store.js';
+import { resolveSessionWorkspace } from './workspace.js';
 const MAX_BODY_BYTES = 1_000_000;
 const SORT_MODES = ['relevance', 'citations', 'year'];
-const LANGUAGES = ['any', 'eng', 'chi'];
+const LANGUAGES = TASK_LANGUAGES;
 const ORIGINS = ['model', 'panel', 'history', 'api'];
 const SCREENINGS = ['unread', 'read', 'included', 'excluded'];
 const PDF_STATUSES = [
     'unknown',
     'available',
     'downloaded',
+    'missing',
     'html-only',
     'paywalled',
     'blocked',
@@ -39,6 +46,13 @@ function isTrusted(req) {
         return false;
     return isLoopbackHost(req.headers.host);
 }
+/** 面板把当前会话 id 放在请求头里；iframe 预览等无法带头部的 GET 用查询参数。 */
+function readSessionId(req, params) {
+    const header = req.headers[SESSION_HEADER];
+    const raw = typeof header === 'string' && header.trim() !== '' ? header : params.get('session') ?? '';
+    const id = raw.trim();
+    return id === '' ? undefined : id;
+}
 /** 写端点要求自定义头 + JSON：跨站请求无法在不触发预检的情况下伪造。 */
 function hasWriteIntent(req) {
     if (req.headers[WRITE_HEADER] !== '1')
@@ -72,48 +86,6 @@ function sendFile(res, file) {
     });
     res.end(body);
 }
-/** 内联 PDF，支持单段 Range（预览器会按需拉取）。 */
-function sendPdf(req, res, buffer, filename) {
-    const total = buffer.length;
-    const disposition = `inline; filename*=UTF-8''${encodeURIComponent(filename)}`;
-    const range = req.headers.range;
-    if (range === undefined) {
-        res.writeHead(200, {
-            'content-type': 'application/pdf',
-            'content-length': total,
-            'content-disposition': disposition,
-            'accept-ranges': 'bytes',
-            'cache-control': 'no-store',
-        });
-        res.end(buffer);
-        return;
-    }
-    const match = /^bytes=(\d*)-(\d*)$/u.exec(range.trim());
-    const start = match?.[1] === '' ? Number.NaN : Number(match?.[1]);
-    const end = match?.[2] === '' ? Number.NaN : Number(match?.[2]);
-    if (match === null || (!Number.isFinite(start) && !Number.isFinite(end))) {
-        res.writeHead(416, { 'content-range': `bytes */${total}`, 'cache-control': 'no-store' });
-        res.end();
-        return;
-    }
-    const from = Number.isFinite(start) ? Math.max(0, start) : Math.max(0, total - (end || 0));
-    const to = Number.isFinite(end) ? Math.min(end, total - 1) : total - 1;
-    if (from > to || from >= total) {
-        res.writeHead(416, { 'content-range': `bytes */${total}`, 'cache-control': 'no-store' });
-        res.end();
-        return;
-    }
-    const chunk = buffer.subarray(from, to + 1);
-    res.writeHead(206, {
-        'content-type': 'application/pdf',
-        'content-length': chunk.length,
-        'content-range': `bytes ${from}-${to}/${total}`,
-        'content-disposition': disposition,
-        'accept-ranges': 'bytes',
-        'cache-control': 'no-store',
-    });
-    res.end(chunk);
-}
 function sendHtml(res, status, html) {
     res.writeHead(status, {
         'content-type': 'text/html; charset=utf-8',
@@ -121,6 +93,52 @@ function sendHtml(res, status, html) {
     });
     res.end(html);
 }
+/**
+ * 在系统文件管理器中定位文件：macOS Finder、Windows 资源管理器、Linux 桌面。
+ * 只接受已存在的 PDF 文件或目录（批量下载的「打开下载目录」用后者），
+ * 避免这个端点被当成任意路径探测。
+ */
+async function revealInFileManager(rawPath, downloadRoot) {
+    const target = resolve(rawPath);
+    let info = await stat(target).catch(() => null);
+    // 下载目录可能一篇都还没下过：打开前补建，避免点「打开下载目录」毫无反应。
+    if (info === null && downloadRoot !== undefined && resolve(downloadRoot) === target) {
+        await mkdir(target, { recursive: true });
+        info = await stat(target).catch(() => null);
+    }
+    if (info === null)
+        throw new SearchError('not_found', '文件不存在或已被移动');
+    if (info.isDirectory()) {
+        const [command, args] = process.platform === 'darwin'
+            ? ['open', [target]]
+            : process.platform === 'win32'
+                ? ['explorer.exe', [target]]
+                : ['xdg-open', [target]];
+        await spawnDetached(command, args);
+        return { path: target };
+    }
+    if (!info.isFile() || extname(target).toLowerCase() !== '.pdf') {
+        throw new SearchError('bad_request', '只能定位 PDF 文件或下载目录');
+    }
+    const [command, args] = process.platform === 'darwin'
+        ? ['open', ['-R', target]]
+        : process.platform === 'win32'
+            ? ['explorer.exe', [`/select,${target}`]]
+            : ['xdg-open', [dirname(target)]];
+    await spawnDetached(command, args);
+    return { path: target };
+}
+/** 启动系统命令后立即脱离，不等待它的生命周期。 */
+async function spawnDetached(command, args) {
+    await new Promise((settle, rejectSpawn) => {
+        const child = spawn(command, args, { detached: true, stdio: 'ignore' });
+        child.once('error', rejectSpawn);
+        child.once('spawn', () => {
+            child.unref();
+            settle();
+        });
+    });
+}
 function sendHtmlError(res, error) {
     const { status, error: apiError } = toApiError(error);
     sendHtml(res, status === 500 ? 502 : status, `<!doctype html><html lang="zh-CN"><head><meta charset="utf-8"><title>全文不可用</title></head><body style="font:14px/1.7 -apple-system,'PingFang SC',sans-serif;padding:16px"><p>全文不可用：${escapeHtml(apiError.message)}</p></body></html>`);
@@ -291,26 +309,6 @@ function readAnnotations(value) {
             if (groupId !== undefined)
                 input.groupId = groupId;
         }
-        if (record['shortlistOrder'] === null)
-            input.shortlistOrder = null;
-        else if (record['shortlistOrder'] !== undefined) {
-            const order = readInt(record['shortlistOrder'], 'shortlistOrder', 0, 100_000);
-            if (order !== undefined)
-                input.shortlistOrder = order;
-        }
-        if (record['note'] !== undefined) {
-            if (typeof record['note'] !== 'string') {
-                throw new SearchError('bad_request', 'note 必须是字符串');
-            }
-            input.note = record['note'].slice(0, 4_000);
-        }
-        if (record['tags'] !== undefined) {
-            const tags = readStringArray(record['tags'], 'tags', 20, 40);
-            if (tags !== undefined && tags !== null)
-                input.tags = tags;
-            else if (tags === null)
-                input.tags = [];
-        }
         return input;
     });
 }
@@ -561,7 +559,7 @@ function parseSegments(pathname) {
         }
     });
 }
-async function handleTasks(service, req, res, seg, params) {
+async function handleTasks(service, req, res, seg, params, sessionId) {
     const id = seg[1];
     if (id === undefined) {
         if (req.method === 'GET') {
@@ -570,7 +568,8 @@ async function handleTasks(service, req, res, seg, params) {
         }
         if (req.method === 'POST') {
             const input = readCreateInput(await readJsonBody(req));
-            const task = await service.create({ ...input, origin: 'panel' });
+            const workspace = await resolveSessionWorkspace(sessionId);
+            const task = await service.create({ ...input, origin: 'panel', workspace });
             ok(res, await service.getDetail(task.id), 201);
             return;
         }
@@ -580,7 +579,7 @@ async function handleTasks(service, req, res, seg, params) {
     const action = seg[2];
     if (action === undefined) {
         if (req.method === 'GET') {
-            ok(res, await service.getDetail(id));
+            ok(res, await service.getDetail(id, await resolveSessionWorkspace(sessionId)));
             return;
         }
         if (req.method === 'PATCH') {
@@ -629,7 +628,7 @@ async function handleTasks(service, req, res, seg, params) {
             fail(res, 405, 'bad_request', 'activate 仅支持 POST');
             return;
         }
-        const task = await service.activate(id);
+        const task = await service.activate(id, await resolveSessionWorkspace(sessionId));
         ok(res, await service.getDetail(task.id));
         return;
     }
@@ -654,7 +653,7 @@ async function handleTasks(service, req, res, seg, params) {
                 fail(res, 405, 'bad_request', '论文详情仅支持 GET');
                 return;
             }
-            ok(res, await service.paperDetail(id, paperId));
+            ok(res, await service.paperDetail(id, paperId, await resolveSessionWorkspace(sessionId)));
             return;
         }
         if (op === 'entities') {
@@ -665,27 +664,20 @@ async function handleTasks(service, req, res, seg, params) {
             ok(res, await service.entities(id, paperId));
             return;
         }
-        if (op === 'download') {
+        if (op === 'recheck') {
             if (req.method !== 'POST') {
-                fail(res, 405, 'bad_request', 'download 仅支持 POST');
+                fail(res, 405, 'bad_request', 'recheck 仅支持 POST');
                 return;
             }
-            ok(res, await service.downloadPdf(id, paperId));
+            ok(res, await service.recheckPdf(id, paperId));
             return;
         }
-        if (op === 'pdf') {
-            if (req.method !== 'GET') {
-                fail(res, 405, 'bad_request', 'pdf 仅支持 GET');
+        if (op === 'download') {
+            if (req.method !== 'POST') {
+                fail(res, 405, 'bad_request', 'download 仅支持 POST');
                 return;
             }
-            try {
-                const file = await service.previewPdf(id, paperId);
-                sendPdf(req, res, file.buffer, file.filename);
-            }
-            catch (error) {
-                const { status, error: apiError } = toApiError(error);
-                fail(res, status === 500 ? 502 : status, apiError.code, apiError.message);
-            }
+            ok(res, await service.downloadPdf(id, paperId, await resolveSessionWorkspace(sessionId)));
             return;
         }
         if (op === 'fulltext') {
@@ -711,7 +703,7 @@ async function handleTasks(service, req, res, seg, params) {
         }
         const body = await readJsonBody(req);
         const ids = readStringArray(body['paperIds'], 'paperIds', 50, 80);
-        ok(res, await service.zoteroSync(id, ids === undefined || ids === null ? undefined : ids));
+        ok(res, await service.zoteroSync(id, ids === undefined || ids === null ? undefined : ids, await resolveSessionWorkspace(sessionId)));
         return;
     }
     if (action === 'compare') {
@@ -737,7 +729,8 @@ async function handleTasks(service, req, res, seg, params) {
             throw new SearchError('conflict', '任务正在运行');
         // 面板里主动重跑 = 用户正在这个任务里工作，把它设为「当前任务」。
         service.setCurrentTask(id);
-        void service.run(id, mode, 'panel').catch((error) => {
+        const workspace = await resolveSessionWorkspace(sessionId);
+        void service.run(id, mode, 'panel', workspace).catch((error) => {
             console.warn('[dsh-plugin-litsearch] 任务执行失败：', error);
         });
         ok(res, await service.getDetail(id), 202);
@@ -791,6 +784,26 @@ export function createHandler(service) {
                 ok(res, await service.eventsSince(Number.isFinite(sinceRev) ? sinceRev : 0));
                 return;
             }
+            if (head === 'files') {
+                if (seg[1] !== 'reveal') {
+                    fail(res, 404, 'not_found', `未知文件操作 ${seg[1] ?? ''}`);
+                    return;
+                }
+                if (req.method !== 'POST') {
+                    fail(res, 405, 'bad_request', 'reveal 仅支持 POST');
+                    return;
+                }
+                const body = await readJsonBody(req);
+                const path = readString(body['path'], 'path', 4096);
+                if (path === undefined || path === '') {
+                    throw new SearchError('bad_request', '缺少 path 参数');
+                }
+                if (!isAbsolute(path))
+                    throw new SearchError('bad_request', 'path 必须是绝对路径');
+                const workspace = await resolveSessionWorkspace(readSessionId(req, url.searchParams));
+                ok(res, await revealInFileManager(path, resolvePdfRoot(workspace)));
+                return;
+            }
             if (head === 'settings') {
                 if (seg[1] === 'probe' && req.method === 'POST') {
                     const source = readString((await readJsonBody(req))['source'], 'source', 40);
@@ -810,14 +823,15 @@ export function createHandler(service) {
                     return;
                 }
                 if (seg[1] === 'storage' && req.method === 'POST') {
-                    ok(res, await service.storageProbe());
+                    const workspace = await resolveSessionWorkspace(readSessionId(req, url.searchParams));
+                    ok(res, await service.storageProbe(workspace));
                     return;
                 }
                 fail(res, 404, 'not_found', `未知设置端点 ${seg[1] ?? ''}`);
                 return;
             }
             if (head === 'tasks') {
-                await handleTasks(service, req, res, seg, url.searchParams);
+                await handleTasks(service, req, res, seg, url.searchParams, readSessionId(req, url.searchParams));
                 return;
             }
             if (head === 'export') {

package/lib/host/tasks/filter.d.ts

@@ -37,6 +37,8 @@ export declare function filterPapers(task: SearchTask, filter?: PaperFilter): Pa
 export declare function screeningOf(task: SearchTask, paper: Paper): Screening;
 /** 有效全文状态：没有人工/下载记录时，上游给的 PDF 直链视为可用。 */
 export declare function pdfStatusOf(task: SearchTask, paper: Paper): PdfStatus;
-export declare function matchesLanguage(paper: Paper, language: TaskLanguage): boolean;
+export declare function matchesLanguage(paper: {
+    language?: string;
+}, language: TaskLanguage): boolean;
 /** 稳定的筛选指纹（统计缓存 key 的一部分）。 */
 export declare function filterKey(filter: PaperFilter): string;

package/lib/host/tasks/filter.js

@@ -72,11 +72,52 @@ export function pdfStatusOf(task, paper) {
 export function matchesLanguage(paper, language) {
     if (language === 'any')
         return true;
-    const value = (paper.language ?? '').toLowerCase();
-    if (language === 'chi')
-        return value.startsWith('chi') || value.startsWith('zh');
-    return value.startsWith('eng') || value.startsWith('en');
+    const raw = (paper.language ?? '').trim().toLowerCase();
+    // 兼容 `en-US` / `zh-CN` 这类带地区的写法，只取主语言码。
+    const code = LANGUAGE_ALIASES[raw] ?? LANGUAGE_ALIASES[raw.split(/[-_]/u)[0] ?? ''];
+    return code !== undefined && code === language;
 }
+/**
+ * 上游语言标记 → 统一的 PubMed 三字母码。
+ *
+ * PubMed / Europe PMC 给三字母码，OpenAlex 等给两字母码，这里都收敛到同一套，
+ * 避免语言筛选把同一种语言判成两种。
+ */
+const LANGUAGE_ALIASES = {
+    eng: 'eng',
+    en: 'eng',
+    english: 'eng',
+    chi: 'chi',
+    zho: 'chi',
+    zh: 'chi',
+    chinese: 'chi',
+    jpn: 'jpn',
+    ja: 'jpn',
+    japanese: 'jpn',
+    kor: 'kor',
+    ko: 'kor',
+    korean: 'kor',
+    fre: 'fre',
+    fra: 'fre',
+    fr: 'fre',
+    french: 'fre',
+    ger: 'ger',
+    deu: 'ger',
+    de: 'ger',
+    german: 'ger',
+    spa: 'spa',
+    es: 'spa',
+    spanish: 'spa',
+    rus: 'rus',
+    ru: 'rus',
+    russian: 'rus',
+    ita: 'ita',
+    it: 'ita',
+    italian: 'ita',
+    por: 'por',
+    pt: 'por',
+    portuguese: 'por',
+};
 /** 稳定的筛选指纹（统计缓存 key 的一部分）。 */
 export function filterKey(filter) {
     return [

package/lib/host/tasks/runner.js

@@ -8,6 +8,7 @@
 import { DEFAULT_TASK_MAX_ROUNDS, MAX_LIMIT, getSettings } from '../../search/config.js';
 import { searchSources } from '../../search/aggregate.js';
 import { mergeHits } from '../../search/identity.js';
+import { matchesLanguage } from './filter.js';
 import { finishRound, startRound } from './rounds.js';
 import { countScreening } from './states.js';
 class RunCanceled extends Error {
@@ -198,11 +199,3 @@ function matchesConditions(paper, conditions) {
     }
     return true;
 }
-function matchesLanguage(paper, language) {
-    if (language === 'any')
-        return true;
-    const value = (paper.language ?? '').toLowerCase();
-    if (language === 'chi')
-        return value.startsWith('chi') || value.startsWith('zh');
-    return value.startsWith('eng') || value.startsWith('en');
-}

package/lib/host/tasks/service.d.ts

@@ -26,6 +26,8 @@ export interface CreateTaskInput {
     sort?: SortMode;
     origin: TaskOrigin;
     fromTaskId?: string;
+    /** 发起方所在会话的工作目录；下载默认落在这里。 */
+    workspace?: string;
     runNow?: boolean;
 }
 export interface TaskPatch {
@@ -76,6 +78,8 @@ export declare class TaskService {
      * 又新建一个同名任务。
      */
     private restore;
+    /** 按设置的总量上限修剪 PDF 去重缓存；失败只告警，不阻断启动。 */
+    private trimPdfCache;
     /** 没有可用快照时，把最近更新的任务设为当前任务并落盘。 */
     private bootstrapCurrentTask;
     get globalRev(): number;
@@ -85,7 +89,7 @@ export declare class TaskService {
     eventsSince(sinceRev: number): Promise<EventsPage>;
     list(filter?: TaskListFilter): Promise<TaskSummary[]>;
     get(id: string): Promise<SearchTask>;
-    getDetail(id: string): Promise<TaskDetail>;
+    getDetail(id: string, workspace?: string): Promise<TaskDetail>;
     /** 任务内轮次历史（时间升序、深拷贝）。 */
     getRounds(id: string): Promise<SearchRound[]>;
     /**
@@ -101,7 +105,9 @@ export declare class TaskService {
     focusTask(id: string, title: string, sessionId?: string): void;
     private persistCurrent;
     /** 用户从任务列表点开某任务 = 进入该任务，之后的模型检索默认落在这里。 */
-    activate(id: string): Promise<SearchTask>;
+    activate(id: string, workspace?: string): Promise<SearchTask>;
+    /** 记下任务最近一次所在的工作目录；只改变下载落点，不改变 rev。 */
+    private applyWorkspace;
     /** 等待该任务当前这一轮结束；返回是否确实已空闲。 */
     waitUntilIdle(id: string, timeoutMs: number, signal?: AbortSignal): Promise<boolean>;
     create(input: CreateTaskInput): Promise<SearchTask>;
@@ -112,7 +118,7 @@ export declare class TaskService {
      *
      * 任务已缓存时同步登记运行态，调用方紧接着调 `waitUntilIdle` 也不会误判为空闲。
      */
-    run(id: string, mode?: 'new' | 'more', origin?: TaskOrigin): Promise<SearchTask>;
+    run(id: string, mode?: 'new' | 'more', origin?: TaskOrigin, workspace?: string): Promise<SearchTask>;
     /**
      * 在同一任务内开新一轮：全量替换检索条件后执行，**不新建任务、不新开 tab**。
      *
@@ -123,7 +129,7 @@ export declare class TaskService {
     cancel(id: string): Promise<SearchTask>;
     results(id: string, filter?: ResultsFilter): Promise<ResultsPage>;
     stats(id: string, filter?: PaperFilter, scope?: 'task' | 'filtered'): Promise<TaskStatsReport>;
-    /** 写入筛读 / 分组 / 必读 / 备注；返回新的 rev 与命中条数。 */
+    /** 写入筛读 / 分组；返回新的 rev 与命中条数。 */
     annotate(id: string, inputs: readonly AnnotationInput[], expectedRev?: number): Promise<AnnotationResult>;
     compare(id: string, withTaskId: string): Promise<CompareResult>;
     /** 生成导出内容；`paperIds` 为空表示按当前筛选导出全部。 */
@@ -136,25 +142,27 @@ export declare class TaskService {
     }>;
     /** 设置页清理：任务文件与 PDF 缓存。 */
     cleanup(scope: CleanupScope): Promise<CleanupResult>;
-    /** 下载单篇 PDF；失败时写入失败状态，若有 EPMC 全文则降级为 html-only。 */
-    downloadPdf(id: string, paperId: string): Promise<PdfDownloadResult>;
-    /** PDF 预览：优先用已落地文件；没有则临时抓取一次但不落盘。 */
-    previewPdf(id: string, paperId: string): Promise<{
-        buffer: Buffer;
-        filename: string;
+    /**
+     * 「查看已下载」前的存在性检查：页面加载后文件可能被删除或移走。
+     * 缺失时把状态改为 `missing` 并广播，所有面板随即显示「下载文件已删除」。
+     */
+    recheckPdf(id: string, paperId: string): Promise<{
+        exists: boolean;
     }>;
+    /** 下载单篇 PDF；失败时写入失败状态，若有 EPMC 全文则降级为 html-only。 */
+    downloadPdf(id: string, paperId: string, workspace?: string): Promise<PdfDownloadResult>;
     /** Europe PMC 全文（JATS → 只含安全标签的 HTML）。 */
     fullText(id: string, paperId: string): Promise<string>;
     /** 单篇论文详情：独立论文页使用，含任务标题与分组。 */
-    paperDetail(id: string, paperId: string): Promise<PaperDetail>;
+    paperDetail(id: string, paperId: string, workspace?: string): Promise<PaperDetail>;
     /** PubTator3 实体标注；无 PMID 时明确告知不可标注。 */
     entities(id: string, paperId: string, signal?: AbortSignal): Promise<PaperEntities>;
     /** 测试 Zotero connector 是否可达，并返回当前选中的分类。 */
     zoteroProbe(): Promise<ZoteroProbeResult>;
     /** 把选中论文同步到 Zotero（含本地 PDF 附件）。 */
-    zoteroSync(id: string, paperIds?: readonly string[]): Promise<ZoteroSyncResult>;
+    zoteroSync(id: string, paperIds?: readonly string[], workspace?: string): Promise<ZoteroSyncResult>;
     /** 下载目录可写性测试（设置页）。 */
-    storageProbe(): Promise<StorageProbeResult>;
+    storageProbe(workspace?: string): Promise<StorageProbeResult>;
     private ensureLocalPdf;
     private setPdfState;
     /** 单源诊断：用固定短查询测一次连通性与延迟（不受冷却影响）。 */