weaver-work-cli 0.1.28 → 0.1.31

package/dist/shortcuts/meeting/operations/attachment-content.js

@@ -0,0 +1,418 @@
+/**
+ * 会议 / 议题附件的「读取」链路（列表 → 下载 → 抽文本）。
+ *
+ * 背景（2026-10-28 用户提出「小E 没有获取附件和文档内容的能力，你有办法吗」）：
+ * 此前 skill 里写着「附件回读只有文件名、拿不到 fileId ⇒ 无法下载」。**这个结论是错的**：
+ * 真机直连 `query_mtDetail_info` 看到的是 `mt_accessorys = "1331337310852210690"` —— 就是 **fileId**。
+ * 拿到 fileId 之后，用 E10 登录态 GET
+ * `/api/file/remotedownload/{fileId}/meeting/true` 就能取回原始字节（实测返回文件内容，不是 401/跳登录）。
+ *
+ * 于是三个 operation：
+ *   - `meeting.attachment.list`     会议（含项目文件）/ 议题资料的附件清单（fileId + 归属字段）
+ *   - `meeting.attachment.download` 按 fileId 或按会议批量落地到本地
+ *   - `meeting.attachment.extract`  下载并抽文本（txt/md/csv/json/docx/xlsx），给 AI 读内容用
+ *
+ * 边界：只做「取内容 + 抽纯文本」。PDF/图片的解析不在会议线内（PDF 走 workbuddy-pdf 技能，图片走视觉模型），
+ * 这里只负责把文件取下来，不假装能读懂所有格式。
+ */
+import { mkdir, writeFile } from 'node:fs/promises';
+import { join, resolve } from 'node:path';
+import { meetingValidationError } from '../errors.js';
+import { meetingStateDir } from './env.js';
+import { meetingSuccess, schemaOf, SCHEMA_ID } from './types.js';
+import { readDocxEntries } from '../invitation/docx.js';
+/** 会议主表里存放 fileId 列表的字段（值形如 "id1,id2" 或 JSON 数组）。 */
+const MEETING_FILE_FIELDS = ['mt_accessorys', 'project_files'];
+const FIELD_LABELS = {
+    mt_accessorys: '会议附件',
+    project_files: '项目文件',
+    related_materials: '议题资料',
+};
+const TEXT_EXTENSIONS = new Set(['.txt', '.md', '.csv', '.json', '.xml', '.html', '.htm', '.log', '.yaml', '.yml', '.tsv']);
+const OFFICE_EXTENSIONS = new Set(['.docx', '.xlsx', '.pptx']);
+/** 雪花 id 形态（15 位以上纯数字）。服务端偶尔会把附件字段返回成**文件名**，必须能识别。 */
+export function looksLikeFileId(value) {
+    return /^\d{15,}$/.test(String(value ?? '').trim());
+}
+/** 把字段值解析成 id 列表（兼容 CSV、JSON 数组字符串、对象数组）。 */
+export function parseFileIdList(value) {
+    if (value === undefined || value === null)
+        return [];
+    if (Array.isArray(value)) {
+        return value
+            .map((item) => (item && typeof item === 'object' ? String(item.id ?? '') : String(item ?? '')))
+            .map((item) => item.trim())
+            .filter((item) => item !== '');
+    }
+    const text = String(value).trim();
+    if (text === '')
+        return [];
+    if (text.startsWith('[')) {
+        try {
+            return parseFileIdList(JSON.parse(text));
+        }
+        catch {
+            // 不是合法 JSON，按 CSV 继续
+        }
+    }
+    return text
+        .split(/[,;，；\s]+/)
+        .map((item) => item.trim())
+        .filter((item) => item !== '' && item !== '[]');
+}
+/** 直接读 EB 详情原始响应里的主表（不经过 CLI 的摘要层，避免又出现"被裁剪"的坑）。 */
+async function readRawMainTable(env, interfacePk, id, operation) {
+    const session = await env.host.getSession();
+    const response = await env.host.request({
+        method: 'POST',
+        path: `/api/ebuilder/form/formdata/v2/getFormDataByPk/${interfacePk}`,
+        operation,
+        body: { datajson: { mainTable: { id }, header: { operator: session.userId } } },
+        acceptBusiness: (value) => {
+            if (!value || typeof value !== 'object')
+                return false;
+            if (typeof value.status === 'boolean')
+                return value.status;
+            return Number(value.code ?? 200) === 200;
+        },
+    });
+    const dj = response?.datajson ?? response?.data ?? {};
+    const rows = dj?.datas ?? dj?.data ?? (Array.isArray(dj) ? dj : []);
+    const first = Array.isArray(rows) ? rows[0] : rows;
+    return first?.mainTable ?? first ?? {};
+}
+/** 会议附件清单（会议主表的 mt_accessorys / project_files）。 */
+export async function listMeetingAttachments(env, mtId) {
+    const main = await readRawMainTable(env, 'query_mtDetail_info', mtId, 'meeting.attachment.list');
+    const refs = [];
+    for (const field of MEETING_FILE_FIELDS) {
+        for (const value of parseFileIdList(main[field])) {
+            refs.push({
+                fileId: value,
+                idLike: looksLikeFileId(value),
+                source: field,
+                sourceLabel: FIELD_LABELS[field] ?? field,
+                ownerId: mtId,
+            });
+        }
+    }
+    return {
+        refs,
+        raw: Object.fromEntries(MEETING_FILE_FIELDS.map((field) => [field, main[field] ?? ''])),
+    };
+}
+/** 议题资料清单（议题主表的 related_materials）。 */
+export async function listTopicAttachments(env, topicId) {
+    const main = await readRawMainTable(env, 'query_mt_topic_info_new', topicId, 'meeting.attachment.list');
+    return parseFileIdList(main.related_materials).map((value) => ({
+        fileId: value,
+        idLike: looksLikeFileId(value),
+        source: 'related_materials',
+        sourceLabel: FIELD_LABELS.related_materials,
+        ownerId: topicId,
+    }));
+}
+function extFromContentType(contentType) {
+    const text = String(contentType ?? '').toLowerCase();
+    if (text.includes('wordprocessingml'))
+        return '.docx';
+    if (text.includes('spreadsheetml'))
+        return '.xlsx';
+    if (text.includes('presentationml'))
+        return '.pptx';
+    if (text.includes('pdf'))
+        return '.pdf';
+    if (text.includes('json'))
+        return '.json';
+    if (text.includes('html'))
+        return '.html';
+    if (text.includes('plain'))
+        return '.txt';
+    return '';
+}
+function nameFromDisposition(disposition) {
+    const text = String(disposition ?? '');
+    const encoded = /filename\*=UTF-8''([^;]+)/i.exec(text);
+    if (encoded) {
+        try {
+            return decodeURIComponent(encoded[1]).trim();
+        }
+        catch {
+            // 落到下一个分支
+        }
+    }
+    const plain = /filename="?([^";]+)"?/i.exec(text);
+    return plain ? plain[1].trim() : '';
+}
+/**
+ * 按 fileId 下载到本地。
+ *
+ * URL 用 `remotedownload/{fileId}/meeting/true`：真机实测 200 且返回原始字节；
+ * 不带 `meeting/true` 的 `/api/file/remotedownload/{fileId}` 是 404（别再试）。
+ */
+export async function downloadMeetingFile(env, fileId, options = {}) {
+    if (!env.host.download) {
+        throw meetingValidationError('download_unsupported', '当前 host 不支持二进制下载（测试替身？），无法获取附件内容');
+    }
+    const response = await env.host.download({
+        path: `/api/file/remotedownload/${encodeURIComponent(fileId)}/meeting/true`,
+    });
+    const dispositionName = nameFromDisposition(response.contentDisposition ?? '');
+    const ext = extFromContentType(response.contentType);
+    const baseName = (options.fileName ?? dispositionName) || `${fileId}${ext}`;
+    const safeName = baseName.replace(/[\\/:*?"<>|]/g, '_');
+    const dir = options.outDir
+        ? resolve(options.outDir)
+        : join(meetingStateDir(env.ctx), 'attachments');
+    await mkdir(dir, { recursive: true });
+    const target = join(dir, safeName);
+    await writeFile(target, response.buffer);
+    return { path: target, bytes: response.buffer.length, contentType: response.contentType };
+}
+function decodeText(buffer) {
+    const utf8 = buffer.toString('utf8');
+    // 中文老文件常见 GBK：UTF-8 解出大量替换字符时退回 gb18030
+    if (utf8.includes('\uFFFD')) {
+        try {
+            return new TextDecoder('gb18030').decode(buffer);
+        }
+        catch {
+            return utf8;
+        }
+    }
+    return utf8;
+}
+function stripXml(xml) {
+    return xml
+        .replace(/<w:p[ >]/g, '\n<w:p ')
+        .replace(/<\/w:p>/g, '\n')
+        .replace(/<[^>]+>/g, '')
+        .replace(/&amp;/g, '&')
+        .replace(/&lt;/g, '<')
+        .replace(/&gt;/g, '>')
+        .replace(/&quot;/g, '"')
+        .replace(/&#39;/g, "'")
+        .replace(/\n{2,}/g, '\n')
+        .trim();
+}
+/** xlsx → TSV：读 sharedStrings + 第一张表，够 AI 读表结构。 */
+function xlsxToText(buffer) {
+    const entries = readDocxEntries(buffer);
+    const find = (suffix) => {
+        const entry = entries.find((item) => item.name === suffix);
+        return entry ? Buffer.from(entry.data).toString('utf8') : '';
+    };
+    const sharedXml = find('xl/sharedStrings.xml');
+    const shared = [...sharedXml.matchAll(/<si>([\s\S]*?)<\/si>/g)].map((match) => stripXml(match[1]).replace(/\n/g, ' '));
+    const sheetPaths = entries.map((item) => item.name).filter((path) => /^xl\/worksheets\/sheet\d+\.xml$/.test(path)).sort();
+    const first = sheetPaths[0];
+    if (!first)
+        return '';
+    const sheetXml = find(first);
+    const lines = [];
+    for (const rowMatch of sheetXml.matchAll(/<row[^>]*>([\s\S]*?)<\/row>/g)) {
+        const cells = [];
+        for (const cellMatch of rowMatch[1].matchAll(/<c\b([^>]*)>([\s\S]*?)<\/c>/g)) {
+            // ⚠️ 类型属性必须从**开标签**里单独取：把 t="s" 塞进一个可选分组 + 非贪婪匹配时，
+            // 惰性量词会先把 `[^>]*?` 收成空串、可选分组整体跳过，结果所有单元格都当数字读（单测抓到过）。
+            const attrs = cellMatch[1] ?? '';
+            const type = /\bt="([^"]+)"/.exec(attrs)?.[1] ?? '';
+            const inner = cellMatch[2] ?? '';
+            const value = /<v>([\s\S]*?)<\/v>/.exec(inner)?.[1] ?? '';
+            const inline = /<is>([\s\S]*?)<\/is>/.exec(inner)?.[1] ?? '';
+            if (type === 's')
+                cells.push(shared[Number(value)] ?? '');
+            else if (type === 'inlineStr')
+                cells.push(stripXml(inline));
+            else
+                cells.push(stripXml(value));
+        }
+        if (cells.some((cell) => cell !== ''))
+            lines.push(cells.join('\t'));
+    }
+    return lines.join('\n');
+}
+function pptxToText(buffer) {
+    const entries = readDocxEntries(buffer);
+    return entries
+        .filter((item) => /^ppt\/slides\/slide\d+\.xml$/.test(item.name))
+        .sort((left, right) => left.name.localeCompare(right.name, undefined, { numeric: true }))
+        .map((item, index) => `# 第 ${index + 1} 页\n${stripXml(Buffer.from(item.data).toString('utf8'))}`)
+        .join('\n\n');
+}
+/** 从下载到的文件里抽纯文本（不支持的格式明确返回 unsupported，不编内容）。 */
+export function extractText(buffer, fileName, contentType = '') {
+    const lower = String(fileName ?? '').toLowerCase();
+    const ext = /\.[a-z0-9]+$/.exec(lower)?.[0] ?? '';
+    if (ext === '.docx' || contentType.includes('wordprocessingml')) {
+        const entries = readDocxEntries(buffer);
+        const document = entries.find((item) => item.name === 'word/document.xml');
+        return { text: document ? stripXml(Buffer.from(document.data).toString('utf8')) : '', kind: 'docx' };
+    }
+    if (ext === '.xlsx' || contentType.includes('spreadsheetml')) {
+        return { text: xlsxToText(buffer), kind: 'xlsx' };
+    }
+    if (ext === '.pptx' || contentType.includes('presentationml')) {
+        return { text: pptxToText(buffer), kind: 'pptx' };
+    }
+    if (TEXT_EXTENSIONS.has(ext) || contentType.startsWith('text/') || contentType.includes('json') || contentType.includes('xml')) {
+        return { text: decodeText(buffer), kind: 'text' };
+    }
+    return { text: '', kind: 'unsupported' };
+}
+/** 附件清单（按会议 / 按议题）。 */
+const attachmentList = {
+    name: 'meeting.attachment.list',
+    resource: 'attachment',
+    method: 'list',
+    action: 'meeting.attachment.list',
+    summary: '查询会议/议题的附件清单（返回 fileId）：会议附件 mt_accessorys、项目文件 project_files、议题资料 related_materials；直接读 EB 详情原始值，不经摘要层',
+    capability: 'attachment',
+    risk: 'read',
+    inputSchema: schemaOf({
+        mtId: { ...SCHEMA_ID, description: '会议 id（查会议附件时必填，与 topicId 二选一）' },
+        topicId: { ...SCHEMA_ID, description: '议题 id（查议题资料时用）' },
+    }, { required: [] }),
+    handler: async (args, env) => {
+        const mtId = String(args.mtId ?? '').trim();
+        const topicId = String(args.topicId ?? '').trim();
+        if (!mtId && !topicId) {
+            throw meetingValidationError('id_required', '必须提供 mtId（会议）或 topicId（议题）');
+        }
+        const refs = [];
+        let raw = {};
+        if (mtId) {
+            const result = await listMeetingAttachments(env, mtId);
+            refs.push(...result.refs);
+            raw = result.raw;
+        }
+        if (topicId)
+            refs.push(...(await listTopicAttachments(env, topicId)));
+        return meetingSuccess('meeting.attachment.list', {
+            mtId: mtId || null,
+            topicId: topicId || null,
+            count: refs.length,
+            items: refs,
+            raw,
+        }, refs.length === 0 ? { warnings: ['该会议/议题当前没有附件（或字段值为空）'] } : {});
+    },
+};
+/** 下载附件到本地。 */
+const attachmentDownload = {
+    name: 'meeting.attachment.download',
+    resource: 'attachment',
+    method: 'download',
+    action: 'meeting.attachment.download',
+    summary: '按 fileId 下载附件到本地（默认 ~/.weaver-work-cli/meeting/attachments/）；也可传 mtId 一次下载该会议全部附件',
+    capability: 'attachment',
+    risk: 'read',
+    inputSchema: schemaOf({
+        fileId: { ...SCHEMA_ID, description: '文件 id（与 mtId 二选一）' },
+        mtId: { ...SCHEMA_ID, description: '会议 id：一次下载该会议全部附件与项目文件' },
+        outDir: { type: 'string', description: '落地目录（默认 <state>/meeting/attachments）' },
+        fileName: { type: 'string', description: '指定文件名（仅单文件时生效）' },
+    }, { required: [] }),
+    handler: async (args, env) => {
+        const fileId = String(args.fileId ?? '').trim();
+        const mtId = String(args.mtId ?? '').trim();
+        if (!fileId && !mtId) {
+            throw meetingValidationError('id_required', '必须提供 fileId 或 mtId');
+        }
+        const outDir = typeof args.outDir === 'string' && args.outDir.trim() !== '' ? String(args.outDir) : undefined;
+        if (fileId) {
+            const one = await downloadMeetingFile(env, fileId, {
+                ...(outDir ? { outDir } : {}),
+                ...(typeof args.fileName === 'string' ? { fileName: String(args.fileName) } : {}),
+            });
+            return meetingSuccess('meeting.attachment.download', { downloaded: [{ fileId, ...one }], count: 1 });
+        }
+        const refs = (await listMeetingAttachments(env, mtId)).refs;
+        if (refs.length === 0) {
+            throw meetingValidationError('attachment_empty', `会议 ${mtId} 没有附件可下载`);
+        }
+        const downloaded = [];
+        const warnings = [];
+        for (const ref of refs) {
+            try {
+                const one = await downloadMeetingFile(env, ref.fileId, { ...(outDir ? { outDir } : {}) });
+                downloaded.push({ fileId: ref.fileId, source: ref.source, ...one });
+            }
+            catch (error) {
+                warnings.push(`附件 ${ref.fileId}（${ref.sourceLabel}）下载失败：${String(error?.message ?? error).slice(0, 120)}`);
+            }
+        }
+        return meetingSuccess('meeting.attachment.download', { mtId, downloaded, count: downloaded.length }, warnings.length > 0 ? { warnings } : {});
+    },
+};
+/** 下载 + 抽文本（给 AI 读内容）。 */
+const attachmentExtract = {
+    name: 'meeting.attachment.extract',
+    resource: 'attachment',
+    method: 'extract',
+    action: 'meeting.attachment.extract',
+    summary: '下载附件并抽取纯文本（txt/md/csv/json/docx/xlsx/pptx），返回正文与 txt 落地路径；pdf/图片等不支持格式会明确标注，不编造内容',
+    capability: 'attachment',
+    risk: 'read',
+    inputSchema: schemaOf({
+        fileId: { ...SCHEMA_ID, description: '文件 id（与 mtId 二选一）' },
+        mtId: { ...SCHEMA_ID, description: '会议 id：抽取该会议全部附件（最多 10 个）' },
+        maxChars: { type: 'number', description: '每个文件返回正文的最大字符数（默认 4000，最多 20000）' },
+        outDir: { type: 'string', description: '落地目录（默认 <state>/meeting/attachments）' },
+    }, { required: [] }),
+    handler: async (args, env) => {
+        const fileId = String(args.fileId ?? '').trim();
+        const mtId = String(args.mtId ?? '').trim();
+        if (!fileId && !mtId) {
+            throw meetingValidationError('id_required', '必须提供 fileId 或 mtId');
+        }
+        const maxChars = Math.max(200, Math.min(20_000, Number(args.maxChars ?? 4000) || 4000));
+        const outDir = typeof args.outDir === 'string' && args.outDir.trim() !== '' ? String(args.outDir) : undefined;
+        const refs = fileId
+            ? [{ fileId, idLike: looksLikeFileId(fileId), source: 'explicit', sourceLabel: '指定文件' }]
+            : (await listMeetingAttachments(env, mtId)).refs.slice(0, 10);
+        if (refs.length === 0) {
+            throw meetingValidationError('attachment_empty', `会议 ${mtId} 没有附件可抽取`);
+        }
+        const files = [];
+        const warnings = [];
+        for (const ref of refs) {
+            try {
+                const downloaded = await downloadMeetingFile(env, ref.fileId, { ...(outDir ? { outDir } : {}) });
+                const { readFile } = await import('node:fs/promises');
+                const buffer = await readFile(downloaded.path);
+                const extracted = extractText(buffer, downloaded.path, downloaded.contentType);
+                const text = extracted.text.slice(0, maxChars);
+                let textPath;
+                const absolutePath = downloaded.path;
+                const isPlainText = /\.(txt|md|csv|json|xml|html?|log|ya?ml|tsv)$/i.test(absolutePath);
+                if (extracted.text && !isPlainText) {
+                    // 原文本身就是纯文本时不另存（否则同名覆盖自己，纯属噪音）
+                    textPath = `${absolutePath.replace(/\.[a-z0-9]+$/i, '')}.txt`;
+                    await writeFile(textPath, extracted.text, 'utf8');
+                }
+                files.push({
+                    fileId: ref.fileId,
+                    source: ref.sourceLabel,
+                    path: downloaded.path,
+                    bytes: downloaded.bytes,
+                    kind: extracted.kind,
+                    chars: extracted.text.length,
+                    ...(textPath ? { textPath } : {}),
+                    text,
+                });
+                if (extracted.kind === 'unsupported') {
+                    warnings.push(`${downloaded.path} 是 ${downloaded.contentType || '未知类型'}，会议线不解析该格式（已下载，PDF 可用 pdf 技能、图片用视觉模型）`);
+                }
+            }
+            catch (error) {
+                warnings.push(`附件 ${ref.fileId} 处理失败：${String(error?.message ?? error).slice(0, 120)}`);
+            }
+        }
+        return meetingSuccess('meeting.attachment.extract', { mtId: mtId || null, count: files.length, files }, warnings.length > 0 ? { warnings } : {});
+    },
+};
+export const meetingAttachmentContentOperations = [
+    attachmentList,
+    attachmentDownload,
+    attachmentExtract,
+];