weflow-cli 1.9.0 → 1.11.0

package/src/services/configService.ts

@@ -28,6 +28,8 @@ interface CliConfig {
   wechatOcSyncBuf: string
   /** 持久化的 context_token: wxid -> token (加密 JSON 字符串) */
   wechatOcContextTokens: string
+  /** 官方通道的**入站媒体**是否下载到本地（默认 'false'）：'true' 才下载，落到 output/wechat-media/ */
+  wechatMediaDownload: string
   // 白名单
   whitelist: string[]
   // 黑名单 (绝对禁止收发)
@@ -109,7 +111,7 @@ const CONFIG_DIR = join(homedir(), '.weflow-cli')
 const CONFIG_FILE = join(CONFIG_DIR, 'config.json')
 
 export class ConfigService {
-  private config: CliConfig = { dbPath: '', wxid: '', decryptKey: '', decryptKey3x: '', dataVersion: '', dbPath3x: '', ntDbPath: '', ntKey: '', ntSalt: '', contactDbPath: '', contactKey: '', contactSalt: '', wechatOcToken: '', wechatOcAccountId: '', wechatOcBaseUrl: '', wechatOcSyncBuf: '', wechatOcContextTokens: '', whitelist: [], blacklist: [], whitelistEntries: [], blacklistEntries: [], vaultRepo: '', aiEngine: 'deepseek', wereadApiKey: '', deepseekApiKey: '', typesafeApiKey: '', dashscopeApiKey: '', snsDbPath: '', snsKey: '', snsSalt: '', favDbPath: '', favKey: '', favPassphrase: '', emoticonSeed: '', assistantPrivacy: 'strict', localModel: '', aiBaseUrl: '', aiModel: '', assistantWhitelist: '', assistantGroupWhitelist: '', assistantGroupRequireMention: 'true', assistantFastRoute: 'off', assistantPanelUser: '', quickReplyContacts: '', skillDirs: '', skillDisabled: '', dailySources: '', dailySourceCategories: '', dailyExcludeTopics: '', dailyAiEnabled: 'true' }
+  private config: CliConfig = { dbPath: '', wxid: '', decryptKey: '', decryptKey3x: '', dataVersion: '', dbPath3x: '', ntDbPath: '', ntKey: '', ntSalt: '', contactDbPath: '', contactKey: '', contactSalt: '', wechatOcToken: '', wechatOcAccountId: '', wechatOcBaseUrl: '', wechatOcSyncBuf: '', wechatOcContextTokens: '', wechatMediaDownload: 'false', whitelist: [], blacklist: [], whitelistEntries: [], blacklistEntries: [], vaultRepo: '', aiEngine: 'deepseek', wereadApiKey: '', deepseekApiKey: '', typesafeApiKey: '', dashscopeApiKey: '', snsDbPath: '', snsKey: '', snsSalt: '', favDbPath: '', favKey: '', favPassphrase: '', emoticonSeed: '', assistantPrivacy: 'strict', localModel: '', aiBaseUrl: '', aiModel: '', assistantWhitelist: '', assistantGroupWhitelist: '', assistantGroupRequireMention: 'true', assistantFastRoute: 'off', assistantPanelUser: '', quickReplyContacts: '', skillDirs: '', skillDisabled: '', dailySources: '', dailySourceCategories: '', dailyExcludeTopics: '', dailyAiEnabled: 'true' }
 
   /** 本进程修改过、待回写的字段 (多进程并发写保护) */
   private dirty = new Set<keyof CliConfig>()
@@ -149,6 +151,7 @@ export class ConfigService {
           wechatOcBaseUrl: data.wechatOcBaseUrl || '',
           wechatOcSyncBuf: data.wechatOcSyncBuf || '',
           wechatOcContextTokens: data.wechatOcContextTokens || '',
+          wechatMediaDownload: data.wechatMediaDownload === 'true' ? 'true' : 'false',
           whitelist,
           blacklist,
           whitelistEntries,
@@ -366,7 +369,7 @@ export class ConfigService {
   }
 
   clear(): void {
-    this.config = { dbPath: '', wxid: '', decryptKey: '', decryptKey3x: '', dataVersion: '', dbPath3x: '', ntDbPath: '', ntKey: '', ntSalt: '', contactDbPath: '', contactKey: '', contactSalt: '', wechatOcToken: '', wechatOcAccountId: '', wechatOcBaseUrl: '', wechatOcSyncBuf: '', wechatOcContextTokens: '', whitelist: [], blacklist: [], whitelistEntries: [], blacklistEntries: [], vaultRepo: '', aiEngine: 'deepseek', wereadApiKey: '', deepseekApiKey: '', typesafeApiKey: '', dashscopeApiKey: '', snsDbPath: '', snsKey: '', snsSalt: '', favDbPath: '', favKey: '', favPassphrase: '', emoticonSeed: '', assistantPrivacy: 'strict', localModel: '', aiBaseUrl: '', aiModel: '', assistantWhitelist: '', assistantGroupWhitelist: '', assistantGroupRequireMention: 'true', assistantFastRoute: 'off', assistantPanelUser: '', quickReplyContacts: '', skillDirs: '', skillDisabled: '', dailySources: '', dailySourceCategories: '', dailyExcludeTopics: '', dailyAiEnabled: 'true' }
+    this.config = { dbPath: '', wxid: '', decryptKey: '', decryptKey3x: '', dataVersion: '', dbPath3x: '', ntDbPath: '', ntKey: '', ntSalt: '', contactDbPath: '', contactKey: '', contactSalt: '', wechatOcToken: '', wechatOcAccountId: '', wechatOcBaseUrl: '', wechatOcSyncBuf: '', wechatOcContextTokens: '', wechatMediaDownload: 'false', whitelist: [], blacklist: [], whitelistEntries: [], blacklistEntries: [], vaultRepo: '', aiEngine: 'deepseek', wereadApiKey: '', deepseekApiKey: '', typesafeApiKey: '', dashscopeApiKey: '', snsDbPath: '', snsKey: '', snsSalt: '', favDbPath: '', favKey: '', favPassphrase: '', emoticonSeed: '', assistantPrivacy: 'strict', localModel: '', aiBaseUrl: '', aiModel: '', assistantWhitelist: '', assistantGroupWhitelist: '', assistantGroupRequireMention: 'true', assistantFastRoute: 'off', assistantPanelUser: '', quickReplyContacts: '', skillDirs: '', skillDisabled: '', dailySources: '', dailySourceCategories: '', dailyExcludeTopics: '', dailyAiEnabled: 'true' }
     // clear 意图是全量重置: 所有字段标记为脏, 覆盖磁盘上的全部旧值
     this.dirty = new Set(Object.keys(this.config) as (keyof CliConfig)[])
     this.save()

package/src/services/exportService.ts

@@ -1,5 +1,5 @@
 import { join, basename, dirname } from 'path'
-import { existsSync, mkdirSync, writeFileSync } from 'fs'
+import { existsSync, mkdirSync, readFileSync, writeFileSync } from 'fs'
 import { execFile } from 'child_process'
 import { promisify } from 'util'
 import { chatService } from './chatService.js'
@@ -15,10 +15,67 @@ export type ExportContract = 'raw' | 'weflow-v1'
 export interface ExportResult { success: boolean; path?: string; count?: number; error?: string }
 
 const execFileAsync = promisify(execFile)
+
+/**
+ * md 导出落盘的媒体清单：`<output>/<talker>_assets/manifest.json`（`local_id:create_time` → 条目）。
+ *
+ * 这是"其余格式也能带图/带文件"的那一半：**md 导出抽取媒体，json/txt/excel 消费同一份清单** —— 匹配
+ * 逻辑（缓存缩略图、资源 md5、原图那档、`msg/file` 里的文档本体）只存在于 Python 那条路上，在这里重写
+ * 一遍必然与它漂移。读不到就返回空表：那说明还没跑过 md/`--media` 导出，行为与从前一致。
+ */
+export interface MediaRecord {
+  /** 相对输出目录的路径；`kind: 'file'` 时指向 `files/` 下的文档。 */
+  path: string
+  kind: 'image' | 'file'
+  /** 文档的原始文件名（图片没有）。 */
+  name?: string
+}
+
+/**
+ * 从清单里取这条消息的媒体条目。
+ *
+ * 键是 `local_id:create_time`：**local_id 会在分片之间重复**（同一会话的表可能存在于多个分片，
+ * 各片自己编号），只按 local_id 查会让 json/txt/excel 引用到别的消息的图。两条消息硬要同 id 同秒
+ * 才算撞上，那个概率可以忽略；回退查裸 local_id 是为了兼容更早写出的清单。
+ */
+export function mediaFor(manifest: Record<string, MediaRecord>, message: { localId?: unknown; createTime?: unknown }): MediaRecord | undefined {
+  const id = String(message.localId)
+  return manifest[`${id}:${message.createTime}`] ?? manifest[id]
+}
+
+export function readMediaManifest(outputDir: string, talker: string): Record<string, MediaRecord> {
+  try {
+    const file = join(outputDir, `${talker}_assets`, 'manifest.json')
+    if (!existsSync(file)) return {}
+    const data = JSON.parse(readFileSync(file, 'utf8'))
+    const map = data && data.messages
+    if (!map || typeof map !== 'object') return {}
+    // v1 的值是裸字符串（只有图片），v2 起是带 `kind` 的对象。升级不该让磁盘上已有的清单失效 ——
+    // 用户不会因为升了个版本就重跑一次导出。
+    const normalized: Record<string, MediaRecord> = {}
+    for (const [key, value] of Object.entries(map as Record<string, unknown>)) {
+      if (typeof value === 'string') {
+        normalized[key] = { path: value, kind: 'image' }
+      } else if (value && typeof value === 'object' && typeof (value as MediaRecord).path === 'string') {
+        const record = value as MediaRecord
+        normalized[key] = { ...record, kind: record.kind === 'file' ? 'file' : 'image' }
+      }
+    }
+    return normalized
+  } catch {
+    return {}
+  }
+}
 function getPackageRoot(): string {
   return resolvePackageRoot(import.meta.url)
 }
 
+/** txt/excel 共用的文字标签：文档要带上原名，光一个相对路径看不出那是什么东西。 */
+function mediaLabel(record: MediaRecord | undefined): string {
+  if (!record) return ''
+  return record.kind === 'file' ? `${record.name || record.path} (${record.path})` : record.path
+}
+
 export class ExportService {
   async exportJson(talker: string, outputDir: string, limit = 0, from?: number, to?: number, contract: ExportContract = 'raw'): Promise<ExportResult> {
     try {
@@ -29,13 +86,26 @@ export class ExportService {
 
       const dir = this.ensureOutputDir(outputDir)
       const filePath = join(dir, `${talker}_messages.json`)
+      // 有清单就带上媒体相对路径（清单由 markdown 导出产生，见 readMediaManifest）
+      const mediaMap = readMediaManifest(dir, talker)
+      const withMedia = Object.keys(mediaMap).length
+        ? messages.map(m => {
+          const media = mediaFor(mediaMap, m)
+          // `media` 保持"相对路径字符串"这个老形状（已有消费方按字符串读）；`mediaKind`/`mediaName`
+          // 是新增的 —— 不带这两项，读 JSON 的人只能靠扩展名猜这是图片还是文档，而导出明明知道。
+          if (!media) return m
+          return media.kind === 'file'
+            ? { ...m, media: media.path, mediaKind: media.kind, mediaName: media.name }
+            : { ...m, media: media.path, mediaKind: media.kind }
+        })
+        : messages
       const payload = contract === 'weflow-v1'
-        ? createWeFlowEnvelope(messages, new Date().toISOString(), {
+        ? createWeFlowEnvelope(withMedia, new Date().toISOString(), {
           requestedFrom: from,
           requestedTo: to,
           requestedLimit: limit,
         })
-        : messages
+        : withMedia
       writeFileSync(filePath, JSON.stringify(payload, null, 2), 'utf8')
       return { success: true, path: filePath, count: messages.length }
     } catch {
@@ -50,10 +120,15 @@ export class ExportService {
         return { success: false, error: '未找到消息' }
       }
 
+      const mediaMap = readMediaManifest(this.ensureOutputDir(outputDir), talker)
       const lines = messages.map(m => {
         const time = new Date(m.createTime * 1000).toLocaleString('zh-CN')
         const sender = m.isSend ? '我' : (m.senderUsername || talker)
-        const content = m.parsedContent || m.rawContent || ''
+        // 有清单就把 `[图片]` / `[文件] 名字` 换成能直接打开的引用，没有就保持原样
+        const media = mediaFor(mediaMap, m)
+        const content = media
+          ? (media.kind === 'file' ? `[文件: ${mediaLabel(media)}]` : `[图片: ${media.path}]`)
+          : (m.parsedContent || m.rawContent || '')
         return `[${time}] ${sender}: ${content}`
       })
 
@@ -66,7 +141,12 @@ export class ExportService {
     }
   }
 
-  async exportHtml(talker: string, outputDir: string, limit = 0, date = '', from?: number, to?: number, quiet = false, options: { fullImages?: boolean } = {}): Promise<ExportResult> {
+  async exportHtml(talker: string, outputDir: string, limit = 0, date = '', from?: number, to?: number, quiet = false, options: { fullImages?: boolean; format?: 'html' | 'md'; assetsOnly?: boolean } = {}): Promise<ExportResult> {
+    const format = options.format ?? 'html'
+    if (format === 'md' && (from !== undefined || to !== undefined)) {
+      // 不静默降级成「导出全部」：范围是调用方明确要的，做不到就说出来。
+      return { success: false, error: 'markdown 导出暂不支持 --from/--to：请用 --date，或整段导出' }
+    }
     try {
       if (from !== undefined || to !== undefined) {
         return this.exportHtmlBasic(talker, outputDir, limit, from, to)
@@ -83,6 +163,9 @@ export class ExportService {
       const passphrase = configService.get('favPassphrase') || cfg.decryptKey || ''
 
       if (!db || !key || !salt) {
+        if (format === 'md') {
+          return { success: false, error: 'markdown 导出需要已配置的数据库与密钥（这条路没有简版回退）' }
+        }
         // Fallback: basic HTML
         let fallbackFrom: number | undefined = from
         let fallbackTo: number | undefined = to
@@ -114,11 +197,16 @@ export class ExportService {
         }
       } catch {}
 
-      const args: string[] = [
-        script,
+      const args: string[] = [script]
+      if (format === 'md') {
+        // 图片落盘 + 清单。资产目录按 **talker** 命名，好让 json/txt/excel 导出能确定性找到它
+        // （它们的文件前缀是 talker，而 md 文件名用的是显示名，不显式指定就对不上）。
+        args.push('--format', 'md', '--assets-dir', join(outputDir || './output', `${talker}_assets`))
+        if (options.assetsOnly) args.push('--assets-only')
+      } else {
         // Paginate: a whole year of chat in one file makes the browser crawl.
-        '--per-page', '100',
-      ]
+        args.push('--per-page', '100')
+      }
       const ownWxid = String(cfg.wxid || '').trim()
       const exportCacheDir = cacheDir && existsSync(cacheDir) ? cacheDir : undefined
       const exportAccountDir = accountDir && existsSync(accountDir) ? accountDir : undefined
@@ -189,11 +277,36 @@ export class ExportService {
       if (date) {
         return { success: false, error: `指定日期导出失败: ${failure}` }
       }
-      // Fallback to basic HTML
+      if (format === 'md') {
+        // markdown 没有「简版」：回退成 HTML 会让调用方以为拿到了 .md。失败就是失败。
+        return { success: false, error: failure }
+      }
+      // Fallback to basic HTML：**这条回退不带内嵌图片**，调用方得知道 ——
+      // 2026-10-13 实测：Python 因一个 AttributeError 退出后，这里静默回退，
+      // 而 CLI 照样打印「✓ 导出成功」，人只会以为导出正常。
+      if (!quiet) console.error('  ⚠ 已回退到简版 HTML（不含内嵌图片）')
       return this.exportHtmlBasic(talker, outputDir, limit, from, to)
     }
   }
 
+  /**
+   * 只抽媒体（图片与文档落盘 + 清单），不产出正文 —— `export … json|txt|excel --media` 先跑这一步。
+   * 与 markdown 导出共用同一条 Python 通路与同一个资产目录，所以清单只有一份。
+   *
+   * `limit`/`date` 必须跟着调用方走：抽取原本固定按"整个会话"跑，而消费它的 json/txt/excel 是按
+   * `--limit` 切片的 —— `--limit 200` 于是把整段历史的媒体都抽出来（文档还要拷成文件），
+   * 实测 67 份里只有 33 份落在导出的 200 条里。
+   */
+  async extractMedia(talker: string, outputDir: string, quiet = false, options: { fullImages?: boolean } = {}, limit = 0, date = ''): Promise<ExportResult> {
+    return this.exportHtml(talker, outputDir, limit, date, undefined, undefined, quiet,
+      { ...options, format: 'md', assetsOnly: true })
+  }
+
+  /** markdown + 图片落盘到 `<output>/<talker>_assets/`（同一条 Python 通路，换个输出形态）。 */
+  async exportMarkdown(talker: string, outputDir: string, limit = 0, date = '', from?: number, to?: number, quiet = false, options: { fullImages?: boolean } = {}): Promise<ExportResult> {
+    return this.exportHtml(talker, outputDir, limit, date, from, to, quiet, { ...options, format: 'md' })
+  }
+
   private async exportHtmlBasic(talker: string, outputDir: string, limit = 0, from?: number, to?: number): Promise<ExportResult> {
     const messages = await chatService.getMessagesInRange(talker, limit, from, to)
     if (messages.length === 0) {
@@ -228,16 +341,20 @@ export class ExportService {
         { header: '发送者', key: 'sender', width: 15 },
         { header: '类型', key: 'type', width: 10 },
         { header: '内容', key: 'content', width: 80 },
-        { header: '消息ID', key: 'msgId', width: 15 }
+        { header: '消息ID', key: 'msgId', width: 15 },
+        // 有清单时填相对路径（清单由 markdown 导出产生）
+        { header: '媒体', key: 'media', width: 40 }
       ]
 
+      const mediaMap = readMediaManifest(this.ensureOutputDir(outputDir), talker)
       for (const m of messages) {
         sheet.addRow({
           time: new Date(m.createTime * 1000).toLocaleString('zh-CN'),
           sender: m.isSend ? '我' : (m.senderUsername || talker),
           type: this.getMessageTypeName(m.localType),
           content: m.parsedContent || m.rawContent || '',
-          msgId: m.localId
+          msgId: m.localId,
+          media: mediaLabel(mediaFor(mediaMap, m))
         })
       }
 

package/src/services/messageQuery.ts

@@ -5,6 +5,35 @@ export interface MessageTimeRange {
   to?: number
 }
 
+/**
+ * 这三种形状的输入会被 `resolveTalker` **原样返回**（见 `src/utils/talkerUtils.ts`），
+ * 于是拼错的 ID 会一路走到「取不到消息」，和「这条会话在窗口内确实没有消息」长得一模一样。
+ * 只看形状，不碰 IO —— 判定所需的名单由调用方传进来。
+ */
+export function looksLikeTalkerId(input: string): boolean {
+  const value = String(input || '')
+  return value.startsWith('wxid_') || value.includes('@chatroom') || value.includes('@openim')
+}
+
+/**
+ * 空结果时补的一句话：**只在"看起来是 ID、但会话与联系人都没有它"时**才给。
+ * 返回空串表示没什么可补的 —— 名字形状的输入本来就由 `resolveTalker` 报错，认识的 ID
+ * 取不到消息是正常结果。
+ *
+ * 这里**不改退出码、也不改 `success`**：既有契约里「没有消息」不是错误（脚本与 MCP 依赖它，
+ * 见 `docs/HEALTH-CHECK.md` 的那一行），所以只把歧义说出来，不动语义。
+ */
+export function emptyMessagesNote(
+  input: string,
+  known: { sessionUsernames: Iterable<string>; contactUsernames: Iterable<string> },
+): string {
+  if (!looksLikeTalkerId(input)) return ''
+  for (const id of known.sessionUsernames) if (id === input) return ''
+  for (const id of known.contactUsernames) if (id === input) return ''
+  return `这个 ID 既不在会话里、也不在联系人里：${input}` +
+    ' —— 先确认有没有拼错（weflow-cli sessions / weflow-cli contacts）'
+}
+
 export async function collectMessagesInRange(
   fetchPage: (limit: number, offset: number) => Promise<Message[]>,
   limit: number,