weflow-cli 1.8.2 → 1.10.0

package/src/services/wechatMedia.ts

@@ -0,0 +1,166 @@
+/**
+ * 官方通道（ilink）**入站媒体**的落地：把 CDN 上的密文拉下来、解密、写到本地磁盘。
+ *
+ * 为什么值得做：入站图片/语音/文件/视频此前只解析出"有这一项"，`filePath` 是空串，于是助手拿到消息
+ * 也不知道图里是什么（那四项的落地当时是显式延后的）。下载与解密的本事本来就有
+ * （`WechatClient.downloadMedia`：`<cdn>/download?encrypted_query_param=…` + AES-128-ECB + 去 PKCS#7），
+ * 缺的只是把它接上入站这条路。
+ *
+ * 字段名与优先级照**厂商自己的实现**（`third-party/WeKnora/internal/im/wechat/longpoll.go`）：
+ * 图片优先用 `image_item.aeskey`（十六进制字符串，**不走 base64**），退回 `media.aes_key`（base64）；
+ * 语音/文件/视频只有 `media.aes_key`。两种编码都由 `WechatClient.parseMediaAesKey` 认。
+ *
+ * 三条边界写在这里，不指望调用方记得：
+ * 1. **它不是默认行为** —— 打开 `wechatMediaDownload` 才发生网络访问；
+ * 2. **文件名永远由本地决定** —— 服务端给的名字只取 basename 并过滤字符，`../` 之类必须在这里被吃掉；
+ * 3. **超过上限就不写**（`WECHAT_MEDIA_MAX_BYTES`），失败不留半个文件 —— 半个文件比没有更难查。
+ */
+import { basename, join } from 'path'
+import { mkdir, writeFile } from 'fs/promises'
+
+/** 单件上限。够大（手机拍的视频也能过），又不至于一次 CDN 事故把磁盘写满 */
+export const WECHAT_MEDIA_MAX_BYTES = 25 * 1024 * 1024
+
+/** 落地目录：默认仓库相对 `output/wechat-media/`（`output/` 已在 .gitignore 里），可用环境变量改（测试用临时目录） */
+export function wechatMediaDir(env: NodeJS.ProcessEnv = process.env): string {
+  const fromEnv = env.WEFLOW_WECHAT_MEDIA_DIR
+  return fromEnv && fromEnv.trim() ? fromEnv : join('output', 'wechat-media')
+}
+
+export type InboundMediaType = 'image' | 'record' | 'file' | 'video'
+
+/** item.type 的数字编码（与服务端 item_list 一致） */
+export function mediaTypeOf(itemType: number): InboundMediaType | null {
+  if (itemType === 2) return 'image'
+  if (itemType === 3) return 'record'
+  if (itemType === 4) return 'file'
+  if (itemType === 5) return 'video'
+  return null
+}
+
+/** 服务端给的名字只当作"一个字符串"用：取 basename、去掉控制字符与路径字符、去掉前导点、限长 */
+export function sanitizeFileName(name: string): string {
+  const base = basename(String(name || '').replace(/\\/g, '/')).trim()
+  return base
+    .replace(/[\u0000-\u001f<>:"|?*]/g, '_')
+    .replace(/^\.+/, '')
+    .trim()
+    .slice(0, 80)
+}
+
+/** 按魔数认扩展名。认不出就返回空串（调用方退回按类型给的后缀），不猜 */
+export function extFromMagic(buf: Buffer): string {
+  const b = buf
+  if (b.length >= 3 && b[0] === 0xff && b[1] === 0xd8 && b[2] === 0xff) return '.jpg'
+  if (b.length >= 8 && b[0] === 0x89 && b[1] === 0x50 && b[2] === 0x4e && b[3] === 0x47) return '.png'
+  if (b.length >= 6 && b.subarray(0, 6).toString('latin1').startsWith('GIF8')) return '.gif'
+  if (b.length >= 12 && b.subarray(0, 4).toString('latin1') === 'RIFF'
+    && b.subarray(8, 12).toString('latin1') === 'WEBP') return '.webp'
+  if (b.length >= 2 && b[0] === 0x42 && b[1] === 0x4d) return '.bmp'
+  if (b.length >= 12 && b.subarray(4, 8).toString('latin1') === 'ftyp') return '.mp4'
+  if (b.length >= 4 && b[0] === 0x1a && b[1] === 0x45 && b[2] === 0xdf && b[3] === 0xa3) return '.webm'
+  if (b.length >= 5 && b.subarray(0, 5).toString('latin1') === '#!AMR') return '.amr'
+  if (b.length >= 4 && b.subarray(0, 4).toString('latin1') === 'OggS') return '.ogg'
+  if (b.length >= 2 && b[0] === 0x49 && b[1] === 0x44 && b[2] === 0x33) return '.mp3'
+  if (b.length >= 12 && b.subarray(0, 4).toString('latin1') === 'RIFF'
+    && b.subarray(8, 12).toString('latin1') === 'WAVE') return '.wav'
+  return ''
+}
+
+/** 该媒体类型在服务端字段树里的节点名 */
+function mediaNode(item: any, mediaType: InboundMediaType): any {
+  if (mediaType === 'image') return item?.image_item
+  if (mediaType === 'record') return item?.voice_item
+  if (mediaType === 'file') return item?.file_item
+  return item?.video_item
+}
+
+/** 下载参数（`encrypt_query_param`）。取不到就没法下载——**这时候不猜、不留空文件** */
+export function pickEncryptParam(item: any, mediaType: InboundMediaType): string {
+  const node = mediaNode(item, mediaType)
+  return String(node?.media?.encrypt_query_param || '')
+}
+
+/**
+ * AES key 的取值优先级。**图片这一支必须照厂商实现**：`image_item.aeskey` 是十六进制字符串，
+ * 比 `media.aes_key`（base64）更靠前；漏了这个优先级，图片会解出一堆乱码却不报错。
+ */
+export function pickAesKeyValue(item: any, mediaType: InboundMediaType): string {
+  const node = mediaNode(item, mediaType)
+  if (mediaType === 'image' && node?.aeskey) return String(node.aeskey)
+  return String(node?.media?.aes_key || '')
+}
+
+export interface SaveMediaOptions {
+  item: any
+  mediaType: InboundMediaType
+  /** 同一条消息里的第几项（一条消息可以带多张图） */
+  index: number
+  /** 名字前缀：优先用服务端给的 client_id */
+  messageId: string
+  dir?: string
+  /** 注入的下载器（默认由调用方传 `client.downloadMedia`），测试里换成假的 */
+  download: (encryptedQueryParam: string, aesKeyValue: string) => Promise<Buffer>
+  /** 文件类消息里服务端给的文件名 */
+  serverFileName?: string
+  maxBytes?: number
+}
+
+export interface SaveMediaResult {
+  /** 成功是磁盘路径；失败是空串（**绝不返回"看起来成功"的半成品**） */
+  path: string
+  /** 失败原因，进日志用 */
+  reason: string
+}
+
+/**
+ * 下载一件入站媒体并落盘。失败一律返回 `{ path: '', reason }` 且**不留下文件**：
+ * 助手与导出拿到空 path 时知道"没有这玩意"，比拿到一个 0 字节文件强。
+ */
+export async function saveInboundMedia(opts: SaveMediaOptions): Promise<SaveMediaResult> {
+  const { item, mediaType, index, messageId, download } = opts
+  const cap = opts.maxBytes ?? WECHAT_MEDIA_MAX_BYTES
+
+  const encryptParam = pickEncryptParam(item, mediaType)
+  if (!encryptParam) return { path: '', reason: '服务端没给 encrypt_query_param' }
+
+  const aesKeyValue = pickAesKeyValue(item, mediaType)
+  if (!aesKeyValue) return { path: '', reason: '服务端没给 aes key' }
+
+  let bytes: Buffer
+  try {
+    bytes = await download(encryptParam, aesKeyValue)
+  } catch (error: any) {
+    return { path: '', reason: `下载或解密失败：${error?.message || error}` }
+  }
+  if (!bytes || bytes.length === 0) return { path: '', reason: '下载结果为空' }
+  if (bytes.length > cap) {
+    return { path: '', reason: `超过上限（${bytes.length} > ${cap} 字节）` }
+  }
+
+  const byMagic = extFromMagic(bytes)
+  const fallbackExt = mediaType === 'image' ? '.jpg' : ''
+  const prefix = sanitizeFileName(messageId) || 'msg'
+  let name: string
+  if (mediaType === 'file' && opts.serverFileName) {
+    const safe = sanitizeFileName(opts.serverFileName)
+    if (safe) {
+      const hasExt = /\.[A-Za-z0-9]{1,8}$/.test(safe)
+      name = `${prefix}-${index}-${hasExt || !byMagic ? safe : safe + byMagic}`
+    } else {
+      name = `${prefix}-${index}${byMagic}`
+    }
+  } else {
+    name = `${prefix}-${index}${byMagic || fallbackExt}`
+  }
+
+  const dir = opts.dir || wechatMediaDir()
+  const full = join(dir, name)
+  try {
+    await mkdir(dir, { recursive: true })
+    await writeFile(full, bytes)
+  } catch (error: any) {
+    return { path: '', reason: `写盘失败：${error?.message || error}` }
+  }
+  return { path: full, reason: '' }
+}

package/src/services/wechatMessageService.ts

@@ -8,6 +8,8 @@ import path from 'path'
 import { WechatClient } from '../core/wechatClient.js'
 import { configService } from './configService.js'
 import { resolveInboundRouting } from './assistantRouting.js'
+import { mediaTypeOf, saveInboundMedia, wechatMediaDir } from './wechatMedia.js'
+import type { InboundMediaType } from './wechatMedia.js'
 import type { WechatOCConfig, WechatLoginSession, WechatInboundMessage, WechatMessageComponent } from '../types.js'
 
 function sleep(ms: number): Promise<void> {
@@ -64,10 +66,16 @@ export class WechatMessageService {
   private messageCallbacks: Array<(msg: WechatInboundMessage) => void> = []
   /** 服务端说过 token 失效了。**只由长轮询置位**（见 `classifyPollResult`），恢复时清掉 */
   private tokenExpired = false
+  /** 入站媒体是否落地。undefined = 每次去读配置（默认关，见 `wechatMedia.ts` 的边界 1） */
+  private mediaDownload?: boolean | string
+  /** 落地目录。undefined = 用 `wechatMediaDir()`（可被环境变量改，测试用临时目录） */
+  private mediaDirOverride?: string
 
   constructor(config: WechatOCConfig = {}) {
     this.config = config
     this.syncBuf = config.syncBuf || ''
+    this.mediaDownload = config.mediaDownload
+    this.mediaDirOverride = config.mediaDir
     this.client = new WechatClient({
       baseUrl: config.baseUrl,
       cdnBaseUrl: config.cdnBaseUrl,
@@ -225,7 +233,7 @@ export class WechatMessageService {
         const msgs: any[] = data.msgs || []
         for (const msg of msgs) {
           if (this.shutdownFlag) return
-          const inbound = this.parseInboundMessage(msg)
+          const inbound = await this.parseInboundMessage(msg)
           if (inbound) {
             for (const cb of this.messageCallbacks) {
               try { cb(inbound) } catch (error: any) {
@@ -409,7 +417,40 @@ export class WechatMessageService {
 
   // ====== Private ======
 
-  private parseInboundMessage(msg: any): WechatInboundMessage | null {
+  /** 入站媒体落地的开关：构造时注入的优先，其次读配置。默认关 —— 关着就不发任何网络请求 */
+  private mediaDownloadEnabled(): boolean {
+    const raw = this.mediaDownload ?? configService.get('wechatMediaDownload')
+    return raw === true || raw === 'true'
+  }
+
+  /**
+   * 下载一件入站媒体并落盘，返回路径；**没开这个功能、或下载失败，都返回空串**。
+   * 失败会往日志里写一行原因 —— "没开"和"开了但失败了"必须分得开，否则出事时两种证据全无。
+   */
+  private async inboundMediaPath(
+    item: any,
+    mediaType: InboundMediaType,
+    index: number,
+    messageId: string,
+  ): Promise<string> {
+    if (!this.mediaDownloadEnabled()) return ''
+    const result = await saveInboundMedia({
+      item,
+      mediaType,
+      index,
+      messageId,
+      dir: this.mediaDirOverride || wechatMediaDir(),
+      serverFileName: mediaType === 'file' ? item?.file_item?.file_name : undefined,
+      download: (param, key) => this.client.downloadMedia(param, key),
+    })
+    if (!result.path) {
+      console.warn(`入站媒体未落地（${mediaType}）：${result.reason}`)
+      return ''
+    }
+    return result.path
+  }
+
+  private async parseInboundMessage(msg: any): Promise<WechatInboundMessage | null> {
     const fromUserId: string = msg.from_user_id || ''
     if (!fromUserId) return null
     const routing = resolveInboundRouting(msg, String(configService.get('wechatOcAccountId') || ''))
@@ -423,19 +464,29 @@ export class WechatMessageService {
 
     const components: WechatMessageComponent[] = []
     const itemList: any[] = msg.item_list || []
+    const messageId: string = msg.client_id || uuidHex()
+    let mediaIndex = 0
 
     for (const item of itemList) {
       const itemType = item.type as number
       if (itemType === 1 && item.text_item) {
         components.push({ type: 'plain', text: item.text_item.text || '' })
-      } else if (itemType === 2 && item.image_item) {
-        components.push({ type: 'image', filePath: '' }) // image download deferred
-      } else if (itemType === 3 && item.voice_item) {
-        components.push({ type: 'record', filePath: '' })
-      } else if (itemType === 4 && item.file_item) {
-        components.push({ type: 'file', name: item.file_item.file_name || '', filePath: '' })
-      } else if (itemType === 5 && item.video_item) {
-        components.push({ type: 'video', filePath: '' })
+        continue
+      }
+      const mediaType = mediaTypeOf(itemType)
+      // 缺节点就整项跳过（与既有行为一致：形状不对的项不进 components）
+      if (!mediaType || !item[`${mediaType === 'record' ? 'voice' : mediaType}_item`]) continue
+      mediaIndex += 1
+      // 落盘（默认关，关着时这里不做任何 IO）
+      const filePath = await this.inboundMediaPath(item, mediaType, mediaIndex, messageId)
+      if (mediaType === 'image') {
+        components.push({ type: 'image', filePath })
+      } else if (mediaType === 'record') {
+        components.push({ type: 'record', filePath })
+      } else if (mediaType === 'file') {
+        components.push({ type: 'file', name: item.file_item.file_name || '', filePath })
+      } else {
+        components.push({ type: 'video', filePath })
       }
     }
 
@@ -454,7 +505,7 @@ export class WechatMessageService {
     const messageStr = textComponents.map(c => (c as { type: 'plain'; text: string }).text).join('')
 
     return {
-      messageId: msg.client_id || uuidHex(),
+      messageId,
       fromUserId,
       ...routing,
       senderNickname: msg.from_user_id || '',

package/src/types.ts

@@ -65,6 +65,10 @@ export interface WechatOCConfig {
   apiTimeoutMs?: number
   typingKeepaliveIntervalS?: number
   typingTicketTtlS?: number
+  /** 入站媒体是否落地（缺省=读 `wechatMediaDownload` 配置，默认关）。见 src/services/wechatMedia.ts */
+  mediaDownload?: boolean | string
+  /** 入站媒体落地目录（缺省=output/wechat-media/，可被 WEFLOW_WECHAT_MEDIA_DIR 改） */
+  mediaDir?: string
 }
 
 export interface WechatLoginSession {

package/src/utils/python.ts

@@ -38,7 +38,7 @@ function candidates(): string[] {
 
 function canRun(cmd: string): boolean {
   try {
-    execFileSync(cmd, ['--version'], { stdio: 'ignore', timeout: 5000 })
+    execFileSync(cmd, ['--version'], { windowsHide: true, stdio: 'ignore', timeout: 5000 })
     return true
   } catch {
     return false
@@ -47,7 +47,7 @@ function canRun(cmd: string): boolean {
 
 function hasSqlcipher(cmd: string): boolean {
   try {
-    execFileSync(cmd, ['-c', 'import sqlcipher3'], { stdio: 'ignore', timeout: 10000 })
+    execFileSync(cmd, ['-c', 'import sqlcipher3'], { windowsHide: true, stdio: 'ignore', timeout: 10000 })
     return true
   } catch {
     return false

package/dist/src/utils/wechatEmoji.d.ts

@@ -1,5 +0,0 @@
-/** Replace known WeChat face codes like [害羞] with emoji; unknown codes pass through. */
-export declare function convertFaces(text: string | undefined | null): string;
-/** Convert faces, then truncate without splitting a surrogate pair. */
-export declare function convertFacesTruncated(text: string | undefined | null, max: number): string;
-//# sourceMappingURL=wechatEmoji.d.ts.map

package/dist/src/utils/wechatEmoji.d.ts.map

@@ -1 +0,0 @@
-{"version":3,"file":"wechatEmoji.d.ts","sourceRoot":"","sources":["../../../src/utils/wechatEmoji.ts"],"names":[],"mappings":"AA4BA,wFAAwF;AACxF,wBAAgB,YAAY,CAAC,IAAI,EAAE,MAAM,GAAG,SAAS,GAAG,IAAI,GAAG,MAAM,CAGpE;AAED,uEAAuE;AACvE,wBAAgB,qBAAqB,CAAC,IAAI,EAAE,MAAM,GAAG,SAAS,GAAG,IAAI,EAAE,GAAG,EAAE,MAAM,GAAG,MAAM,CAE1F"}

package/dist/src/utils/wechatEmoji.js

@@ -1,35 +0,0 @@
-import { readFileSync } from 'fs';
-import { dirname, join } from 'path';
-import { fileURLToPath } from 'url';
-const __filename = fileURLToPath(import.meta.url);
-const __dirname = dirname(__filename);
-// dist/src/utils -> package root. The table is shared with the Python side.
-const DATA_FILE = join(__dirname, '..', '..', '..', 'scripts', 'data', 'wechat_emoji.json');
-const faces = {};
-try {
-    const parsed = JSON.parse(readFileSync(DATA_FILE, 'utf8'));
-    for (const [name, emoji] of Object.entries(parsed)) {
-        // Entries starting with "_" are documentation, not faces.
-        if (!name.startsWith('_') && typeof emoji === 'string') {
-            faces[name] = emoji;
-        }
-    }
-}
-catch {
-    // Missing table just means faces stay as their [名称] source text.
-}
-const escapeRe = (text) => text.replace(/[.*+?^${}()|[\]\\]/g, '\\$&');
-const names = Object.keys(faces).sort((a, b) => b.length - a.length);
-// Longest name first so [左哼哼] can never be shadowed by a shorter prefix.
-const pattern = names.length > 0 ? new RegExp(`\\[(${names.map(escapeRe).join('|')})\\]`, 'g') : null;
-/** Replace known WeChat face codes like [害羞] with emoji; unknown codes pass through. */
-export function convertFaces(text) {
-    if (!text || !pattern)
-        return text || '';
-    return text.replace(pattern, (match, name) => faces[name] ?? match);
-}
-/** Convert faces, then truncate without splitting a surrogate pair. */
-export function convertFacesTruncated(text, max) {
-    return Array.from(convertFaces(text)).slice(0, max).join('');
-}
-//# sourceMappingURL=wechatEmoji.js.map

package/dist/src/utils/wechatEmoji.js.map

@@ -1 +0,0 @@
-{"version":3,"file":"wechatEmoji.js","sourceRoot":"","sources":["../../../src/utils/wechatEmoji.ts"],"names":[],"mappings":"AAAA,OAAO,EAAE,YAAY,EAAE,MAAM,IAAI,CAAA;AACjC,OAAO,EAAE,OAAO,EAAE,IAAI,EAAE,MAAM,MAAM,CAAA;AACpC,OAAO,EAAE,aAAa,EAAE,MAAM,KAAK,CAAA;AAEnC,MAAM,UAAU,GAAG,aAAa,CAAC,MAAM,CAAC,IAAI,CAAC,GAAG,CAAC,CAAA;AACjD,MAAM,SAAS,GAAG,OAAO,CAAC,UAAU,CAAC,CAAA;AAErC,4EAA4E;AAC5E,MAAM,SAAS,GAAG,IAAI,CAAC,SAAS,EAAE,IAAI,EAAE,IAAI,EAAE,IAAI,EAAE,SAAS,EAAE,MAAM,EAAE,mBAAmB,CAAC,CAAA;AAE3F,MAAM,KAAK,GAA2B,EAAE,CAAA;AACxC,IAAI,CAAC;IACH,MAAM,MAAM,GAAG,IAAI,CAAC,KAAK,CAAC,YAAY,CAAC,SAAS,EAAE,MAAM,CAAC,CAA4B,CAAA;IACrF,KAAK,MAAM,CAAC,IAAI,EAAE,KAAK,CAAC,IAAI,MAAM,CAAC,OAAO,CAAC,MAAM,CAAC,EAAE,CAAC;QACnD,0DAA0D;QAC1D,IAAI,CAAC,IAAI,CAAC,UAAU,CAAC,GAAG,CAAC,IAAI,OAAO,KAAK,KAAK,QAAQ,EAAE,CAAC;YACvD,KAAK,CAAC,IAAI,CAAC,GAAG,KAAK,CAAA;QACrB,CAAC;IACH,CAAC;AACH,CAAC;AAAC,MAAM,CAAC;IACP,iEAAiE;AACnE,CAAC;AAED,MAAM,QAAQ,GAAG,CAAC,IAAY,EAAU,EAAE,CAAC,IAAI,CAAC,OAAO,CAAC,qBAAqB,EAAE,MAAM,CAAC,CAAA;AACtF,MAAM,KAAK,GAAG,MAAM,CAAC,IAAI,CAAC,KAAK,CAAC,CAAC,IAAI,CAAC,CAAC,CAAC,EAAE,CAAC,EAAE,EAAE,CAAC,CAAC,CAAC,MAAM,GAAG,CAAC,CAAC,MAAM,CAAC,CAAA;AACpE,yEAAyE;AACzE,MAAM,OAAO,GAAG,KAAK,CAAC,MAAM,GAAG,CAAC,CAAC,CAAC,CAAC,IAAI,MAAM,CAAC,OAAO,KAAK,CAAC,GAAG,CAAC,QAAQ,CAAC,CAAC,IAAI,CAAC,GAAG,CAAC,MAAM,EAAE,GAAG,CAAC,CAAC,CAAC,CAAC,IAAI,CAAA;AAErG,wFAAwF;AACxF,MAAM,UAAU,YAAY,CAAC,IAA+B;IAC1D,IAAI,CAAC,IAAI,IAAI,CAAC,OAAO;QAAE,OAAO,IAAI,IAAI,EAAE,CAAA;IACxC,OAAO,IAAI,CAAC,OAAO,CAAC,OAAO,EAAE,CAAC,KAAK,EAAE,IAAY,EAAE,EAAE,CAAC,KAAK,CAAC,IAAI,CAAC,IAAI,KAAK,CAAC,CAAA;AAC7E,CAAC;AAED,uEAAuE;AACvE,MAAM,UAAU,qBAAqB,CAAC,IAA+B,EAAE,GAAW;IAChF,OAAO,KAAK,CAAC,IAAI,CAAC,YAAY,CAAC,IAAI,CAAC,CAAC,CAAC,KAAK,CAAC,CAAC,EAAE,GAAG,CAAC,CAAC,IAAI,CAAC,EAAE,CAAC,CAAA;AAC9D,CAAC"}