@yaoyuanchao/dingtalk 1.3.4 → 1.3.6

package/CHANGELOG.md

@@ -5,6 +5,46 @@ All notable changes to this project will be documented in this file.
 The format is based on [Keep a Changelog](https://keepachangelog.com/en/1.0.0/),
 and this project adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0.html).
 
+## [1.3.6] - 2026-01-28
+
+### Fixed
+
+- **Stream ACK method name** — corrected `socketResponse()` to `socketCallBackResponse()` (the actual SDK method); previous typo caused ACK to silently fail, triggering DingTalk 60-second retry
+- **Audio message handling** — skip .amr file download when DingTalk ASR recognition text is available; prevents agent from being confused by audio attachment and trying Whisper instead of reading the already-transcribed text
+
+## [1.3.5] - 2026-01-28
+
+### Fixed
+
+- **Outbound `to` parameter parsing** — bare userId (no `dm:` prefix) now correctly treated as DM target; previously silently dropped with ok:true
+- **SessionWebhook response validation** — `sendViaSessionWebhook()` and `sendMarkdownViaSessionWebhook()` now return errcode/errmsg and check `.ok`; failures trigger REST API fallback instead of being silently ignored
+- **Stream ACK timing** — immediately call `socketResponse()` on message receipt to prevent DingTalk 60-second retry timeout; previously awaited full AI processing before ACK
+- **`resolveDeliverText()` type safety** — check `typeof payload.markdown === 'string'` to avoid treating boolean flags as text content
+
+### Changed
+
+- **`parseOutboundTo()` enhanced** — handles `"dm:id"`, `"group:id"`, `"dingtalk:dm:id"`, `"dingtalk:group:id"`, and bare `"id"` (defaults to DM)
+- **`deliverReply()` error propagation** — throws on sessionWebhook rejection to trigger retry + REST API fallback
+- **Media URL merging** — `resolveDeliverText()` merges `payload.mediaUrl`/`payload.imageUrl` into text as markdown image syntax
+- **Webhook functions** return `{ ok, errcode, errmsg }` for proper error inspection
+
+## [1.3.0] - 2026-01-28
+
+### Added
+
+- **Full SDK Pipeline** — runtime feature detection for `dispatchReplyFromConfig` with 9-step SDK integration (routing, session, envelope, dispatch)
+- **Media support** — image download via `downloadPicture()`, audio/video/file recognition via `downloadMediaFile()`
+- **Smart Markdown detection** — `messageFormat: 'auto'` option with regex-based content detection
+- **Thinking indicator** — `showThinking` config option sends "正在思考..." before AI processing
+- **Activity recording** — `runtime.channel.activity.record()` calls for start/stop/message events
+- **`cleanupOldMedia()`** — generalized media cleanup (replaces `cleanupOldPictures`)
+
+### Changed
+
+- **Message extraction refactored** — `extractMessageContent()` switch-case structure for text/richText/picture/audio/video/file
+- **Config schema** — added `showThinking`, `messageFormat: 'auto'` option
+- **`sendMedia()` outbound** — uses markdown image syntax instead of plain URL text
+
 ## [1.2.0] - 2026-01-28
 
 ### 🎉 Major Features - Official Plugin Release

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@yaoyuanchao/dingtalk",
-  "version": "1.3.4",
+  "version": "1.3.6",
   "type": "module",
   "description": "DingTalk channel plugin for Clawdbot with Stream Mode support",
   "license": "MIT",

package/src/channel.ts

@@ -6,14 +6,21 @@ import { probeDingTalk } from './probe.js';
 
 /**
  * Parse outbound `to` address, stripping optional channel prefix.
- * Handles both "dm:id" and "dingtalk:dm:id" formats.
+ * Handles: "dm:id", "group:id", "dingtalk:dm:id", "dingtalk:group:id",
+ * and bare "id" (treated as DM userId).
  */
 function parseOutboundTo(to: string): { type: string; id: string } {
   const parts = to.split(':');
+  // Strip channel prefix: "dingtalk:dm:id" → "dm:id"
   if (parts[0] === 'dingtalk' && parts.length > 2) {
     parts.shift();
   }
-  return { type: parts[0], id: parts.slice(1).join(':') };
+  // Known types
+  if (parts[0] === 'dm' || parts[0] === 'group') {
+    return { type: parts[0], id: parts.slice(1).join(':') };
+  }
+  // Bare ID (no type prefix) — treat as DM userId
+  return { type: 'dm', id: to };
 }
 
 export const dingtalkPlugin = {

package/src/monitor.ts

@@ -53,8 +53,9 @@ export async function startDingTalkMonitor(ctx: DingTalkMonitorContext): Promise
 
   client.registerCallbackListener(TOPIC_ROBOT, async (downstream: any) => {
     // Immediately ACK to prevent DingTalk from retrying (60s timeout)
+    // SDK method is socketCallBackResponse, not socketResponse
     try {
-      client.socketResponse(downstream.headers.messageId, { status: 'SUCCESS' });
+      client.socketCallBackResponse(downstream.headers.messageId, { status: 'SUCCESS' });
     } catch (_) { /* best-effort ACK */ }
 
     try {
@@ -318,11 +319,16 @@ async function processInboundMessage(
   // Extract message content using structured extractor
   const extracted = await extractMessageContent(msg, account, log);
 
-  // Download media if present (picture/audio/video/file)
+  // Download media if present (picture/video/file — but skip audio when ASR text exists)
   let mediaPath: string | undefined;
   let mediaType: string | undefined;
 
-  if (extracted.mediaDownloadCode && account.clientId && account.clientSecret) {
+  // For audio messages with successful ASR recognition, use the text directly
+  // and skip downloading the .amr file (which would confuse the agent into
+  // trying Whisper instead of reading the already-transcribed text).
+  const skipMediaDownload = extracted.messageType === 'audio' && !!extracted.text;
+
+  if (!skipMediaDownload && extracted.mediaDownloadCode && account.clientId && account.clientSecret) {
     const robotCode = account.robotCode || account.clientId;
     try {
       const result = await downloadMediaFile(
@@ -342,6 +348,8 @@ async function processInboundMessage(
     } catch (err) {
       log?.warn?.(`[dingtalk] Media download error: ${err}`);
     }
+  } else if (skipMediaDownload) {
+    log?.info?.("[dingtalk] Audio ASR text available, skipping .amr download");
   }
 
   let rawBody = extracted.text;