@daolmedo/openclaw-twilio-whatsapp 1.1.13 → 1.1.16

package/dist/index.js

@@ -194,8 +194,11 @@ var twilioWhatsappPlugin = {
 };
 
 // src/http-routes.ts
+import { readFile } from "node:fs/promises";
+import path from "node:path";
 import { dispatchInboundDirectDmWithRuntime } from "openclaw/plugin-sdk/direct-dm";
 import { saveMediaBuffer } from "openclaw/plugin-sdk/media-runtime";
+import { transcribeOpenAiCompatibleAudio } from "openclaw/plugin-sdk/media-understanding";
 
 // src/webhook.ts
 import twilio2 from "twilio";
@@ -375,10 +378,29 @@ function registerTwilioWhatsappHttpRoutes(api) {
       const firstAudio = resolvedMedia.find((m) => m.contentType.startsWith("audio/"));
       if (firstAudio) {
         try {
-          const { text } = await api.runtime.mediaUnderstanding.transcribeAudioFile({
-            filePath: firstAudio.path,
+          const cfgAny = stored.cfg;
+          const agentId = cfgAny.bindings?.find(
+            (b) => b.match?.channel === "twilio-whatsapp" && b.match?.accountId === stored.accountId
+          )?.agentId;
+          const agentDir = cfgAny.agents?.list?.find(
+            (a) => a.id === agentId
+          )?.agentDir;
+          const auth = await api.runtime.modelAuth.resolveApiKeyForProvider({
+            provider: "openai",
             cfg: api.config,
-            mime: firstAudio.contentType
+            agentDir
+          });
+          if (!auth.apiKey) throw new Error("No OpenAI API key available for audio transcription");
+          const audioBuffer = await readFile(firstAudio.path);
+          const { text } = await transcribeOpenAiCompatibleAudio({
+            buffer: audioBuffer,
+            fileName: path.basename(firstAudio.path),
+            mime: firstAudio.contentType,
+            apiKey: auth.apiKey,
+            model: "gpt-4o-mini-transcribe",
+            defaultBaseUrl: "https://api.openai.com/v1",
+            defaultModel: "gpt-4o-mini-transcribe",
+            timeoutMs: 12e4
           });
           console.log("[twilio-whatsapp] Transcription result:", text);
           bodyForAgent = text || messageText || "<media:audio>";

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@daolmedo/openclaw-twilio-whatsapp",
-  "version": "1.1.13",
+  "version": "1.1.16",
   "description": "Twilio WhatsApp channel plugin for OpenClaw (Cody ISV)",
   "type": "module",
   "exports": {