mulmocast 2.6.8 → 2.6.9

package/lib/agents/tts_google_agent.js

@@ -2,6 +2,10 @@ import { GraphAILogger } from "graphai";
 import * as textToSpeech from "@google-cloud/text-to-speech";
 import { agentGenerationError, audioAction, audioFileTarget } from "../utils/error_cause.js";
 const client = new textToSpeech.TextToSpeechClient();
+// Hard cap so a hung Google TTS RPC can't pin a beat indefinitely.
+// Most synthesizeSpeech calls return in seconds; 60s leaves headroom
+// for long inputs and slow regions while still failing loud.
+const SYNTHESIZE_TIMEOUT_MS = 60_000;
 const getPrompt = (text, instructions) => {
     if (instructions) {
         return `### DIRECTOR'S NOTES\n${instructions}\n\n#### TRANSCRIPT\n${text}`;
@@ -37,7 +41,7 @@ export const ttsGoogleAgent = async ({ namedInputs, params }) => {
     };
     try {
         // Call the Text-to-Speech API
-        const [response] = await client.synthesizeSpeech(request);
+        const [response] = await client.synthesizeSpeech(request, { timeout: SYNTHESIZE_TIMEOUT_MS });
         return { buffer: response.audioContent };
     }
     catch (e) {
@@ -47,11 +51,23 @@ export const ttsGoogleAgent = async ({ namedInputs, params }) => {
             };
         }
         GraphAILogger.info(e);
-        throw new Error("TTS Google Error", {
+        // gRPC errors from @google-cloud/text-to-speech are ServiceError
+        // (extends Error with a `details` string). Surface that human-readable
+        // text so callers don't see only "TTS Google Error".
+        throw new Error(`TTS Google Error: ${grpcErrorDetail(e)}`, {
             cause: agentGenerationError("ttsGoogleAgent", audioAction, audioFileTarget),
         });
     }
 };
+const grpcErrorDetail = (e) => {
+    if (e instanceof Error) {
+        const details = e.details;
+        if (typeof details === "string" && details)
+            return details;
+        return e.message;
+    }
+    return String(e);
+};
 const ttsGoogleAgentInfo = {
     name: "ttsGoogleAgent",
     agent: ttsGoogleAgent,

package/lib/types/provider2agent.js

@@ -44,7 +44,7 @@ export const provider2TTSAgent = {
         models: ["mock-model"],
     },
 };
-export const gptImages = ["gpt-image-1.5", "gpt-image-1", "gpt-image-1-mini"];
+export const gptImages = ["gpt-image-2", "gpt-image-1.5", "gpt-image-1", "gpt-image-1-mini"];
 export const provider2ImageAgent = {
     openai: {
         agentName: "imageOpenaiAgent",

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "mulmocast",
-  "version": "2.6.8",
+  "version": "2.6.9",
   "description": "",
   "type": "module",
   "main": "lib/index.node.js",
@@ -109,11 +109,11 @@
     "dotenv": "^17.4.2",
     "fluent-ffmpeg": "^2.1.3",
     "graphai": "^2.0.16",
-    "jsdom": "^29.0.2",
+    "jsdom": "^29.1.0",
     "marked": "^18.0.2",
     "mulmocast-vision": "^1.0.9",
-    "ora": "^9.3.0",
-    "puppeteer": "^24.41.0",
+    "ora": "^9.4.0",
+    "puppeteer": "^24.42.0",
     "replicate": "^1.4.0",
     "yaml": "^2.8.3",
     "yargs": "^18.0.0",
@@ -136,7 +136,7 @@
     "prettier": "^3.8.3",
     "tsx": "^4.21.0",
     "typescript": "6.0.3",
-    "typescript-eslint": "^8.58.1"
+    "typescript-eslint": "^8.59.1"
   },
   "engines": {
     "node": ">=22.0.0"

package/scripts/test/README.md

@@ -106,7 +106,7 @@ Language setting tests
 Provider-specific feature tests
 
 - [**test_hello_google.json**](./test_hello_google.json) - Google TTS専用テスト / Google TTS specific test
-- [**gpt.json**](./gpt.json) - GPTモデルテスト / GPT model test
+- [**test_gpt_image.json**](./test_gpt_image.json) - GPT image model test
 - [**mulmo_story.json**](./mulmo_story.json) - ストーリー形式テスト / Story format test
 - [**nano_banana.json**](./nano_banana.json) - カスタムサンプル / Custom sample
 

package/scripts/test/{gpt.json → test_gpt_image.json}

@@ -21,6 +21,14 @@
     }
   },
   "beats": [
+    {
+      "speaker": "Host",
+      "text": "How are you?",
+      "imagePrompt": "A witch in Harajuku",
+      "imageParams": {
+        "model": "gpt-image-2"
+      }
+    },
     {
       "speaker": "Host",
       "text": "How are you?",