mulmocast 2.1.4 → 2.1.6

package/lib/agents/tts_elevenlabs_agent.js

@@ -1,6 +1,6 @@
 import { GraphAILogger } from "graphai";
 import { provider2TTSAgent } from "../utils/provider2agent.js";
-import { apiKeyMissingError, agentIncorrectAPIKeyError, agentGenerationError, audioAction, audioFileTarget } from "../utils/error_cause.js";
+import { apiKeyMissingError, agentVoiceLimitReachedError, agentIncorrectAPIKeyError, agentGenerationError, audioAction, audioFileTarget, } from "../utils/error_cause.js";
 export const ttsElevenlabsAgent = async ({ namedInputs, params, config, }) => {
     const { text } = namedInputs;
     const { voice, model, stability, similarityBoost, suppressError } = params;
@@ -57,6 +57,12 @@ export const ttsElevenlabsAgent = async ({ namedInputs, params, config, }) => {
                 cause: agentIncorrectAPIKeyError("ttsElevenlabsAgent", audioAction, audioFileTarget),
             });
         }
+        const errorDetail = await response.json();
+        if (errorDetail.detail.status === "voice_limit_reached") {
+            throw new Error("Failed to generate audio: 400 You have reached your maximum amount of custom voices with ElevenLabs", {
+                cause: agentVoiceLimitReachedError("ttsElevenlabsAgent", audioAction, audioFileTarget),
+            });
+        }
         throw new Error(`Eleven Labs API error: ${response.status} ${response.statusText}`, {
             cause: agentGenerationError("ttsElevenlabsAgent", audioAction, audioFileTarget),
         });

package/lib/agents/tts_gemini_agent.js

@@ -5,7 +5,7 @@ import { agentIncorrectAPIKeyError, apiKeyMissingError, agentGenerationError, au
 import { pcmToMp3 } from "../utils/ffmpeg_utils.js";
 export const ttsGeminiAgent = async ({ namedInputs, params, config, }) => {
     const { text } = namedInputs;
-    const { voice, suppressError } = params;
+    const { model, voice, suppressError } = params;
     const apiKey = config?.apiKey;
     if (!apiKey) {
         throw new Error("Google GenAI API key is required (GEMINI_API_KEY)", {
@@ -15,7 +15,7 @@ export const ttsGeminiAgent = async ({ namedInputs, params, config, }) => {
     try {
         const ai = new GoogleGenAI({ apiKey });
         const response = await ai.models.generateContent({
-            model: "gemini-2.5-flash-preview-tts",
+            model: model ?? provider2TTSAgent.gemini.defaultModel,
             contents: [{ parts: [{ text }] }],
             config: {
                 responseModalities: ["AUDIO"],

package/lib/types/agent.d.ts

@@ -121,6 +121,7 @@ export type KotodamaTTSAgentParams = TTSAgentParams & {
 };
 export type GoogleTTSAgentParams = TTSAgentParams & {
     speed: number;
+    model: string;
 };
 export type ElevenlabsTTSAgentParams = TTSAgentParams & {
     model: string;

package/lib/utils/error_cause.d.ts

@@ -54,6 +54,7 @@ export declare const apiKeyInvalidType = "apiKeyInvalid";
 export declare const apiRateLimitErrorType = "apiRateLimit";
 export declare const apiKeyMissingType = "apiKeyMissing";
 export declare const invalidResponseType = "invalidResponse";
+export declare const voiceLimitReachedType = "voice_limit_reached";
 export declare const movieAction = "movie";
 export declare const imageAction = "images";
 export declare const audioAction = "audio";
@@ -174,6 +175,12 @@ export declare const agentIncorrectAPIKeyError: (agentName: string, action: stri
     target: string;
     agentName: string;
 };
+export declare const agentVoiceLimitReachedError: (agentName: string, action: string, target: string) => {
+    type: string;
+    action: string;
+    target: string;
+    agentName: string;
+};
 export declare const agentAPIRateLimitError: (agentName: string, action: string, target: string, beatIndex?: number) => {
     beatIndex?: number | undefined;
     type: string;

package/lib/utils/error_cause.js

@@ -55,6 +55,7 @@ export const apiKeyInvalidType = "apiKeyInvalid";
 export const apiRateLimitErrorType = "apiRateLimit";
 export const apiKeyMissingType = "apiKeyMissing";
 export const invalidResponseType = "invalidResponse";
+export const voiceLimitReachedType = "voice_limit_reached";
 // Actions
 export const movieAction = "movie";
 export const imageAction = "images";
@@ -194,6 +195,14 @@ export const agentIncorrectAPIKeyError = (agentName, action, target, beatIndex)
         ...(beatIndex !== undefined && { beatIndex }),
     };
 };
+export const agentVoiceLimitReachedError = (agentName, action, target) => {
+    return {
+        type: voiceLimitReachedType,
+        action,
+        target,
+        agentName,
+    };
+};
 // Agent API/Rate Limit Errors
 export const agentAPIRateLimitError = (agentName, action, target, beatIndex) => {
     return {

package/lib/utils/filters.js

@@ -17,6 +17,10 @@ export const nijovoiceTextAgentFilter = async (context, next) => {
 };
 export const fileCacheAgentFilter = async (context, next) => {
     const { force, file, index, mulmoContext, sessionType, id, withBackup } = context.namedInputs.cache;
+    const fileName = path.resolve(path.dirname(file), id + ".mp3");
+    if (fs.existsSync(file)) {
+        fs.copyFileSync(file, fileName);
+    }
     const shouldUseCache = async () => {
         if (force && force.some((element) => element)) {
             return false;

package/lib/utils/provider2agent.d.ts

@@ -20,7 +20,9 @@ export declare const provider2TTSAgent: {
     gemini: {
         agentName: string;
         hasLimitedConcurrency: boolean;
+        defaultModel: string;
         defaultVoice: string;
+        models: string[];
         keyName: string;
     };
     elevenlabs: {

package/lib/utils/provider2agent.js

@@ -21,7 +21,9 @@ export const provider2TTSAgent = {
     gemini: {
         agentName: "ttsGeminiAgent",
         hasLimitedConcurrency: false,
+        defaultModel: "gemini-2.5-flash-preview-tts",
         defaultVoice: "Kore",
+        models: ["gemini-2.5-flash-preview-tts", "gemini-2.5-pro-preview-tts"],
         keyName: "GEMINI_API_KEY",
     },
     elevenlabs: {

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "mulmocast",
-  "version": "2.1.4",
+  "version": "2.1.6",
   "description": "",
   "type": "module",
   "main": "lib/index.node.js",

package/scripts/test/test_audio_gemini.json

@@ -12,6 +12,7 @@
       },
       "Presenter2": {
         "provider": "gemini",
+        "model": "gemini-2.5-pro-preview-tts",
         "voiceId": "Puck"
       }
     }