mulmocast 2.1.21 → 2.1.23

package/lib/agents/image_openai_agent.js

@@ -1,21 +1,22 @@
 import fs from "fs";
 import path from "path";
 import { GraphAILogger } from "graphai";
-import OpenAI, { toFile, AuthenticationError, RateLimitError, APIError } from "openai";
+import { toFile, AuthenticationError, RateLimitError, APIError } from "openai";
+import { createOpenAIClient } from "../utils/openai_client.js";
 import { provider2ImageAgent, gptImages } from "../types/provider2agent.js";
 import { apiKeyMissingError, agentGenerationError, openAIAgentGenerationError, agentIncorrectAPIKeyError, agentAPIRateLimitError, agentInvalidResponseError, imageAction, imageFileTarget, } from "../utils/error_cause.js";
 // https://platform.openai.com/docs/guides/image-generation
 export const imageOpenaiAgent = async ({ namedInputs, params, config, }) => {
     const { prompt, referenceImages } = namedInputs;
     const { moderation, canvasSize, quality } = params;
-    const { apiKey, baseURL } = { ...config };
+    const { apiKey, baseURL, apiVersion } = { ...config };
     if (!apiKey) {
         throw new Error("OpenAI API key is required (OPENAI_API_KEY)", {
             cause: apiKeyMissingError("imageOpenaiAgent", imageAction, "OPENAI_API_KEY"),
         });
     }
     const model = params.model ?? provider2ImageAgent["openai"].defaultModel;
-    const openai = new OpenAI({ apiKey, baseURL });
+    const openai = createOpenAIClient({ apiKey, baseURL, apiVersion });
     const size = (() => {
         if (gptImages.includes(model)) {
             if (canvasSize.width > canvasSize.height) {

package/lib/agents/tts_openai_agent.js

@@ -1,17 +1,18 @@
 import { GraphAILogger } from "graphai";
-import OpenAI, { AuthenticationError, RateLimitError } from "openai";
+import { AuthenticationError, RateLimitError } from "openai";
 import { provider2TTSAgent } from "../types/provider2agent.js";
+import { createOpenAIClient } from "../utils/openai_client.js";
 import { apiKeyMissingError, agentIncorrectAPIKeyError, agentAPIRateLimitError, agentGenerationError, audioAction, audioFileTarget, } from "../utils/error_cause.js";
 export const ttsOpenaiAgent = async ({ namedInputs, params, config, }) => {
     const { text } = namedInputs;
     const { model, voice, suppressError, instructions, speed } = params;
-    const { apiKey, baseURL } = config ?? {};
+    const { apiKey, baseURL, apiVersion } = config ?? {};
     if (!apiKey) {
         throw new Error("OpenAI API key is required (OPENAI_API_KEY)", {
             cause: apiKeyMissingError("ttsOpenaiAgent", audioAction, "OPENAI_API_KEY"),
         });
     }
-    const openai = new OpenAI({ apiKey, baseURL });
+    const openai = createOpenAIClient({ apiKey, baseURL, apiVersion });
     try {
         const tts_options = {
             model: model ?? provider2TTSAgent.openai.defaultModel,

package/lib/types/agent.d.ts

@@ -58,6 +58,7 @@ export type ReplicateImageAgentParams = {
 export type OpenAIImageAgentConfig = {
     baseURL?: string;
     apiKey?: string;
+    apiVersion?: string;
 };
 export type GoogleImageAgentConfig = {
     projectId?: string;

package/lib/types/provider2agent.d.ts

@@ -139,6 +139,7 @@ export declare const provider2LLMAgent: {
         readonly defaultModel: "gpt-5";
         readonly keyName: "OPENAI_API_KEY";
         readonly baseURLKeyName: "OPENAI_BASE_URL";
+        readonly apiVersionKeyName: "OPENAI_API_VERSION";
         readonly max_tokens: 8192;
         readonly models: readonly ["gpt-5", "gpt-5-nano", "gpt-5-mini", "gpt-4.1", "gpt-4.1-mini", "gpt-4.1-nano", "o3", "o3-mini", "o3-pro", "o1", "o1-pro", "gpt-4o", "gpt-4o-mini"];
     };

package/lib/types/provider2agent.js

@@ -252,6 +252,7 @@ export const provider2LLMAgent = {
         defaultModel: "gpt-5",
         keyName: "OPENAI_API_KEY",
         baseURLKeyName: "OPENAI_BASE_URL",
+        apiVersionKeyName: "OPENAI_API_VERSION",
         max_tokens: 8192,
         models: [
             "gpt-5",

package/lib/types/schema.d.ts

@@ -43,6 +43,8 @@ export declare const speakerDataSchema: z.ZodObject<{
         [x: string]: string;
     }>>>;
     model: z.ZodOptional<z.ZodString>;
+    baseURL: z.ZodOptional<z.ZodString>;
+    apiVersion: z.ZodOptional<z.ZodString>;
 }, z.core.$strict>;
 export declare const speakerSchema: z.ZodObject<{
     displayName: z.ZodOptional<z.ZodRecord<z.ZodString, z.ZodString>>;
@@ -59,6 +61,8 @@ export declare const speakerSchema: z.ZodObject<{
         [x: string]: string;
     }>>>;
     model: z.ZodOptional<z.ZodString>;
+    baseURL: z.ZodOptional<z.ZodString>;
+    apiVersion: z.ZodOptional<z.ZodString>;
     lang: z.ZodOptional<z.ZodRecord<z.ZodString, z.ZodObject<{
         displayName: z.ZodOptional<z.ZodRecord<z.ZodString, z.ZodString>>;
         voiceId: z.ZodString;
@@ -74,6 +78,8 @@ export declare const speakerSchema: z.ZodObject<{
             [x: string]: string;
         }>>>;
         model: z.ZodOptional<z.ZodString>;
+        baseURL: z.ZodOptional<z.ZodString>;
+        apiVersion: z.ZodOptional<z.ZodString>;
     }, z.core.$strict>>>;
 }, z.core.$strict>;
 export declare const speakerDictionarySchema: z.ZodRecord<z.ZodString, z.ZodObject<{
@@ -91,6 +97,8 @@ export declare const speakerDictionarySchema: z.ZodRecord<z.ZodString, z.ZodObje
         [x: string]: string;
     }>>>;
     model: z.ZodOptional<z.ZodString>;
+    baseURL: z.ZodOptional<z.ZodString>;
+    apiVersion: z.ZodOptional<z.ZodString>;
     lang: z.ZodOptional<z.ZodRecord<z.ZodString, z.ZodObject<{
         displayName: z.ZodOptional<z.ZodRecord<z.ZodString, z.ZodString>>;
         voiceId: z.ZodString;
@@ -106,6 +114,8 @@ export declare const speakerDictionarySchema: z.ZodRecord<z.ZodString, z.ZodObje
             [x: string]: string;
         }>>>;
         model: z.ZodOptional<z.ZodString>;
+        baseURL: z.ZodOptional<z.ZodString>;
+        apiVersion: z.ZodOptional<z.ZodString>;
     }, z.core.$strict>>>;
 }, z.core.$strict>>;
 export declare const mulmoSpeechParamsSchema: z.ZodDefault<z.ZodObject<{
@@ -124,6 +134,8 @@ export declare const mulmoSpeechParamsSchema: z.ZodDefault<z.ZodObject<{
             [x: string]: string;
         }>>>;
         model: z.ZodOptional<z.ZodString>;
+        baseURL: z.ZodOptional<z.ZodString>;
+        apiVersion: z.ZodOptional<z.ZodString>;
         lang: z.ZodOptional<z.ZodRecord<z.ZodString, z.ZodObject<{
             displayName: z.ZodOptional<z.ZodRecord<z.ZodString, z.ZodString>>;
             voiceId: z.ZodString;
@@ -139,6 +151,8 @@ export declare const mulmoSpeechParamsSchema: z.ZodDefault<z.ZodObject<{
                 [x: string]: string;
             }>>>;
             model: z.ZodOptional<z.ZodString>;
+            baseURL: z.ZodOptional<z.ZodString>;
+            apiVersion: z.ZodOptional<z.ZodString>;
         }, z.core.$strict>>>;
     }, z.core.$strict>>;
 }, z.core.$strip>>;
@@ -437,6 +451,8 @@ export declare const mulmoBeatImageParamsSchema: z.ZodObject<{
     quality: z.ZodOptional<z.ZodString>;
     style: z.ZodOptional<z.ZodString>;
     moderation: z.ZodOptional<z.ZodString>;
+    baseURL: z.ZodOptional<z.ZodString>;
+    apiVersion: z.ZodOptional<z.ZodString>;
 }, z.core.$strict>;
 export declare const mulmoImageParamsSchema: z.ZodObject<{
     provider: z.ZodOptional<z.ZodEnum<{
@@ -446,6 +462,8 @@ export declare const mulmoImageParamsSchema: z.ZodObject<{
     quality: z.ZodOptional<z.ZodString>;
     style: z.ZodOptional<z.ZodString>;
     moderation: z.ZodOptional<z.ZodString>;
+    baseURL: z.ZodOptional<z.ZodString>;
+    apiVersion: z.ZodOptional<z.ZodString>;
     images: z.ZodOptional<z.ZodRecord<z.ZodString, z.ZodUnion<readonly [z.ZodObject<{
         type: z.ZodLiteral<"image">;
         source: z.ZodDiscriminatedUnion<[z.ZodObject<{
@@ -891,6 +909,8 @@ export declare const mulmoBeatSchema: z.ZodObject<{
         quality: z.ZodOptional<z.ZodString>;
         style: z.ZodOptional<z.ZodString>;
         moderation: z.ZodOptional<z.ZodString>;
+        baseURL: z.ZodOptional<z.ZodString>;
+        apiVersion: z.ZodOptional<z.ZodString>;
     }, z.core.$strict>>;
     audioParams: z.ZodOptional<z.ZodObject<{
         padding: z.ZodOptional<z.ZodNumber>;
@@ -1223,6 +1243,8 @@ export declare const mulmoPresentationStyleSchema: z.ZodObject<{
                 [x: string]: string;
             }>>>;
             model: z.ZodOptional<z.ZodString>;
+            baseURL: z.ZodOptional<z.ZodString>;
+            apiVersion: z.ZodOptional<z.ZodString>;
             lang: z.ZodOptional<z.ZodRecord<z.ZodString, z.ZodObject<{
                 displayName: z.ZodOptional<z.ZodRecord<z.ZodString, z.ZodString>>;
                 voiceId: z.ZodString;
@@ -1238,6 +1260,8 @@ export declare const mulmoPresentationStyleSchema: z.ZodObject<{
                     [x: string]: string;
                 }>>>;
                 model: z.ZodOptional<z.ZodString>;
+                baseURL: z.ZodOptional<z.ZodString>;
+                apiVersion: z.ZodOptional<z.ZodString>;
             }, z.core.$strict>>>;
         }, z.core.$strict>>;
     }, z.core.$strip>>;
@@ -1249,6 +1273,8 @@ export declare const mulmoPresentationStyleSchema: z.ZodObject<{
         quality: z.ZodOptional<z.ZodString>;
         style: z.ZodOptional<z.ZodString>;
         moderation: z.ZodOptional<z.ZodString>;
+        baseURL: z.ZodOptional<z.ZodString>;
+        apiVersion: z.ZodOptional<z.ZodString>;
         images: z.ZodOptional<z.ZodRecord<z.ZodString, z.ZodUnion<readonly [z.ZodObject<{
             type: z.ZodLiteral<"image">;
             source: z.ZodDiscriminatedUnion<[z.ZodObject<{
@@ -1584,6 +1610,8 @@ export declare const mulmoScriptSchema: z.ZodObject<{
                 [x: string]: string;
             }>>>;
             model: z.ZodOptional<z.ZodString>;
+            baseURL: z.ZodOptional<z.ZodString>;
+            apiVersion: z.ZodOptional<z.ZodString>;
             lang: z.ZodOptional<z.ZodRecord<z.ZodString, z.ZodObject<{
                 displayName: z.ZodOptional<z.ZodRecord<z.ZodString, z.ZodString>>;
                 voiceId: z.ZodString;
@@ -1599,6 +1627,8 @@ export declare const mulmoScriptSchema: z.ZodObject<{
                     [x: string]: string;
                 }>>>;
                 model: z.ZodOptional<z.ZodString>;
+                baseURL: z.ZodOptional<z.ZodString>;
+                apiVersion: z.ZodOptional<z.ZodString>;
             }, z.core.$strict>>>;
         }, z.core.$strict>>;
     }, z.core.$strip>>;
@@ -1610,6 +1640,8 @@ export declare const mulmoScriptSchema: z.ZodObject<{
         quality: z.ZodOptional<z.ZodString>;
         style: z.ZodOptional<z.ZodString>;
         moderation: z.ZodOptional<z.ZodString>;
+        baseURL: z.ZodOptional<z.ZodString>;
+        apiVersion: z.ZodOptional<z.ZodString>;
         images: z.ZodOptional<z.ZodRecord<z.ZodString, z.ZodUnion<readonly [z.ZodObject<{
             type: z.ZodLiteral<"image">;
             source: z.ZodDiscriminatedUnion<[z.ZodObject<{
@@ -2049,6 +2081,8 @@ export declare const mulmoScriptSchema: z.ZodObject<{
             quality: z.ZodOptional<z.ZodString>;
             style: z.ZodOptional<z.ZodString>;
             moderation: z.ZodOptional<z.ZodString>;
+            baseURL: z.ZodOptional<z.ZodString>;
+            apiVersion: z.ZodOptional<z.ZodString>;
         }, z.core.$strict>>;
         audioParams: z.ZodOptional<z.ZodObject<{
             padding: z.ZodOptional<z.ZodNumber>;
@@ -2456,6 +2490,8 @@ export declare const mulmoStudioSchema: z.ZodObject<{
                     [x: string]: string;
                 }>>>;
                 model: z.ZodOptional<z.ZodString>;
+                baseURL: z.ZodOptional<z.ZodString>;
+                apiVersion: z.ZodOptional<z.ZodString>;
                 lang: z.ZodOptional<z.ZodRecord<z.ZodString, z.ZodObject<{
                     displayName: z.ZodOptional<z.ZodRecord<z.ZodString, z.ZodString>>;
                     voiceId: z.ZodString;
@@ -2471,6 +2507,8 @@ export declare const mulmoStudioSchema: z.ZodObject<{
                         [x: string]: string;
                     }>>>;
                     model: z.ZodOptional<z.ZodString>;
+                    baseURL: z.ZodOptional<z.ZodString>;
+                    apiVersion: z.ZodOptional<z.ZodString>;
                 }, z.core.$strict>>>;
             }, z.core.$strict>>;
         }, z.core.$strip>>;
@@ -2482,6 +2520,8 @@ export declare const mulmoStudioSchema: z.ZodObject<{
             quality: z.ZodOptional<z.ZodString>;
             style: z.ZodOptional<z.ZodString>;
             moderation: z.ZodOptional<z.ZodString>;
+            baseURL: z.ZodOptional<z.ZodString>;
+            apiVersion: z.ZodOptional<z.ZodString>;
             images: z.ZodOptional<z.ZodRecord<z.ZodString, z.ZodUnion<readonly [z.ZodObject<{
                 type: z.ZodLiteral<"image">;
                 source: z.ZodDiscriminatedUnion<[z.ZodObject<{
@@ -2921,6 +2961,8 @@ export declare const mulmoStudioSchema: z.ZodObject<{
                 quality: z.ZodOptional<z.ZodString>;
                 style: z.ZodOptional<z.ZodString>;
                 moderation: z.ZodOptional<z.ZodString>;
+                baseURL: z.ZodOptional<z.ZodString>;
+                apiVersion: z.ZodOptional<z.ZodString>;
             }, z.core.$strict>>;
             audioParams: z.ZodOptional<z.ZodObject<{
                 padding: z.ZodOptional<z.ZodNumber>;
@@ -3264,6 +3306,8 @@ export declare const mulmoPromptTemplateSchema: z.ZodObject<{
                     [x: string]: string;
                 }>>>;
                 model: z.ZodOptional<z.ZodString>;
+                baseURL: z.ZodOptional<z.ZodString>;
+                apiVersion: z.ZodOptional<z.ZodString>;
                 lang: z.ZodOptional<z.ZodRecord<z.ZodString, z.ZodObject<{
                     displayName: z.ZodOptional<z.ZodRecord<z.ZodString, z.ZodString>>;
                     voiceId: z.ZodString;
@@ -3279,6 +3323,8 @@ export declare const mulmoPromptTemplateSchema: z.ZodObject<{
                         [x: string]: string;
                     }>>>;
                     model: z.ZodOptional<z.ZodString>;
+                    baseURL: z.ZodOptional<z.ZodString>;
+                    apiVersion: z.ZodOptional<z.ZodString>;
                 }, z.core.$strict>>>;
             }, z.core.$strict>>;
         }, z.core.$strip>>;
@@ -3290,6 +3336,8 @@ export declare const mulmoPromptTemplateSchema: z.ZodObject<{
             quality: z.ZodOptional<z.ZodString>;
             style: z.ZodOptional<z.ZodString>;
             moderation: z.ZodOptional<z.ZodString>;
+            baseURL: z.ZodOptional<z.ZodString>;
+            apiVersion: z.ZodOptional<z.ZodString>;
             images: z.ZodOptional<z.ZodRecord<z.ZodString, z.ZodUnion<readonly [z.ZodObject<{
                 type: z.ZodLiteral<"image">;
                 source: z.ZodDiscriminatedUnion<[z.ZodObject<{
@@ -3619,6 +3667,8 @@ export declare const mulmoPromptTemplateFileSchema: z.ZodObject<{
                     [x: string]: string;
                 }>>>;
                 model: z.ZodOptional<z.ZodString>;
+                baseURL: z.ZodOptional<z.ZodString>;
+                apiVersion: z.ZodOptional<z.ZodString>;
                 lang: z.ZodOptional<z.ZodRecord<z.ZodString, z.ZodObject<{
                     displayName: z.ZodOptional<z.ZodRecord<z.ZodString, z.ZodString>>;
                     voiceId: z.ZodString;
@@ -3634,6 +3684,8 @@ export declare const mulmoPromptTemplateFileSchema: z.ZodObject<{
                         [x: string]: string;
                     }>>>;
                     model: z.ZodOptional<z.ZodString>;
+                    baseURL: z.ZodOptional<z.ZodString>;
+                    apiVersion: z.ZodOptional<z.ZodString>;
                 }, z.core.$strict>>>;
             }, z.core.$strict>>;
         }, z.core.$strip>>;
@@ -3645,6 +3697,8 @@ export declare const mulmoPromptTemplateFileSchema: z.ZodObject<{
             quality: z.ZodOptional<z.ZodString>;
             style: z.ZodOptional<z.ZodString>;
             moderation: z.ZodOptional<z.ZodString>;
+            baseURL: z.ZodOptional<z.ZodString>;
+            apiVersion: z.ZodOptional<z.ZodString>;
             images: z.ZodOptional<z.ZodRecord<z.ZodString, z.ZodUnion<readonly [z.ZodObject<{
                 type: z.ZodLiteral<"image">;
                 source: z.ZodDiscriminatedUnion<[z.ZodObject<{

package/lib/types/schema.js

@@ -39,6 +39,8 @@ export const speakerDataSchema = z
     speechOptions: speechOptionsSchema.optional(),
     provider: text2SpeechProviderSchema.optional(),
     model: z.string().optional().describe("TTS model to use for this speaker"),
+    baseURL: z.string().optional(), // Azure/custom endpoint URL
+    apiVersion: z.string().optional(), // Azure API version (e.g., "2025-04-01-preview")
 })
     .strict();
 export const speakerSchema = speakerDataSchema.extend({
@@ -240,6 +242,8 @@ export const mulmoBeatImageParamsSchema = z
     quality: z.string().optional(), // optional image quality (model specific)
     style: z.string().optional(), // optional image style
     moderation: z.string().optional(), // optional image style
+    baseURL: z.string().optional(), // Azure/custom endpoint URL
+    apiVersion: z.string().optional(), // Azure API version (e.g., "2025-04-01-preview")
 })
     .strict();
 export const mulmoImageParamsSchema = mulmoBeatImageParamsSchema

package/lib/utils/context.d.ts

@@ -24,6 +24,8 @@ export declare const createStudioData: (_mulmoScript: MulmoScript, fileName: str
                 } | undefined;
                 provider?: string | undefined;
                 model?: string | undefined;
+                baseURL?: string | undefined;
+                apiVersion?: string | undefined;
                 lang?: Record<string, {
                     voiceId: string;
                     displayName?: Record<string, string> | undefined;
@@ -37,6 +39,8 @@ export declare const createStudioData: (_mulmoScript: MulmoScript, fileName: str
                     } | undefined;
                     provider?: string | undefined;
                     model?: string | undefined;
+                    baseURL?: string | undefined;
+                    apiVersion?: string | undefined;
                 }> | undefined;
             }>;
         };
@@ -46,6 +50,8 @@ export declare const createStudioData: (_mulmoScript: MulmoScript, fileName: str
             quality?: string | undefined;
             style?: string | undefined;
             moderation?: string | undefined;
+            baseURL?: string | undefined;
+            apiVersion?: string | undefined;
             images?: Record<string, {
                 type: "image";
                 source: {
@@ -369,6 +375,8 @@ export declare const createStudioData: (_mulmoScript: MulmoScript, fileName: str
                 quality?: string | undefined;
                 style?: string | undefined;
                 moderation?: string | undefined;
+                baseURL?: string | undefined;
+                apiVersion?: string | undefined;
             } | undefined;
             audioParams?: {
                 movieVolume: number;
@@ -664,6 +672,8 @@ export declare const initializeContextFromFiles: (files: FileObject, raiseError:
                     } | undefined;
                     provider?: string | undefined;
                     model?: string | undefined;
+                    baseURL?: string | undefined;
+                    apiVersion?: string | undefined;
                     lang?: Record<string, {
                         voiceId: string;
                         displayName?: Record<string, string> | undefined;
@@ -677,6 +687,8 @@ export declare const initializeContextFromFiles: (files: FileObject, raiseError:
                         } | undefined;
                         provider?: string | undefined;
                         model?: string | undefined;
+                        baseURL?: string | undefined;
+                        apiVersion?: string | undefined;
                     }> | undefined;
                 }>;
             };
@@ -686,6 +698,8 @@ export declare const initializeContextFromFiles: (files: FileObject, raiseError:
                 quality?: string | undefined;
                 style?: string | undefined;
                 moderation?: string | undefined;
+                baseURL?: string | undefined;
+                apiVersion?: string | undefined;
                 images?: Record<string, {
                     type: "image";
                     source: {
@@ -1009,6 +1023,8 @@ export declare const initializeContextFromFiles: (files: FileObject, raiseError:
                     quality?: string | undefined;
                     style?: string | undefined;
                     moderation?: string | undefined;
+                    baseURL?: string | undefined;
+                    apiVersion?: string | undefined;
                 } | undefined;
                 audioParams?: {
                     movieVolume: number;
@@ -1311,6 +1327,8 @@ export declare const initializeContextFromFiles: (files: FileObject, raiseError:
                 } | undefined;
                 provider?: string | undefined;
                 model?: string | undefined;
+                baseURL?: string | undefined;
+                apiVersion?: string | undefined;
                 lang?: Record<string, {
                     voiceId: string;
                     displayName?: Record<string, string> | undefined;
@@ -1324,6 +1342,8 @@ export declare const initializeContextFromFiles: (files: FileObject, raiseError:
                     } | undefined;
                     provider?: string | undefined;
                     model?: string | undefined;
+                    baseURL?: string | undefined;
+                    apiVersion?: string | undefined;
                 }> | undefined;
             }>;
         };
@@ -1333,6 +1353,8 @@ export declare const initializeContextFromFiles: (files: FileObject, raiseError:
             quality?: string | undefined;
             style?: string | undefined;
             moderation?: string | undefined;
+            baseURL?: string | undefined;
+            apiVersion?: string | undefined;
             images?: Record<string, {
                 type: "image";
                 source: {

package/lib/utils/openai_client.d.ts

@@ -0,0 +1,17 @@
+import OpenAI from "openai";
+export interface OpenAIClientOptions {
+    apiKey?: string;
+    baseURL?: string;
+    apiVersion?: string;
+}
+/**
+ * Detects if the given URL is an Azure OpenAI endpoint
+ * Safely parses the URL and checks if the hostname ends with ".openai.azure.com"
+ */
+export declare const isAzureEndpoint: (baseURL: string | undefined) => boolean;
+/**
+ * Creates an OpenAI or AzureOpenAI client based on the baseURL
+ * - If baseURL contains ".openai.azure.com", returns AzureOpenAI client
+ * - Otherwise, returns standard OpenAI client
+ */
+export declare const createOpenAIClient: (options: OpenAIClientOptions) => OpenAI;

package/lib/utils/openai_client.js

@@ -0,0 +1,35 @@
+import OpenAI, { AzureOpenAI } from "openai";
+/**
+ * Detects if the given URL is an Azure OpenAI endpoint
+ * Safely parses the URL and checks if the hostname ends with ".openai.azure.com"
+ */
+export const isAzureEndpoint = (baseURL) => {
+    if (!baseURL)
+        return false;
+    try {
+        const url = new URL(baseURL);
+        return url.hostname.endsWith(".openai.azure.com");
+    }
+    catch {
+        return false;
+    }
+};
+/**
+ * Creates an OpenAI or AzureOpenAI client based on the baseURL
+ * - If baseURL contains ".openai.azure.com", returns AzureOpenAI client
+ * - Otherwise, returns standard OpenAI client
+ */
+export const createOpenAIClient = (options) => {
+    const { apiKey, baseURL, apiVersion } = options;
+    if (isAzureEndpoint(baseURL)) {
+        return new AzureOpenAI({
+            apiKey,
+            endpoint: baseURL,
+            apiVersion: apiVersion ?? "2025-04-01-preview",
+        });
+    }
+    return new OpenAI({
+        apiKey,
+        baseURL,
+    });
+};

package/lib/utils/utils.js

@@ -53,6 +53,9 @@ export const settings2GraphAIConfig = (settings, env) => {
             if (info.baseURLKeyName) {
                 config[info.agentName].baseURL = getKey(prefix, info.baseURLKeyName);
             }
+            if (info.apiVersionKeyName) {
+                config[info.agentName].apiVersion = getKey(prefix, info.apiVersionKeyName);
+            }
         });
     };
     const config = {};

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "mulmocast",
-  "version": "2.1.21",
+  "version": "2.1.23",
   "description": "",
   "type": "module",
   "main": "lib/index.node.js",
@@ -83,7 +83,7 @@
     "@graphai/gemini_agent": "^2.0.4",
     "@graphai/groq_agent": "^2.0.2",
     "@graphai/input_agents": "^1.0.2",
-    "@graphai/openai_agent": "^2.0.8",
+    "@graphai/openai_agent": "^2.0.9",
     "@graphai/stream_agent_filter": "^2.0.2",
     "@graphai/vanilla": "^2.0.12",
     "@graphai/vanilla_node_agents": "^2.0.4",
@@ -101,7 +101,7 @@
     "marked": "^17.0.1",
     "mulmocast-vision": "^1.0.8",
     "ora": "^9.1.0",
-    "puppeteer": "^24.36.0",
+    "puppeteer": "^24.36.1",
     "replicate": "^1.4.0",
     "yaml": "^2.8.2",
     "yargs": "^18.0.0",
@@ -116,11 +116,11 @@
     "eslint": "^9.39.2",
     "eslint-config-prettier": "^10.1.8",
     "eslint-plugin-prettier": "^5.5.5",
-    "eslint-plugin-sonarjs": "^3.0.5",
+    "eslint-plugin-sonarjs": "^3.0.6",
     "prettier": "^3.8.1",
     "tsx": "^4.21.0",
     "typescript": "^5.9.3",
-    "typescript-eslint": "^8.53.1"
+    "typescript-eslint": "^8.54.0"
   },
   "engines": {
     "node": ">=20.0.0"

package/scripts/test/test_audio_azure.json

@@ -0,0 +1,24 @@
+{
+  "$mulmocast": {
+    "version": "1.1"
+  },
+  "lang": "en",
+  "title": "Test Audio with Azure OpenAI TTS",
+  "speechParams": {
+    "speakers": {
+      "Presenter": {
+        "provider": "openai",
+        "voiceId": "alloy",
+        "model": "tts"
+      }
+    }
+  },
+  "beats": [
+    {
+      "text": "Hello, this is a test of Azure OpenAI text to speech."
+    },
+    {
+      "text": "MulmoCast now supports Azure OpenAI for both image generation and speech synthesis."
+    }
+  ]
+}

package/scripts/test/test_images_azure.json

@@ -0,0 +1,26 @@
+{
+  "$mulmocast": {
+    "version": "1.1"
+  },
+  "lang": "en",
+  "title": "Test Images with Azure OpenAI",
+  "imageParams": {
+    "provider": "openai",
+    "model": "gpt-image-1.5"
+  },
+  "audioParams": {
+    "suppressSpeech": true
+  },
+  "beats": [
+    {
+      "text": "",
+      "imagePrompt": "A beautiful sunset over mountains, photorealistic style",
+      "duration": 3
+    },
+    {
+      "text": "",
+      "imagePrompt": "A cute robot waving hello, digital art style",
+      "duration": 3
+    }
+  ]
+}

package/lib/agents/tts_nijivoice_agent.d.ts

@@ -1,5 +0,0 @@
-import type { AgentFunction, AgentFunctionInfo } from "graphai";
-import type { NijivoiceTTSAgentParams, AgentBufferResult, AgentTextInputs, AgentErrorResult, AgentConfig } from "../types/agent.js";
-export declare const ttsNijivoiceAgent: AgentFunction<NijivoiceTTSAgentParams, AgentBufferResult | AgentErrorResult, AgentTextInputs, AgentConfig>;
-declare const ttsNijivoiceAgentInfo: AgentFunctionInfo;
-export default ttsNijivoiceAgentInfo;

package/lib/agents/tts_nijivoice_agent.js

@@ -1,76 +0,0 @@
-import { GraphAILogger } from "graphai";
-import { apiKeyMissingError, agentGenerationError, audioAction, audioFileTarget } from "../utils/error_cause.js";
-/*
-const errorMessage = [
-  "TTS NijiVoice: No API key. ",
-  "You have the following options:",
-  "1. Obtain an API key from Niji Voice (https://platform.nijivoice.com/) and set it as the NIJIVOICE_API_KEY environment variable.",
-  '2. Use OpenAI\'s TTS instead of Niji Voice by changing speechParams.provider from "nijivoice" to "openai".',
-].join("\n");
-*/
-export const ttsNijivoiceAgent = async ({ params, namedInputs, config, }) => {
-    const { suppressError, voice, speed, speed_global } = params;
-    const { apiKey } = config ?? {};
-    const { text } = namedInputs;
-    if (!apiKey) {
-        throw new Error("NijiVoice API key is required (NIJIVOICE_API_KEY)", {
-            cause: apiKeyMissingError("ttsNijivoiceAgent", audioAction, "NIJIVOICE_API_KEY"),
-        });
-    }
-    const url = `https://api.nijivoice.com/api/platform/v1/voice-actors/${voice}/generate-voice`;
-    const options = {
-        method: "POST",
-        headers: {
-            "x-api-key": apiKey,
-            accept: "application/json",
-            "content-type": "application/json",
-        },
-        body: JSON.stringify({
-            format: "mp3",
-            speed: String(speed ?? speed_global ?? "1.0"),
-            script: text,
-        }),
-    };
-    try {
-        const voiceRes = await fetch(url, options);
-        const voiceJson = await voiceRes.json();
-        if (voiceJson?.generatedVoice?.audioFileDownloadUrl) {
-            const audioRes = await fetch(voiceJson.generatedVoice.audioFileDownloadUrl);
-            const buffer = Buffer.from(await audioRes.arrayBuffer());
-            return { buffer };
-        }
-        if (suppressError) {
-            return {
-                error: voiceJson,
-            };
-        }
-        GraphAILogger.info(voiceJson);
-        throw new Error("TTS Nijivoice Error", {
-            cause: agentGenerationError("ttsNijivoiceAgent", audioAction, audioFileTarget),
-        });
-    }
-    catch (e) {
-        if (suppressError) {
-            return {
-                error: e,
-            };
-        }
-        GraphAILogger.info(e);
-        throw new Error("TTS Nijivoice Error", {
-            cause: agentGenerationError("ttsNijivoiceAgent", audioAction, audioFileTarget),
-        });
-    }
-};
-const ttsNijivoiceAgentInfo = {
-    name: "ttsNijivoiceAgent",
-    agent: ttsNijivoiceAgent,
-    mock: ttsNijivoiceAgent,
-    samples: [],
-    description: "TTS nijivoice agent",
-    category: ["tts"],
-    author: "Receptron Team",
-    repository: "https://github.com/receptron/mulmocast-cli/",
-    license: "MIT",
-    environmentVariables: ["NIJIVOICE_API_KEY"],
-};
-export default ttsNijivoiceAgentInfo;

package/lib/agents/utils.d.ts

@@ -1 +0,0 @@
-export {};

package/lib/agents/utils.js

@@ -1 +0,0 @@
-export {};

package/lib/utils/const.d.ts

@@ -1,15 +0,0 @@
-export declare const currentMulmoScriptVersion = "1.1";
-export declare const outDirName = "output";
-export declare const audioDirName = "audio";
-export declare const imageDirName = "images";
-export declare const cacheDirName = "cache";
-export declare const pdf_modes: string[];
-export declare const pdf_sizes: string[];
-export declare const languages: string[];
-export declare const storyToScriptGenerateMode: {
-    stepWise: string;
-    oneStep: string;
-};
-export declare const bundleTargetLang: string[];
-export declare const ASPECT_RATIOS: string[];
-export declare const PRO_ASPECT_RATIOS: string[];

package/lib/utils/const.js

@@ -1,15 +0,0 @@
-export const currentMulmoScriptVersion = "1.1";
-export const outDirName = "output";
-export const audioDirName = "audio";
-export const imageDirName = "images";
-export const cacheDirName = "cache";
-export const pdf_modes = ["slide", "talk", "handout"];
-export const pdf_sizes = ["letter", "a4"];
-export const languages = ["en", "ja", "fr", "es", "de", "zh-CN", "zh-TW", "ko", "it", "pt", "ar", "hi"];
-export const storyToScriptGenerateMode = {
-    stepWise: "step_wise",
-    oneStep: "one_step",
-};
-export const bundleTargetLang = ["ja", "en"];
-export const ASPECT_RATIOS = ["1:1", "9:16", "16:9"];
-export const PRO_ASPECT_RATIOS = ["1:1", "2:3", "3:2", "3:4", "4:3", "4:5", "5:4", "9:16", "16:9", "21:9"];

package/lib/utils/provider2agent.d.ts

@@ -1,191 +0,0 @@
-export declare const provider2TTSAgent: {
-    nijivoice: {
-        agentName: string;
-        hasLimitedConcurrency: boolean;
-        keyName: string;
-    };
-    openai: {
-        agentName: string;
-        hasLimitedConcurrency: boolean;
-        defaultModel: string;
-        defaultVoice: string;
-        keyName: string;
-        baseURLKeyName: string;
-    };
-    google: {
-        agentName: string;
-        hasLimitedConcurrency: boolean;
-        keyName: string;
-    };
-    gemini: {
-        agentName: string;
-        hasLimitedConcurrency: boolean;
-        defaultModel: string;
-        defaultVoice: string;
-        models: string[];
-        keyName: string;
-    };
-    elevenlabs: {
-        agentName: string;
-        hasLimitedConcurrency: boolean;
-        defaultModel: string;
-        models: string[];
-        keyName: string;
-    };
-    kotodama: {
-        agentName: string;
-        hasLimitedConcurrency: boolean;
-        defaultVoice: string;
-        defaultDecoration: string;
-        keyName: string;
-    };
-    mock: {
-        agentName: string;
-        hasLimitedConcurrency: boolean;
-        defaultModel: string;
-        models: string[];
-    };
-};
-export declare const gptImages: string[];
-export declare const provider2ImageAgent: {
-    openai: {
-        agentName: string;
-        defaultModel: string;
-        models: string[];
-        keyName: string;
-        baseURLKeyName: string;
-    };
-    google: {
-        agentName: string;
-        defaultModel: string;
-        models: string[];
-        keyName: string;
-    };
-    replicate: {
-        agentName: string;
-        defaultModel: string;
-        models: string[];
-        keyName: string;
-    };
-    mock: {
-        agentName: string;
-        defaultModel: string;
-        models: string[];
-        keyName: string;
-    };
-};
-export type ReplicateModel = `${string}/${string}`;
-export declare const provider2MovieAgent: {
-    replicate: {
-        agentName: string;
-        defaultModel: ReplicateModel;
-        keyName: string;
-        models: string[];
-        modelParams: Record<ReplicateModel, {
-            durations: number[];
-            start_image: string | undefined;
-            last_image?: string;
-            price_per_sec: number;
-        }>;
-    };
-    google: {
-        agentName: string;
-        defaultModel: string;
-        models: string[];
-        keyName: string;
-        modelParams: {
-            "veo-3.1-generate-preview": {
-                durations: number[];
-            };
-            "veo-3.0-generate-001": {
-                durations: number[];
-            };
-            "veo-2.0-generate-001": {
-                durations: number[];
-            };
-        };
-    };
-    mock: {
-        agentName: string;
-        defaultModel: string;
-        models: string[];
-        keyName: string;
-        modelParams: {};
-    };
-};
-export declare const provider2SoundEffectAgent: {
-    replicate: {
-        agentName: string;
-        defaultModel: ReplicateModel;
-        keyName: string;
-        models: ReplicateModel[];
-        modelParams: Record<ReplicateModel, {
-            identifier?: `${string}/${string}:${string}`;
-        }>;
-    };
-};
-export declare const provider2LipSyncAgent: {
-    replicate: {
-        agentName: string;
-        defaultModel: ReplicateModel;
-        keyName: string;
-        models: ReplicateModel[];
-        modelParams: Record<ReplicateModel, {
-            identifier?: `${string}/${string}:${string}` | `${string}/${string}`;
-            video?: string;
-            audio: string;
-            image?: string;
-        }>;
-    };
-};
-export declare const provider2LLMAgent: {
-    readonly openai: {
-        readonly agentName: "openAIAgent";
-        readonly defaultModel: "gpt-5";
-        readonly keyName: "OPENAI_API_KEY";
-        readonly baseURLKeyName: "OPENAI_BASE_URL";
-        readonly max_tokens: 8192;
-        readonly models: readonly ["gpt-5", "gpt-5-nano", "gpt-5-mini", "gpt-4.1", "gpt-4.1-mini", "gpt-4.1-nano", "o3", "o3-mini", "o3-pro", "o1", "o1-pro", "gpt-4o", "gpt-4o-mini"];
-    };
-    readonly anthropic: {
-        readonly agentName: "anthropicAgent";
-        readonly defaultModel: "claude-3-7-sonnet-20250219";
-        readonly max_tokens: 8192;
-        readonly models: readonly ["claude-opus-4-1-20250805", "claude-opus-4-20250514", "claude-sonnet-4-20250514", "claude-3-7-sonnet-20250219", "claude-3-haiku-20240307"];
-        readonly keyName: "ANTHROPIC_API_KEY";
-        readonly apiKeyNameOverride: "ANTHROPIC_API_TOKEN";
-    };
-    readonly gemini: {
-        readonly agentName: "geminiAgent";
-        readonly defaultModel: "gemini-2.5-flash";
-        readonly max_tokens: 8192;
-        readonly models: readonly ["gemini-2.5-pro", "gemini-2.5-flash", "gemini-2.5-flash-lite", "gemini-2.0-flash"];
-        readonly keyName: "GEMINI_API_KEY";
-    };
-    readonly groq: {
-        readonly agentName: "groqAgent";
-        readonly defaultModel: "llama-3.1-8b-instant";
-        readonly keyName: "GROQ_API_KEY";
-        readonly max_tokens: 4096;
-        readonly models: readonly ["llama-3.1-8b-instant", "llama-3.3-70b-versatile", "deepseek-r1-distill-llama-70b", "openai/gpt-oss-120b", "openai/gpt-oss-20b"];
-    };
-    readonly mock: {
-        readonly agentName: "mediaMockAgent";
-        readonly defaultModel: "mock";
-        readonly max_tokens: 4096;
-        readonly models: readonly ["mock"];
-    };
-};
-export declare const defaultProviders: {
-    tts: keyof typeof provider2TTSAgent;
-    text2image: keyof typeof provider2ImageAgent;
-    text2movie: keyof typeof provider2MovieAgent;
-    text2Html: keyof typeof provider2LLMAgent;
-    llm: keyof typeof provider2LLMAgent;
-    soundEffect: keyof typeof provider2SoundEffectAgent;
-    lipSync: keyof typeof provider2LipSyncAgent;
-};
-export declare const llm: (keyof typeof provider2LLMAgent)[];
-export type LLM = keyof typeof provider2LLMAgent;
-export declare const htmlLLMProvider: string[];
-export declare const getModelDuration: (provider: keyof typeof provider2MovieAgent, model: string, movieDuration?: number) => number | undefined;

package/lib/utils/provider2agent.js

@@ -1,326 +0,0 @@
-// node & browser
-export const provider2TTSAgent = {
-    nijivoice: {
-        agentName: "ttsNijivoiceAgent",
-        hasLimitedConcurrency: true,
-        keyName: "NIJIVOICE_API_KEY",
-    },
-    openai: {
-        agentName: "ttsOpenaiAgent",
-        hasLimitedConcurrency: false,
-        defaultModel: "gpt-4o-mini-tts",
-        defaultVoice: "shimmer",
-        keyName: "OPENAI_API_KEY",
-        baseURLKeyName: "OPENAI_BASE_URL",
-    },
-    google: {
-        agentName: "ttsGoogleAgent",
-        hasLimitedConcurrency: false,
-        keyName: "GEMINI_API_KEY",
-    },
-    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: {
-        agentName: "ttsElevenlabsAgent",
-        hasLimitedConcurrency: true,
-        defaultModel: "eleven_multilingual_v2",
-        // Models | ElevenLabs Documentation
-        // https://elevenlabs.io/docs/models
-        models: ["eleven_multilingual_v2", "eleven_turbo_v2_5", "eleven_turbo_v2", "eleven_flash_v2_5", "eleven_flash_v2"],
-        keyName: "ELEVENLABS_API_KEY",
-    },
-    kotodama: {
-        agentName: "ttsKotodamaAgent",
-        hasLimitedConcurrency: true,
-        defaultVoice: "Atla",
-        defaultDecoration: "neutral",
-        keyName: "KOTODAMA_API_KEY",
-    },
-    mock: {
-        agentName: "mediaMockAgent",
-        hasLimitedConcurrency: true,
-        defaultModel: "mock-model",
-        models: ["mock-model"],
-    },
-};
-export const gptImages = ["gpt-image-1.5", "gpt-image-1", "gpt-image-1-mini"];
-export const provider2ImageAgent = {
-    openai: {
-        agentName: "imageOpenaiAgent",
-        defaultModel: "gpt-image-1",
-        models: ["dall-e-3", ...gptImages],
-        keyName: "OPENAI_API_KEY",
-        baseURLKeyName: "OPENAI_BASE_URL",
-    },
-    google: {
-        agentName: "imageGenAIAgent",
-        defaultModel: "gemini-2.5-flash-image",
-        models: ["imagen-4.0-generate-preview-06-06", "imagen-4.0-ultra-generate-preview-06-06", "gemini-2.5-flash-image", "gemini-3-pro-image-preview"],
-        keyName: "GEMINI_API_KEY",
-    },
-    replicate: {
-        agentName: "imageReplicateAgent",
-        defaultModel: "bytedance/seedream-4",
-        models: ["bytedance/seedream-4", "qwen/qwen-image"],
-        keyName: "REPLICATE_API_TOKEN",
-    },
-    mock: {
-        agentName: "mediaMockAgent",
-        defaultModel: "mock-model",
-        models: ["mock-model"],
-        keyName: "",
-    },
-};
-export const provider2MovieAgent = {
-    replicate: {
-        agentName: "movieReplicateAgent",
-        defaultModel: "bytedance/seedance-1-lite",
-        keyName: "REPLICATE_API_TOKEN",
-        models: [
-            "bytedance/seedance-1-lite",
-            "bytedance/seedance-1-pro",
-            "kwaivgi/kling-v1.6-pro",
-            "kwaivgi/kling-v2.1",
-            "kwaivgi/kling-v2.1-master",
-            "google/veo-2",
-            "google/veo-3",
-            "google/veo-3-fast",
-            "minimax/video-01",
-            "minimax/hailuo-02",
-            "minimax/hailuo-02-fast",
-            "pixverse/pixverse-v4.5",
-            "wan-video/wan-2.2-i2v-fast",
-            "wan-video/wan-2.2-t2v-fast",
-        ],
-        modelParams: {
-            "bytedance/seedance-1-lite": {
-                durations: [5, 10],
-                start_image: "image",
-                last_image: "last_frame_image",
-                price_per_sec: 0.036, // in USD
-            },
-            "bytedance/seedance-1-pro": {
-                durations: [5, 10],
-                start_image: "image",
-                last_image: "last_frame_image",
-                price_per_sec: 0.15,
-            },
-            "kwaivgi/kling-v1.6-pro": {
-                durations: [5, 10],
-                start_image: "start_image",
-                price_per_sec: 0.095,
-            },
-            "kwaivgi/kling-v2.1": {
-                durations: [5, 10],
-                start_image: "start_image",
-                price_per_sec: 0.05,
-            },
-            "kwaivgi/kling-v2.1-master": {
-                durations: [5, 10],
-                start_image: "start_image",
-                price_per_sec: 0.28,
-            },
-            "google/veo-2": {
-                durations: [5, 6, 7, 8],
-                start_image: "image",
-                price_per_sec: 0.5,
-            },
-            "google/veo-3": {
-                durations: [8],
-                start_image: "image",
-                price_per_sec: 0.75,
-            },
-            "google/veo-3-fast": {
-                durations: [8],
-                start_image: "image",
-                price_per_sec: 0.4,
-            },
-            "minimax/video-01": {
-                durations: [6],
-                start_image: "first_frame_image",
-                price_per_sec: 0.5,
-            },
-            "minimax/hailuo-02": {
-                durations: [6], // NOTE: 10 for only 720p
-                start_image: "first_frame_image",
-                price_per_sec: 0.08,
-            },
-            "minimax/hailuo-02-fast": {
-                durations: [6, 10], // NOTE: 512P
-                start_image: "first_frame_image",
-                price_per_sec: 0.0166,
-            },
-            "pixverse/pixverse-v4.5": {
-                durations: [5, 8],
-                start_image: "image",
-                last_image: "last_frame_image",
-                price_per_sec: 0.12,
-            },
-            "wan-video/wan-2.2-i2v-fast": {
-                durations: [5],
-                start_image: "image",
-                price_per_sec: 0.012,
-            },
-            "wan-video/wan-2.2-t2v-fast": {
-                durations: [5],
-                start_image: undefined,
-                price_per_sec: 0.012,
-            },
-        },
-    },
-    google: {
-        agentName: "movieGenAIAgent",
-        defaultModel: "veo-2.0-generate-001",
-        models: ["veo-2.0-generate-001", "veo-3.0-generate-001", "veo-3.1-generate-preview"],
-        keyName: "GEMINI_API_KEY",
-        modelParams: {
-            "veo-3.1-generate-preview": {
-                durations: [4, 6, 8],
-            },
-            "veo-3.0-generate-001": {
-                durations: [4, 6, 8],
-            },
-            "veo-2.0-generate-001": {
-                durations: [5, 6, 7, 8],
-            },
-        },
-    },
-    mock: {
-        agentName: "mediaMockAgent",
-        defaultModel: "mock-model",
-        models: ["mock-model"],
-        keyName: "",
-        modelParams: {},
-    },
-};
-export const provider2SoundEffectAgent = {
-    replicate: {
-        agentName: "soundEffectReplicateAgent",
-        defaultModel: "zsxkib/mmaudio",
-        keyName: "REPLICATE_API_TOKEN",
-        models: ["zsxkib/mmaudio"],
-        modelParams: {
-            "zsxkib/mmaudio": {
-                identifier: "zsxkib/mmaudio:62871fb59889b2d7c13777f08deb3b36bdff88f7e1d53a50ad7694548a41b484",
-            },
-        },
-    },
-};
-export const provider2LipSyncAgent = {
-    replicate: {
-        agentName: "lipSyncReplicateAgent",
-        defaultModel: "bytedance/omni-human",
-        keyName: "REPLICATE_API_TOKEN",
-        models: ["bytedance/latentsync", "tmappdev/lipsync", "bytedance/omni-human"],
-        modelParams: {
-            "bytedance/latentsync": {
-                identifier: "bytedance/latentsync:637ce1919f807ca20da3a448ddc2743535d2853649574cd52a933120e9b9e293",
-                video: "video",
-                audio: "audio",
-            },
-            "tmappdev/lipsync": {
-                identifier: "tmappdev/lipsync:c54ce2fe673ea59b857b91250b3d71a2cd304a78f2370687632805c8405fbf4c",
-                video: "video_input",
-                audio: "audio_input",
-            },
-            "bytedance/omni-human": {
-                identifier: "bytedance/omni-human",
-                image: "image",
-                audio: "audio",
-                price_per_sec: 0.14,
-            },
-            /* NOTE: This model does not work with large base64 urls.
-            "sync/lipsync-2": {
-              video: "video",
-              audio: "audio",
-            },
-            */
-            /* NOTE: This model does not work well for some unknown reason.
-            "kwaivgi/kling-lip-sync": {
-              video: "video_url",
-              audio: "audio_file",
-            },
-            */
-        },
-    },
-};
-// : Record<LLM, { agent: string; defaultModel: string; max_tokens: number }>
-export const provider2LLMAgent = {
-    openai: {
-        agentName: "openAIAgent",
-        defaultModel: "gpt-5",
-        keyName: "OPENAI_API_KEY",
-        baseURLKeyName: "OPENAI_BASE_URL",
-        max_tokens: 8192,
-        models: [
-            "gpt-5",
-            "gpt-5-nano",
-            "gpt-5-mini",
-            "gpt-4.1",
-            "gpt-4.1-mini",
-            "gpt-4.1-nano",
-            "o3",
-            "o3-mini",
-            "o3-pro",
-            "o1",
-            "o1-pro",
-            "gpt-4o",
-            "gpt-4o-mini",
-        ],
-    },
-    anthropic: {
-        agentName: "anthropicAgent",
-        defaultModel: "claude-3-7-sonnet-20250219",
-        max_tokens: 8192,
-        models: ["claude-opus-4-1-20250805", "claude-opus-4-20250514", "claude-sonnet-4-20250514", "claude-3-7-sonnet-20250219", "claude-3-haiku-20240307"],
-        keyName: "ANTHROPIC_API_KEY",
-        apiKeyNameOverride: "ANTHROPIC_API_TOKEN",
-        // GraphAI is currently using ANTHROPIC_API_KEY, but the official name is ANTHROPIC_API_TOKEN.
-    },
-    gemini: {
-        agentName: "geminiAgent",
-        defaultModel: "gemini-2.5-flash",
-        max_tokens: 8192,
-        models: ["gemini-2.5-pro", "gemini-2.5-flash", "gemini-2.5-flash-lite", "gemini-2.0-flash"],
-        keyName: "GEMINI_API_KEY",
-    },
-    groq: {
-        agentName: "groqAgent",
-        defaultModel: "llama-3.1-8b-instant",
-        keyName: "GROQ_API_KEY",
-        max_tokens: 4096,
-        models: ["llama-3.1-8b-instant", "llama-3.3-70b-versatile", "deepseek-r1-distill-llama-70b", "openai/gpt-oss-120b", "openai/gpt-oss-20b"],
-    },
-    mock: {
-        agentName: "mediaMockAgent",
-        defaultModel: "mock",
-        max_tokens: 4096,
-        models: ["mock"],
-    },
-};
-export const defaultProviders = {
-    tts: "openai",
-    text2image: "openai",
-    text2movie: "replicate",
-    text2Html: "openai",
-    llm: "openai",
-    soundEffect: "replicate",
-    lipSync: "replicate",
-};
-export const llm = Object.keys(provider2LLMAgent);
-export const htmlLLMProvider = ["openai", "anthropic", "mock"];
-export const getModelDuration = (provider, model, movieDuration) => {
-    const modelParams = provider2MovieAgent[provider]?.modelParams;
-    const { durations } = modelParams[model];
-    if (durations && movieDuration) {
-        const largerDurations = durations.filter((d) => d >= movieDuration);
-        return largerDurations.length > 0 ? largerDurations[0] : durations[durations.length - 1];
-    }
-    return durations?.[0];
-};