mulmocast 2.1.22 → 2.1.24

package/lib/agents/image_genai_agent.js

@@ -59,12 +59,20 @@ export const imageGenAIAgent = async ({ namedInputs, params, config, }) => {
     const { prompt, referenceImages } = namedInputs;
     const model = params.model ?? provider2ImageAgent["google"].defaultModel;
     const apiKey = config?.apiKey;
-    if (!apiKey) {
-        throw new Error("Google GenAI API key is required (GEMINI_API_KEY)", {
-            cause: apiKeyMissingError("imageGenAIAgent", imageAction, "GEMINI_API_KEY"),
-        });
-    }
-    const ai = new GoogleGenAI({ apiKey });
+    const ai = params.vertexai_project
+        ? new GoogleGenAI({
+            vertexai: true,
+            project: params.vertexai_project,
+            location: params.vertexai_location ?? "us-central1",
+        })
+        : (() => {
+            if (!apiKey) {
+                throw new Error("Google GenAI API key is required (GEMINI_API_KEY)", {
+                    cause: apiKeyMissingError("imageGenAIAgent", imageAction, "GEMINI_API_KEY"),
+                });
+            }
+            return new GoogleGenAI({ apiKey });
+        })();
     if (model === "gemini-2.5-flash-image" || model === "gemini-3-pro-image-preview") {
         const contentParams = (() => {
             const contents = getGeminiContents(prompt, referenceImages);

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/movie_genai_agent.js

@@ -1,4 +1,4 @@
-import { readFileSync } from "fs";
+import { readFileSync, writeFileSync } from "fs";
 import { GraphAILogger, sleep } from "graphai";
 import { GoogleGenAI, PersonGeneration } from "@google/genai";
 import { apiKeyMissingError, agentGenerationError, agentInvalidResponseError, imageAction, movieFileTarget, videoDurationTarget, hasCause, } from "../utils/error_cause.js";
@@ -35,12 +35,19 @@ const loadImageAsBase64 = (imagePath) => {
         mimeType: "image/png",
     };
 };
-const downloadVideo = async (ai, video, movieFile) => {
-    await ai.files.download({
-        file: video,
-        downloadPath: movieFile,
-    });
-    await sleep(5000); // HACK: Without this, the file is not ready yet.
+const downloadVideo = async (ai, video, movieFile, isVertexAI) => {
+    if (isVertexAI) {
+        // Vertex AI returns videoBytes directly
+        writeFileSync(movieFile, Buffer.from(video.videoBytes, "base64"));
+    }
+    else {
+        // Gemini API requires download via uri
+        await ai.files.download({
+            file: video,
+            downloadPath: movieFile,
+        });
+        await sleep(5000); // HACK: Without this, the file is not ready yet.
+    }
     return { saved: movieFile };
 };
 const createVeo31Payload = (model, prompt, aspectRatio, source) => ({
@@ -53,7 +60,7 @@ const createVeo31Payload = (model, prompt, aspectRatio, source) => ({
     },
     ...source,
 });
-const generateExtendedVideo = async (ai, model, prompt, aspectRatio, imagePath, requestedDuration, movieFile) => {
+const generateExtendedVideo = async (ai, model, prompt, aspectRatio, imagePath, requestedDuration, movieFile, isVertexAI) => {
     const initialDuration = 8;
     const maxExtensionDuration = 8;
     const extensionsNeeded = Math.ceil((requestedDuration - initialDuration) / maxExtensionDuration);
@@ -85,9 +92,9 @@ const generateExtendedVideo = async (ai, model, prompt, aspectRatio, imagePath,
             cause: agentInvalidResponseError("movieGenAIAgent", imageAction, movieFileTarget),
         });
     }
-    return downloadVideo(ai, result.video, movieFile);
+    return downloadVideo(ai, result.video, movieFile, isVertexAI);
 };
-const generateStandardVideo = async (ai, model, prompt, aspectRatio, imagePath, duration, movieFile) => {
+const generateStandardVideo = async (ai, model, prompt, aspectRatio, imagePath, duration, movieFile, isVertexAI) => {
     const isVeo3 = model === "veo-3.0-generate-001" || model === "veo-3.1-generate-preview";
     const payload = {
         model,
@@ -102,18 +109,13 @@ const generateStandardVideo = async (ai, model, prompt, aspectRatio, imagePath,
     const operation = await ai.models.generateVideos(payload);
     const response = await pollUntilDone(ai, operation);
     const video = getVideoFromResponse(response);
-    return downloadVideo(ai, video, movieFile);
+    return downloadVideo(ai, video, movieFile, isVertexAI);
 };
 export const movieGenAIAgent = async ({ namedInputs, params, config, }) => {
     const { prompt, imagePath, movieFile } = namedInputs;
     const aspectRatio = getAspectRatio(params.canvasSize, ASPECT_RATIOS);
     const model = params.model ?? provider2MovieAgent.google.defaultModel;
     const apiKey = config?.apiKey;
-    if (!apiKey) {
-        throw new Error("Google GenAI API key is required (GEMINI_API_KEY)", {
-            cause: apiKeyMissingError("movieGenAIAgent", imageAction, "GEMINI_API_KEY"),
-        });
-    }
     try {
         const requestedDuration = params.duration ?? 8;
         const duration = getModelDuration("google", model, requestedDuration);
@@ -122,13 +124,27 @@ export const movieGenAIAgent = async ({ namedInputs, params, config, }) => {
                 cause: agentGenerationError("movieGenAIAgent", imageAction, videoDurationTarget),
             });
         }
-        const ai = new GoogleGenAI({ apiKey });
+        const isVertexAI = !!params.vertexai_project;
+        const ai = isVertexAI
+            ? new GoogleGenAI({
+                vertexai: true,
+                project: params.vertexai_project,
+                location: params.vertexai_location ?? "us-central1",
+            })
+            : (() => {
+                if (!apiKey) {
+                    throw new Error("Google GenAI API key is required (GEMINI_API_KEY)", {
+                        cause: apiKeyMissingError("movieGenAIAgent", imageAction, "GEMINI_API_KEY"),
+                    });
+                }
+                return new GoogleGenAI({ apiKey });
+            })();
         // Veo 3.1: Video extension mode for videos longer than 8s
         if (model === "veo-3.1-generate-preview" && requestedDuration > 8 && params.canvasSize) {
-            return generateExtendedVideo(ai, model, prompt, aspectRatio, imagePath, requestedDuration, movieFile);
+            return generateExtendedVideo(ai, model, prompt, aspectRatio, imagePath, requestedDuration, movieFile, isVertexAI);
         }
         // Standard mode
-        return generateStandardVideo(ai, model, prompt, aspectRatio, imagePath, duration, movieFile);
+        return generateStandardVideo(ai, model, prompt, aspectRatio, imagePath, duration, movieFile, isVertexAI);
     }
     catch (error) {
         GraphAILogger.info("Failed to generate movie:", error.message);

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/methods/mulmo_presentation_style.d.ts

@@ -171,6 +171,8 @@ export declare const MulmoPresentationStyleMethods: {
                 type: "custom";
                 filter: string;
             })[] | undefined;
+            vertexai_project?: string | undefined;
+            vertexai_location?: string | undefined;
             speed?: number | undefined;
         };
         keyName: string;

package/lib/types/agent.d.ts

@@ -39,6 +39,8 @@ export type ImageAgentParams = {
         width: number;
         height: number;
     };
+    vertexai_project?: string;
+    vertexai_location?: string;
 };
 export type OpenAIImageAgentParams = ImageAgentParams & {
     moderation: OpenAIImageModeration | null | undefined;
@@ -58,6 +60,7 @@ export type ReplicateImageAgentParams = {
 export type OpenAIImageAgentConfig = {
     baseURL?: string;
     apiKey?: string;
+    apiVersion?: string;
 };
 export type GoogleImageAgentConfig = {
     projectId?: string;
@@ -73,6 +76,8 @@ export type MovieAgentInputs = AgentPromptInputs & {
 };
 export type GoogleMovieAgentParams = ImageAgentParams & {
     duration?: number;
+    vertexai_project?: string;
+    vertexai_location?: string;
 };
 export type ReplicateMovieAgentParams = {
     model: `${string}/${string}` | undefined;

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"];
     };
@@ -154,7 +155,7 @@ export declare const provider2LLMAgent: {
         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"];
+        readonly models: readonly ["gemini-3-pro-preview", "gemini-3-flash-preview", "gemini-2.5-pro", "gemini-2.5-flash", "gemini-2.5-flash-lite"];
         readonly keyName: "GEMINI_API_KEY";
     };
     readonly groq: {

package/lib/types/provider2agent.js

@@ -56,7 +56,13 @@ export const provider2ImageAgent = {
     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"],
+        models: [
+            "imagen-4.0-generate-001",
+            "imagen-4.0-ultra-generate-001",
+            "imagen-4.0-fast-generate-001",
+            "gemini-2.5-flash-image",
+            "gemini-3-pro-image-preview",
+        ],
         keyName: "GEMINI_API_KEY",
     },
     replicate: {
@@ -252,6 +258,7 @@ export const provider2LLMAgent = {
         defaultModel: "gpt-5",
         keyName: "OPENAI_API_KEY",
         baseURLKeyName: "OPENAI_BASE_URL",
+        apiVersionKeyName: "OPENAI_API_VERSION",
         max_tokens: 8192,
         models: [
             "gpt-5",
@@ -282,7 +289,7 @@ export const provider2LLMAgent = {
         agentName: "geminiAgent",
         defaultModel: "gemini-2.5-flash",
         max_tokens: 8192,
-        models: ["gemini-2.5-pro", "gemini-2.5-flash", "gemini-2.5-flash-lite"],
+        models: ["gemini-3-pro-preview", "gemini-3-flash-preview", "gemini-2.5-pro", "gemini-2.5-flash", "gemini-2.5-flash-lite"],
         keyName: "GEMINI_API_KEY",
     },
     groq: {

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 +451,10 @@ 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>;
+    vertexai_project: z.ZodOptional<z.ZodString>;
+    vertexai_location: z.ZodOptional<z.ZodString>;
 }, z.core.$strict>;
 export declare const mulmoImageParamsSchema: z.ZodObject<{
     provider: z.ZodOptional<z.ZodEnum<{
@@ -446,6 +464,10 @@ export declare const mulmoImageParamsSchema: z.ZodObject<{
     quality: z.ZodOptional<z.ZodString>;
     style: z.ZodOptional<z.ZodString>;
     moderation: z.ZodOptional<z.ZodString>;
+    baseURL: z.ZodOptional<z.ZodString>;
+    apiVersion: z.ZodOptional<z.ZodString>;
+    vertexai_project: z.ZodOptional<z.ZodString>;
+    vertexai_location: z.ZodOptional<z.ZodString>;
     images: z.ZodOptional<z.ZodRecord<z.ZodString, z.ZodUnion<readonly [z.ZodObject<{
         type: z.ZodLiteral<"image">;
         source: z.ZodDiscriminatedUnion<[z.ZodObject<{
@@ -763,6 +785,8 @@ export declare const mulmoMovieParamsSchema: z.ZodObject<{
         type: z.ZodLiteral<"custom">;
         filter: z.ZodString;
     }, z.core.$strip>]>>>;
+    vertexai_project: z.ZodOptional<z.ZodString>;
+    vertexai_location: z.ZodOptional<z.ZodString>;
 }, z.core.$strip>;
 export declare const mulmoBeatSchema: z.ZodObject<{
     speaker: z.ZodOptional<z.ZodString>;
@@ -891,6 +915,10 @@ export declare const mulmoBeatSchema: z.ZodObject<{
         quality: z.ZodOptional<z.ZodString>;
         style: z.ZodOptional<z.ZodString>;
         moderation: z.ZodOptional<z.ZodString>;
+        baseURL: z.ZodOptional<z.ZodString>;
+        apiVersion: z.ZodOptional<z.ZodString>;
+        vertexai_project: z.ZodOptional<z.ZodString>;
+        vertexai_location: z.ZodOptional<z.ZodString>;
     }, z.core.$strict>>;
     audioParams: z.ZodOptional<z.ZodObject<{
         padding: z.ZodOptional<z.ZodNumber>;
@@ -1123,6 +1151,8 @@ export declare const mulmoBeatSchema: z.ZodObject<{
             type: z.ZodLiteral<"custom">;
             filter: z.ZodString;
         }, z.core.$strip>]>>>;
+        vertexai_project: z.ZodOptional<z.ZodString>;
+        vertexai_location: z.ZodOptional<z.ZodString>;
         speed: z.ZodOptional<z.ZodNumber>;
     }, z.core.$strip>>;
     soundEffectParams: z.ZodOptional<z.ZodObject<{
@@ -1223,6 +1253,8 @@ 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 +1270,8 @@ 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.$strip>>;
@@ -1249,6 +1283,10 @@ export declare const mulmoPresentationStyleSchema: z.ZodObject<{
         quality: z.ZodOptional<z.ZodString>;
         style: z.ZodOptional<z.ZodString>;
         moderation: z.ZodOptional<z.ZodString>;
+        baseURL: z.ZodOptional<z.ZodString>;
+        apiVersion: z.ZodOptional<z.ZodString>;
+        vertexai_project: z.ZodOptional<z.ZodString>;
+        vertexai_location: z.ZodOptional<z.ZodString>;
         images: z.ZodOptional<z.ZodRecord<z.ZodString, z.ZodUnion<readonly [z.ZodObject<{
             type: z.ZodLiteral<"image">;
             source: z.ZodDiscriminatedUnion<[z.ZodObject<{
@@ -1493,6 +1531,8 @@ export declare const mulmoPresentationStyleSchema: z.ZodObject<{
             type: z.ZodLiteral<"custom">;
             filter: z.ZodString;
         }, z.core.$strip>]>>>;
+        vertexai_project: z.ZodOptional<z.ZodString>;
+        vertexai_location: z.ZodOptional<z.ZodString>;
     }, z.core.$strip>>>;
     soundEffectParams: z.ZodDefault<z.ZodOptional<z.ZodObject<{
         provider: z.ZodOptional<z.ZodDefault<z.ZodEnum<{
@@ -1584,6 +1624,8 @@ export declare const mulmoScriptSchema: z.ZodObject<{
                 [x: string]: string;
             }>>>;
             model: z.ZodOptional<z.ZodString>;
+            baseURL: z.ZodOptional<z.ZodString>;
+            apiVersion: z.ZodOptional<z.ZodString>;
             lang: z.ZodOptional<z.ZodRecord<z.ZodString, z.ZodObject<{
                 displayName: z.ZodOptional<z.ZodRecord<z.ZodString, z.ZodString>>;
                 voiceId: z.ZodString;
@@ -1599,6 +1641,8 @@ export declare const mulmoScriptSchema: z.ZodObject<{
                     [x: string]: string;
                 }>>>;
                 model: z.ZodOptional<z.ZodString>;
+                baseURL: z.ZodOptional<z.ZodString>;
+                apiVersion: z.ZodOptional<z.ZodString>;
             }, z.core.$strict>>>;
         }, z.core.$strict>>;
     }, z.core.$strip>>;
@@ -1610,6 +1654,10 @@ 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>;
+        vertexai_project: z.ZodOptional<z.ZodString>;
+        vertexai_location: 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<{
@@ -1854,6 +1902,8 @@ export declare const mulmoScriptSchema: z.ZodObject<{
             type: z.ZodLiteral<"custom">;
             filter: z.ZodString;
         }, z.core.$strip>]>>>;
+        vertexai_project: z.ZodOptional<z.ZodString>;
+        vertexai_location: z.ZodOptional<z.ZodString>;
     }, z.core.$strip>>>;
     soundEffectParams: z.ZodDefault<z.ZodOptional<z.ZodObject<{
         provider: z.ZodOptional<z.ZodDefault<z.ZodEnum<{
@@ -2049,6 +2099,10 @@ 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>;
+            vertexai_project: z.ZodOptional<z.ZodString>;
+            vertexai_location: z.ZodOptional<z.ZodString>;
         }, z.core.$strict>>;
         audioParams: z.ZodOptional<z.ZodObject<{
             padding: z.ZodOptional<z.ZodNumber>;
@@ -2281,6 +2335,8 @@ export declare const mulmoScriptSchema: z.ZodObject<{
                 type: z.ZodLiteral<"custom">;
                 filter: z.ZodString;
             }, z.core.$strip>]>>>;
+            vertexai_project: z.ZodOptional<z.ZodString>;
+            vertexai_location: z.ZodOptional<z.ZodString>;
             speed: z.ZodOptional<z.ZodNumber>;
         }, z.core.$strip>>;
         soundEffectParams: z.ZodOptional<z.ZodObject<{
@@ -2456,6 +2512,8 @@ export declare const mulmoStudioSchema: z.ZodObject<{
                     [x: string]: string;
                 }>>>;
                 model: z.ZodOptional<z.ZodString>;
+                baseURL: z.ZodOptional<z.ZodString>;
+                apiVersion: z.ZodOptional<z.ZodString>;
                 lang: z.ZodOptional<z.ZodRecord<z.ZodString, z.ZodObject<{
                     displayName: z.ZodOptional<z.ZodRecord<z.ZodString, z.ZodString>>;
                     voiceId: z.ZodString;
@@ -2471,6 +2529,8 @@ export declare const mulmoStudioSchema: z.ZodObject<{
                         [x: string]: string;
                     }>>>;
                     model: z.ZodOptional<z.ZodString>;
+                    baseURL: z.ZodOptional<z.ZodString>;
+                    apiVersion: z.ZodOptional<z.ZodString>;
                 }, z.core.$strict>>>;
             }, z.core.$strict>>;
         }, z.core.$strip>>;
@@ -2482,6 +2542,10 @@ 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>;
+            vertexai_project: z.ZodOptional<z.ZodString>;
+            vertexai_location: 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<{
@@ -2726,6 +2790,8 @@ export declare const mulmoStudioSchema: z.ZodObject<{
                 type: z.ZodLiteral<"custom">;
                 filter: z.ZodString;
             }, z.core.$strip>]>>>;
+            vertexai_project: z.ZodOptional<z.ZodString>;
+            vertexai_location: z.ZodOptional<z.ZodString>;
         }, z.core.$strip>>>;
         soundEffectParams: z.ZodDefault<z.ZodOptional<z.ZodObject<{
             provider: z.ZodOptional<z.ZodDefault<z.ZodEnum<{
@@ -2921,6 +2987,10 @@ 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>;
+                vertexai_project: z.ZodOptional<z.ZodString>;
+                vertexai_location: z.ZodOptional<z.ZodString>;
             }, z.core.$strict>>;
             audioParams: z.ZodOptional<z.ZodObject<{
                 padding: z.ZodOptional<z.ZodNumber>;
@@ -3153,6 +3223,8 @@ export declare const mulmoStudioSchema: z.ZodObject<{
                     type: z.ZodLiteral<"custom">;
                     filter: z.ZodString;
                 }, z.core.$strip>]>>>;
+                vertexai_project: z.ZodOptional<z.ZodString>;
+                vertexai_location: z.ZodOptional<z.ZodString>;
                 speed: z.ZodOptional<z.ZodNumber>;
             }, z.core.$strip>>;
             soundEffectParams: z.ZodOptional<z.ZodObject<{
@@ -3264,6 +3336,8 @@ export declare const mulmoPromptTemplateSchema: z.ZodObject<{
                     [x: string]: string;
                 }>>>;
                 model: z.ZodOptional<z.ZodString>;
+                baseURL: z.ZodOptional<z.ZodString>;
+                apiVersion: z.ZodOptional<z.ZodString>;
                 lang: z.ZodOptional<z.ZodRecord<z.ZodString, z.ZodObject<{
                     displayName: z.ZodOptional<z.ZodRecord<z.ZodString, z.ZodString>>;
                     voiceId: z.ZodString;
@@ -3279,6 +3353,8 @@ export declare const mulmoPromptTemplateSchema: z.ZodObject<{
                         [x: string]: string;
                     }>>>;
                     model: z.ZodOptional<z.ZodString>;
+                    baseURL: z.ZodOptional<z.ZodString>;
+                    apiVersion: z.ZodOptional<z.ZodString>;
                 }, z.core.$strict>>>;
             }, z.core.$strict>>;
         }, z.core.$strip>>;
@@ -3290,6 +3366,10 @@ 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>;
+            vertexai_project: z.ZodOptional<z.ZodString>;
+            vertexai_location: 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<{
@@ -3534,6 +3614,8 @@ export declare const mulmoPromptTemplateSchema: z.ZodObject<{
                 type: z.ZodLiteral<"custom">;
                 filter: z.ZodString;
             }, z.core.$strip>]>>>;
+            vertexai_project: z.ZodOptional<z.ZodString>;
+            vertexai_location: z.ZodOptional<z.ZodString>;
         }, z.core.$strip>>>;
         soundEffectParams: z.ZodDefault<z.ZodOptional<z.ZodObject<{
             provider: z.ZodOptional<z.ZodDefault<z.ZodEnum<{
@@ -3619,6 +3701,8 @@ 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 +3718,8 @@ 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.$strip>>;
@@ -3645,6 +3731,10 @@ export declare const mulmoPromptTemplateFileSchema: z.ZodObject<{
             quality: z.ZodOptional<z.ZodString>;
             style: z.ZodOptional<z.ZodString>;
             moderation: z.ZodOptional<z.ZodString>;
+            baseURL: z.ZodOptional<z.ZodString>;
+            apiVersion: z.ZodOptional<z.ZodString>;
+            vertexai_project: z.ZodOptional<z.ZodString>;
+            vertexai_location: z.ZodOptional<z.ZodString>;
             images: z.ZodOptional<z.ZodRecord<z.ZodString, z.ZodUnion<readonly [z.ZodObject<{
                 type: z.ZodLiteral<"image">;
                 source: z.ZodDiscriminatedUnion<[z.ZodObject<{
@@ -3889,6 +3979,8 @@ export declare const mulmoPromptTemplateFileSchema: z.ZodObject<{
                 type: z.ZodLiteral<"custom">;
                 filter: z.ZodString;
             }, z.core.$strip>]>>>;
+            vertexai_project: z.ZodOptional<z.ZodString>;
+            vertexai_location: z.ZodOptional<z.ZodString>;
         }, z.core.$strip>>>;
         soundEffectParams: z.ZodDefault<z.ZodOptional<z.ZodObject<{
             provider: z.ZodOptional<z.ZodDefault<z.ZodEnum<{

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 +242,10 @@ 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")
+    vertexai_project: z.string().optional(), // Google Cloud Project ID for Vertex AI
+    vertexai_location: z.string().optional(), // Vertex AI location (default: us-central1)
 })
     .strict();
 export const mulmoImageParamsSchema = mulmoBeatImageParamsSchema
@@ -322,6 +328,8 @@ export const mulmoMovieParamsSchema = z.object({
     fillOption: mulmoFillOptionSchema.optional(), // for movie.ts
     transition: mulmoTransitionSchema.optional(), // for movie.ts
     filters: z.array(mulmoVideoFilterSchema).optional(), // for movie.ts
+    vertexai_project: z.string().optional(), // Google Cloud Project ID for Vertex AI
+    vertexai_location: z.string().optional(), // Vertex AI location (default: us-central1)
 });
 export const mulmoBeatSchema = z
     .object({