mulmocast 2.1.22 → 2.1.24

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 +50,10 @@ export declare const createStudioData: (_mulmoScript: MulmoScript, fileName: str
             quality?: string | undefined;
             style?: string | undefined;
             moderation?: string | undefined;
+            baseURL?: string | undefined;
+            apiVersion?: string | undefined;
+            vertexai_project?: string | undefined;
+            vertexai_location?: string | undefined;
             images?: Record<string, {
                 type: "image";
                 source: {
@@ -219,6 +227,8 @@ export declare const createStudioData: (_mulmoScript: MulmoScript, fileName: str
                 type: "custom";
                 filter: string;
             })[] | undefined;
+            vertexai_project?: string | undefined;
+            vertexai_location?: string | undefined;
         };
         soundEffectParams: {
             provider?: string | undefined;
@@ -369,6 +379,10 @@ export declare const createStudioData: (_mulmoScript: MulmoScript, fileName: str
                 quality?: string | undefined;
                 style?: string | undefined;
                 moderation?: string | undefined;
+                baseURL?: string | undefined;
+                apiVersion?: string | undefined;
+                vertexai_project?: string | undefined;
+                vertexai_location?: string | undefined;
             } | undefined;
             audioParams?: {
                 movieVolume: number;
@@ -530,6 +544,8 @@ export declare const createStudioData: (_mulmoScript: MulmoScript, fileName: str
                     type: "custom";
                     filter: string;
                 })[] | undefined;
+                vertexai_project?: string | undefined;
+                vertexai_location?: string | undefined;
                 speed?: number | undefined;
             } | undefined;
             soundEffectParams?: {
@@ -664,6 +680,8 @@ 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 +695,8 @@ export declare const initializeContextFromFiles: (files: FileObject, raiseError:
                         } | undefined;
                         provider?: string | undefined;
                         model?: string | undefined;
+                        baseURL?: string | undefined;
+                        apiVersion?: string | undefined;
                     }> | undefined;
                 }>;
             };
@@ -686,6 +706,10 @@ export declare const initializeContextFromFiles: (files: FileObject, raiseError:
                 quality?: string | undefined;
                 style?: string | undefined;
                 moderation?: string | undefined;
+                baseURL?: string | undefined;
+                apiVersion?: string | undefined;
+                vertexai_project?: string | undefined;
+                vertexai_location?: string | undefined;
                 images?: Record<string, {
                     type: "image";
                     source: {
@@ -859,6 +883,8 @@ export declare const initializeContextFromFiles: (files: FileObject, raiseError:
                     type: "custom";
                     filter: string;
                 })[] | undefined;
+                vertexai_project?: string | undefined;
+                vertexai_location?: string | undefined;
             };
             soundEffectParams: {
                 provider?: string | undefined;
@@ -1009,6 +1035,10 @@ export declare const initializeContextFromFiles: (files: FileObject, raiseError:
                     quality?: string | undefined;
                     style?: string | undefined;
                     moderation?: string | undefined;
+                    baseURL?: string | undefined;
+                    apiVersion?: string | undefined;
+                    vertexai_project?: string | undefined;
+                    vertexai_location?: string | undefined;
                 } | undefined;
                 audioParams?: {
                     movieVolume: number;
@@ -1170,6 +1200,8 @@ export declare const initializeContextFromFiles: (files: FileObject, raiseError:
                         type: "custom";
                         filter: string;
                     })[] | undefined;
+                    vertexai_project?: string | undefined;
+                    vertexai_location?: string | undefined;
                     speed?: number | undefined;
                 } | undefined;
                 soundEffectParams?: {
@@ -1311,6 +1343,8 @@ 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 +1358,8 @@ export declare const initializeContextFromFiles: (files: FileObject, raiseError:
                     } | undefined;
                     provider?: string | undefined;
                     model?: string | undefined;
+                    baseURL?: string | undefined;
+                    apiVersion?: string | undefined;
                 }> | undefined;
             }>;
         };
@@ -1333,6 +1369,10 @@ export declare const initializeContextFromFiles: (files: FileObject, raiseError:
             quality?: string | undefined;
             style?: string | undefined;
             moderation?: string | undefined;
+            baseURL?: string | undefined;
+            apiVersion?: string | undefined;
+            vertexai_project?: string | undefined;
+            vertexai_location?: string | undefined;
             images?: Record<string, {
                 type: "image";
                 source: {
@@ -1506,6 +1546,8 @@ export declare const initializeContextFromFiles: (files: FileObject, raiseError:
                 type: "custom";
                 filter: string;
             })[] | undefined;
+            vertexai_project?: string | undefined;
+            vertexai_location?: string | undefined;
         };
         soundEffectParams: {
             provider?: string | undefined;

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": "2.1.22",
+  "version": "2.1.24",
   "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/scripts/test/test_vertexai.json

@@ -0,0 +1,21 @@
+{
+  "$schema": "../../assets/schema.json",
+  "title": "Vertex AI Test",
+  "description": "Test Vertex AI integration",
+  "lang": "en",
+  "imageParams": {
+    "provider": "google",
+    "model": "imagen-4.0-generate-001",
+    "vertexai_project": "${GOOGLE_PROJECT_ID}",
+    "vertexai_location": "us-central1"
+  },
+  "speechParams": {
+    "provider": "mock"
+  },
+  "beats": [
+    {
+      "text": "Hello",
+      "imagePrompt": "A simple blue circle on white background"
+    }
+  ]
+}

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;