@aigne/gemini 0.14.16 → 1.74.0-beta

package/dist/gemini-video-model.d.cts

@@ -0,0 +1,117 @@
+import { AgentInvokeOptions, FileUnionContent, VideoModel, VideoModelInput, VideoModelOptions, VideoModelOutput } from "@aigne/core";
+import { GoogleGenAI } from "@google/genai";
+
+//#region src/gemini-video-model.d.ts
+/**
+ * Input options for Gemini Video Model
+ */
+interface GeminiVideoModelInput extends VideoModelInput {
+  /**
+       * Text describing content that should not appear in the video
+       */
+  negativePrompt?: string;
+  /**
+       * Aspect ratio of the video
+       *
+       * Veo 3.1: "16:9" (default, 720p and 1080p), "9:16" (720p and 1080p)
+       * Veo 3: "16:9" (default, 720p and 1080p), "9:16" (720p and 1080p)
+       */
+  aspectRatio?: "16:9" | "9:16";
+  /**
+       * Resolution of the video
+       *
+       * Veo 3.1: "720p" (default), "1080p" (only supports 8 seconds duration)
+       * Veo 3: "720p" (default), "1080p" (16:9 only)
+       */
+  size?: "720p" | "1080p";
+  /**
+       * Duration of the generated video in seconds
+       *
+       * Veo 3.1: "4", "6", "8"
+       * Veo 3: "4", "6", "8"
+       */
+  seconds?: "4" | "6" | "8";
+  /**
+       * Control person generation
+       *
+       * For text-to-video and image-to-video:
+       * - Veo 3.1: "allow_all" for image-to-video, frame interpolation and reference images; only "allow_adult" for text-to-video
+       * - Veo 3: "allow_all" for image-to-video; only "allow_adult" for text-to-video
+       * - Veo 2: "allow_all", "allow_adult", "dont_allow"
+       */
+  personGeneration?: string;
+  /**
+       * Last frame for video generation (frame interpolation)
+       */
+  lastFrame?: FileUnionContent;
+  /**
+       * Reference images for video generation
+       * Only supported in Veo 3.1 models
+       */
+  referenceImages?: FileUnionContent[];
+}
+/**
+ * Output from Gemini Video Model
+ */
+interface GeminiVideoModelOutput extends VideoModelOutput {}
+/**
+ * Configuration options for Gemini Video Model
+ */
+interface GeminiVideoModelOptions extends VideoModelOptions<GeminiVideoModelInput, GeminiVideoModelOutput> {
+  /**
+       * API key for Gemini API
+       *
+       * If not provided, will look for GEMINI_API_KEY in environment variables
+       */
+  apiKey?: string;
+  /**
+       * Base URL for Gemini API
+       *
+       * Useful for proxies or alternate endpoints
+       */
+  baseURL?: string;
+  /**
+       * Gemini model to use
+       *
+       * Defaults to 'veo-3.1-generate-preview'
+       */
+  model?: string;
+  /**
+       * Additional model options to control behavior
+       */
+  modelOptions?: Omit<Partial<GeminiVideoModelInput>, "model">;
+  /**
+       * Client options for Gemini API
+       */
+  clientOptions?: Record<string, any>;
+  /**
+       * Polling interval in milliseconds for checking video generation status
+       *
+       * Defaults to 10000ms (10 seconds)
+       */
+  pollingInterval?: number;
+}
+declare class GeminiVideoModel extends VideoModel<GeminiVideoModelInput, GeminiVideoModelOutput> {
+  options?: GeminiVideoModelOptions | undefined;
+  constructor(options?: GeminiVideoModelOptions | undefined);
+  /**
+       * @hidden
+       */
+  protected _client?: GoogleGenAI;
+  protected apiKeyEnvName: string;
+  get client(): GoogleGenAI;
+  get credential(): {
+    url: string | undefined;
+    apiKey: string | undefined;
+    model: string;
+  };
+  get modelOptions(): Omit<Partial<GeminiVideoModelInput>, "model"> | undefined;
+  downloadToFile(dir: string, videoId: string, videoFile: {
+    uri?: string;
+    videoBytes?: any;
+  }): Promise<string>;
+  process(input: GeminiVideoModelInput, options: AgentInvokeOptions): Promise<GeminiVideoModelOutput>;
+}
+//#endregion
+export { GeminiVideoModel, GeminiVideoModelInput, GeminiVideoModelOptions, GeminiVideoModelOutput };
+//# sourceMappingURL=gemini-video-model.d.cts.map

package/dist/gemini-video-model.d.cts.map

@@ -0,0 +1 @@
+{"version":3,"file":"gemini-video-model.d.cts","names":[],"sources":["../src/gemini-video-model.ts"],"mappings":";;;;;AAqBA;;UAAiB,qBAAA,SAA8B,eAAA;EAAA;;AAuD/C;EAvD+C,cAAA;EAAA;;AAuD/C;AAKA;;;EA5D+C,WAAA;EAAA;;AAuD/C;AAKA;;;EA5D+C,IAAA;EAAA;;AAuD/C;AAKA;;;EA5D+C,OAAA;EAAA;;AAuD/C;AAKA;;;;;EA5D+C,gBAAA;EAAA;;AAuD/C;EAvD+C,SAAA,GA2CjC,gBAAA;EAAA;;;;EAAA,eAAA,GAMM,gBAAA;AAAA;AAAA;;;AAAA,UAMH,sBAAA,SAA+B,gBAAA;AAAA;AAKhD;;AALgD,UAK/B,uBAAA,SACP,iBAAA,CAAkB,qBAAA,EAAuB,sBAAA;EAAA;;;;;EAAA,MAAA;EAAA;;;;;EAAA,OAAA;EAAA;;;;;EAAA,KAAA;EAAA;;;EAAA,YAAA,GAyBlC,IAAA,CAAK,OAAA,CAAQ,qBAAA;EAAA;;;EAAA,aAAA,GAKZ,MAAA;EAAA;;;AA6BlB;;EA7BkB,eAAA;AAAA;AAAA,cA6BL,gBAAA,SAAyB,UAAA,CAAW,qBAAA,EAAuB,sBAAA;EAAA,OAAA,GAChC,uBAAA;EAAA,YAAA,OAAA,GAAA,uBAAA;EAAA;;;EAAA,UAAA,OAAA,GAalB,WAAA;EAAA,UAAA,aAAA;EAAA,IAAA,OAAA,GAIV,WAAA;EAAA,IAAA,WAAA;IAAA,GAAA;IAAA,MAAA;IAAA,KAAA;EAAA;EAAA,IAAA,aAAA,GAmBM,IAAA,CAAA,OAAA,CAAA,qBAAA;EAAA,eAAA,GAAA,UAAA,OAAA,UAAA,SAAA;IAAA,GAAA;IAAA,UAAA;EAAA,IAQb,OAAA;EAAA,QAAA,KAAA,EAaM,qBAAA,EAAA,OAAA,EACE,kBAAA,GACR,OAAA,CAAQ,sBAAA;AAAA"}

package/dist/gemini-video-model.d.mts

@@ -0,0 +1,117 @@
+import { AgentInvokeOptions, FileUnionContent, VideoModel, VideoModelInput, VideoModelOptions, VideoModelOutput } from "@aigne/core";
+import { GoogleGenAI } from "@google/genai";
+
+//#region src/gemini-video-model.d.ts
+/**
+ * Input options for Gemini Video Model
+ */
+interface GeminiVideoModelInput extends VideoModelInput {
+  /**
+       * Text describing content that should not appear in the video
+       */
+  negativePrompt?: string;
+  /**
+       * Aspect ratio of the video
+       *
+       * Veo 3.1: "16:9" (default, 720p and 1080p), "9:16" (720p and 1080p)
+       * Veo 3: "16:9" (default, 720p and 1080p), "9:16" (720p and 1080p)
+       */
+  aspectRatio?: "16:9" | "9:16";
+  /**
+       * Resolution of the video
+       *
+       * Veo 3.1: "720p" (default), "1080p" (only supports 8 seconds duration)
+       * Veo 3: "720p" (default), "1080p" (16:9 only)
+       */
+  size?: "720p" | "1080p";
+  /**
+       * Duration of the generated video in seconds
+       *
+       * Veo 3.1: "4", "6", "8"
+       * Veo 3: "4", "6", "8"
+       */
+  seconds?: "4" | "6" | "8";
+  /**
+       * Control person generation
+       *
+       * For text-to-video and image-to-video:
+       * - Veo 3.1: "allow_all" for image-to-video, frame interpolation and reference images; only "allow_adult" for text-to-video
+       * - Veo 3: "allow_all" for image-to-video; only "allow_adult" for text-to-video
+       * - Veo 2: "allow_all", "allow_adult", "dont_allow"
+       */
+  personGeneration?: string;
+  /**
+       * Last frame for video generation (frame interpolation)
+       */
+  lastFrame?: FileUnionContent;
+  /**
+       * Reference images for video generation
+       * Only supported in Veo 3.1 models
+       */
+  referenceImages?: FileUnionContent[];
+}
+/**
+ * Output from Gemini Video Model
+ */
+interface GeminiVideoModelOutput extends VideoModelOutput {}
+/**
+ * Configuration options for Gemini Video Model
+ */
+interface GeminiVideoModelOptions extends VideoModelOptions<GeminiVideoModelInput, GeminiVideoModelOutput> {
+  /**
+       * API key for Gemini API
+       *
+       * If not provided, will look for GEMINI_API_KEY in environment variables
+       */
+  apiKey?: string;
+  /**
+       * Base URL for Gemini API
+       *
+       * Useful for proxies or alternate endpoints
+       */
+  baseURL?: string;
+  /**
+       * Gemini model to use
+       *
+       * Defaults to 'veo-3.1-generate-preview'
+       */
+  model?: string;
+  /**
+       * Additional model options to control behavior
+       */
+  modelOptions?: Omit<Partial<GeminiVideoModelInput>, "model">;
+  /**
+       * Client options for Gemini API
+       */
+  clientOptions?: Record<string, any>;
+  /**
+       * Polling interval in milliseconds for checking video generation status
+       *
+       * Defaults to 10000ms (10 seconds)
+       */
+  pollingInterval?: number;
+}
+declare class GeminiVideoModel extends VideoModel<GeminiVideoModelInput, GeminiVideoModelOutput> {
+  options?: GeminiVideoModelOptions | undefined;
+  constructor(options?: GeminiVideoModelOptions | undefined);
+  /**
+       * @hidden
+       */
+  protected _client?: GoogleGenAI;
+  protected apiKeyEnvName: string;
+  get client(): GoogleGenAI;
+  get credential(): {
+    url: string | undefined;
+    apiKey: string | undefined;
+    model: string;
+  };
+  get modelOptions(): Omit<Partial<GeminiVideoModelInput>, "model"> | undefined;
+  downloadToFile(dir: string, videoId: string, videoFile: {
+    uri?: string;
+    videoBytes?: any;
+  }): Promise<string>;
+  process(input: GeminiVideoModelInput, options: AgentInvokeOptions): Promise<GeminiVideoModelOutput>;
+}
+//#endregion
+export { GeminiVideoModel, GeminiVideoModelInput, GeminiVideoModelOptions, GeminiVideoModelOutput };
+//# sourceMappingURL=gemini-video-model.d.mts.map

package/dist/gemini-video-model.d.mts.map

@@ -0,0 +1 @@
+{"version":3,"file":"gemini-video-model.d.mts","names":[],"sources":["../src/gemini-video-model.ts"],"mappings":";;;;;AAqBA;;UAAiB,qBAAA,SAA8B,eAAA;EAAA;;AAuD/C;EAvD+C,cAAA;EAAA;;AAuD/C;AAKA;;;EA5D+C,WAAA;EAAA;;AAuD/C;AAKA;;;EA5D+C,IAAA;EAAA;;AAuD/C;AAKA;;;EA5D+C,OAAA;EAAA;;AAuD/C;AAKA;;;;;EA5D+C,gBAAA;EAAA;;AAuD/C;EAvD+C,SAAA,GA2CjC,gBAAA;EAAA;;;;EAAA,eAAA,GAMM,gBAAA;AAAA;AAAA;;;AAAA,UAMH,sBAAA,SAA+B,gBAAA;AAAA;AAKhD;;AALgD,UAK/B,uBAAA,SACP,iBAAA,CAAkB,qBAAA,EAAuB,sBAAA;EAAA;;;;;EAAA,MAAA;EAAA;;;;;EAAA,OAAA;EAAA;;;;;EAAA,KAAA;EAAA;;;EAAA,YAAA,GAyBlC,IAAA,CAAK,OAAA,CAAQ,qBAAA;EAAA;;;EAAA,aAAA,GAKZ,MAAA;EAAA;;;AA6BlB;;EA7BkB,eAAA;AAAA;AAAA,cA6BL,gBAAA,SAAyB,UAAA,CAAW,qBAAA,EAAuB,sBAAA;EAAA,OAAA,GAChC,uBAAA;EAAA,YAAA,OAAA,GAAA,uBAAA;EAAA;;;EAAA,UAAA,OAAA,GAalB,WAAA;EAAA,UAAA,aAAA;EAAA,IAAA,OAAA,GAIV,WAAA;EAAA,IAAA,WAAA;IAAA,GAAA;IAAA,MAAA;IAAA,KAAA;EAAA;EAAA,IAAA,aAAA,GAmBM,IAAA,CAAA,OAAA,CAAA,qBAAA;EAAA,eAAA,GAAA,UAAA,OAAA,UAAA,SAAA;IAAA,GAAA;IAAA,UAAA;EAAA,IAQb,OAAA;EAAA,QAAA,KAAA,EAaM,qBAAA,EAAA,OAAA,EACE,kBAAA,GACR,OAAA,CAAQ,sBAAA;AAAA"}

package/dist/gemini-video-model.mjs

@@ -0,0 +1,149 @@
+import { waitFileSizeStable } from "./utils.mjs";
+import { VideoModel, fileUnionContentSchema, videoModelInputSchema } from "@aigne/core";
+import { logger } from "@aigne/core/utils/logger";
+import { checkArguments } from "@aigne/core/utils/type-utils";
+import { nodejs } from "@aigne/utils/nodejs";
+import { GoogleGenAI } from "@google/genai";
+import { z } from "zod";
+
+//#region src/gemini-video-model.ts
+const DEFAULT_MODEL = "veo-3.1-generate-preview";
+const DEFAULT_SECONDS = 8;
+const geminiVideoModelInputSchema = videoModelInputSchema.extend({
+	negativePrompt: z.string().optional(),
+	aspectRatio: z.enum(["16:9", "9:16"]).optional(),
+	size: z.enum(["720p", "1080p"]).optional(),
+	seconds: z.enum([
+		"4",
+		"6",
+		"8"
+	]).optional(),
+	personGeneration: z.string().optional(),
+	lastFrame: fileUnionContentSchema.optional(),
+	referenceImages: fileUnionContentSchema.array().optional()
+});
+const geminiVideoModelOptionsSchema = z.object({
+	apiKey: z.string().optional(),
+	baseURL: z.string().optional(),
+	model: z.string().optional(),
+	modelOptions: z.object({}).optional(),
+	clientOptions: z.object({}).optional(),
+	pollingInterval: z.number().optional()
+});
+var GeminiVideoModel = class extends VideoModel {
+	constructor(options) {
+		super({
+			...options,
+			description: options?.description ?? "Generate videos using Google Gemini Veo models",
+			inputSchema: geminiVideoModelInputSchema
+		});
+		this.options = options;
+		if (options) checkArguments(this.name, geminiVideoModelOptionsSchema, options);
+	}
+	/**
+	* @hidden
+	*/
+	_client;
+	apiKeyEnvName = "GEMINI_API_KEY";
+	get client() {
+		const { apiKey } = this.credential;
+		if (!apiKey) throw new Error(`${this.name} requires an API key. Please provide it via \`options.apiKey\`, or set the \`${this.apiKeyEnvName}\` environment variable`);
+		this._client ??= new GoogleGenAI({
+			apiKey,
+			...this.options?.clientOptions
+		});
+		return this._client;
+	}
+	get credential() {
+		return {
+			url: this.options?.baseURL || process.env.GEMINI_BASE_URL,
+			apiKey: this.options?.apiKey || process.env[this.apiKeyEnvName],
+			model: this.options?.model || DEFAULT_MODEL
+		};
+	}
+	get modelOptions() {
+		return this.options?.modelOptions;
+	}
+	async downloadToFile(dir, videoId, videoFile) {
+		logger.debug("Downloading video content...");
+		const localPath = nodejs.path.join(dir, `${videoId}.mp4`);
+		await this.client.files.download({
+			file: videoFile,
+			downloadPath: localPath
+		});
+		logger.debug(`Generated video saved to ${localPath}`);
+		await waitFileSizeStable(localPath);
+		return (await nodejs.fs.readFile(localPath)).toString("base64");
+	}
+	async process(input, options) {
+		const model = input.model ?? input.modelOptions?.model ?? this.credential.model;
+		const mergedInput = {
+			...this.modelOptions,
+			...input
+		};
+		if (mergedInput.referenceImages && !model.includes("veo-3.1")) throw new Error("referenceImages is only supported in Veo 3.1 models");
+		const config = {};
+		if (mergedInput.negativePrompt) config.negativePrompt = mergedInput.negativePrompt;
+		if (mergedInput.aspectRatio) config.aspectRatio = mergedInput.aspectRatio;
+		if (mergedInput.size) config.resolution = mergedInput.size;
+		if (mergedInput.seconds) config.durationSeconds = parseInt(mergedInput.seconds, 10);
+		if (mergedInput.personGeneration) config.personGeneration = mergedInput.personGeneration;
+		if (mergedInput.lastFrame) config.lastFrame = await this.transformFileType("file", mergedInput.lastFrame).then((file) => {
+			return {
+				imageBytes: file.data,
+				mimeType: file.mimeType
+			};
+		});
+		if (mergedInput.referenceImages) config.referenceImages = await Promise.all(mergedInput.referenceImages.map(async (image) => {
+			return await this.transformFileType("file", image).then((file) => {
+				return { image: {
+					imageBytes: file.data,
+					mimeType: file.mimeType
+				} };
+			});
+		}));
+		const params = {
+			model,
+			prompt: mergedInput.prompt,
+			config
+		};
+		if (mergedInput.image) params.image = await this.transformFileType("file", mergedInput.image).then((file) => {
+			return {
+				imageBytes: file.data,
+				mimeType: file.mimeType
+			};
+		});
+		let operation = await this.client.models.generateVideos(params);
+		logger.debug("Video generation started...");
+		const pollingInterval = this.options?.pollingInterval ?? 1e4;
+		while (!operation.done) {
+			logger.debug("Waiting for video generation to complete...");
+			await new Promise((resolve) => setTimeout(resolve, pollingInterval));
+			operation = await this.client.operations.getVideosOperation({ operation });
+		}
+		if (!operation.response?.generatedVideos?.[0]?.video) throw new Error("Video generation failed: No video generated");
+		const videoFile = operation.response.generatedVideos[0].video;
+		if (!videoFile) throw new Error("Video generation failed: No video file returned");
+		const dir = nodejs.path.join(nodejs.os.tmpdir(), options?.context?.id || "");
+		await nodejs.fs.mkdir(dir, { recursive: true });
+		const videoId = Date.now().toString();
+		return {
+			videos: [{
+				type: "file",
+				data: await this.downloadToFile(dir, videoId, videoFile),
+				mimeType: "video/mp4",
+				filename: `${videoId}.mp4`
+			}],
+			usage: {
+				inputTokens: 0,
+				outputTokens: 0
+			},
+			model,
+			seconds: mergedInput.seconds ? parseInt(mergedInput.seconds, 10) : DEFAULT_SECONDS
+		};
+	}
+};
+
+//#endregion
+export { GeminiVideoModel };
+//# sourceMappingURL=gemini-video-model.mjs.map

package/dist/gemini-video-model.mjs.map

@@ -0,0 +1 @@
+{"version":3,"file":"gemini-video-model.mjs","names":[],"sources":["../src/gemini-video-model.ts"],"sourcesContent":["import type {\n  AgentInvokeOptions,\n  FileUnionContent,\n  VideoModelInput,\n  VideoModelOptions,\n  VideoModelOutput,\n} from \"@aigne/core\";\nimport { fileUnionContentSchema, VideoModel, videoModelInputSchema } from \"@aigne/core\";\nimport { logger } from \"@aigne/core/utils/logger\";\nimport { checkArguments } from \"@aigne/core/utils/type-utils\";\nimport { nodejs } from \"@aigne/utils/nodejs\";\nimport { type GenerateVideosParameters, GoogleGenAI } from \"@google/genai\";\nimport { type ZodType, z } from \"zod\";\nimport { waitFileSizeStable } from \"./utils.js\";\n\nconst DEFAULT_MODEL = \"veo-3.1-generate-preview\";\nconst DEFAULT_SECONDS = 8;\n\n/**\n * Input options for Gemini Video Model\n */\nexport interface GeminiVideoModelInput extends VideoModelInput {\n  /**\n   * Text describing content that should not appear in the video\n   */\n  negativePrompt?: string;\n\n  /**\n   * Aspect ratio of the video\n   *\n   * Veo 3.1: \"16:9\" (default, 720p and 1080p), \"9:16\" (720p and 1080p)\n   * Veo 3: \"16:9\" (default, 720p and 1080p), \"9:16\" (720p and 1080p)\n   */\n  aspectRatio?: \"16:9\" | \"9:16\";\n\n  /**\n   * Resolution of the video\n   *\n   * Veo 3.1: \"720p\" (default), \"1080p\" (only supports 8 seconds duration)\n   * Veo 3: \"720p\" (default), \"1080p\" (16:9 only)\n   */\n  size?: \"720p\" | \"1080p\";\n\n  /**\n   * Duration of the generated video in seconds\n   *\n   * Veo 3.1: \"4\", \"6\", \"8\"\n   * Veo 3: \"4\", \"6\", \"8\"\n   */\n  seconds?: \"4\" | \"6\" | \"8\";\n\n  /**\n   * Control person generation\n   *\n   * For text-to-video and image-to-video:\n   * - Veo 3.1: \"allow_all\" for image-to-video, frame interpolation and reference images; only \"allow_adult\" for text-to-video\n   * - Veo 3: \"allow_all\" for image-to-video; only \"allow_adult\" for text-to-video\n   * - Veo 2: \"allow_all\", \"allow_adult\", \"dont_allow\"\n   */\n  personGeneration?: string;\n\n  /**\n   * Last frame for video generation (frame interpolation)\n   */\n  lastFrame?: FileUnionContent;\n\n  /**\n   * Reference images for video generation\n   * Only supported in Veo 3.1 models\n   */\n  referenceImages?: FileUnionContent[];\n}\n\n/**\n * Output from Gemini Video Model\n */\nexport interface GeminiVideoModelOutput extends VideoModelOutput {}\n\n/**\n * Configuration options for Gemini Video Model\n */\nexport interface GeminiVideoModelOptions\n  extends VideoModelOptions<GeminiVideoModelInput, GeminiVideoModelOutput> {\n  /**\n   * API key for Gemini API\n   *\n   * If not provided, will look for GEMINI_API_KEY in environment variables\n   */\n  apiKey?: string;\n\n  /**\n   * Base URL for Gemini API\n   *\n   * Useful for proxies or alternate endpoints\n   */\n  baseURL?: string;\n\n  /**\n   * Gemini model to use\n   *\n   * Defaults to 'veo-3.1-generate-preview'\n   */\n  model?: string;\n\n  /**\n   * Additional model options to control behavior\n   */\n  modelOptions?: Omit<Partial<GeminiVideoModelInput>, \"model\">;\n\n  /**\n   * Client options for Gemini API\n   */\n  clientOptions?: Record<string, any>;\n\n  /**\n   * Polling interval in milliseconds for checking video generation status\n   *\n   * Defaults to 10000ms (10 seconds)\n   */\n  pollingInterval?: number;\n}\n\nconst geminiVideoModelInputSchema: ZodType<GeminiVideoModelInput> = videoModelInputSchema.extend({\n  negativePrompt: z.string().optional(),\n  aspectRatio: z.enum([\"16:9\", \"9:16\"]).optional(),\n  size: z.enum([\"720p\", \"1080p\"]).optional(),\n  seconds: z.enum([\"4\", \"6\", \"8\"]).optional(),\n  personGeneration: z.string().optional(),\n  lastFrame: fileUnionContentSchema.optional(),\n  referenceImages: fileUnionContentSchema.array().optional(),\n});\n\nconst geminiVideoModelOptionsSchema = z.object({\n  apiKey: z.string().optional(),\n  baseURL: z.string().optional(),\n  model: z.string().optional(),\n  modelOptions: z.object({}).optional(),\n  clientOptions: z.object({}).optional(),\n  pollingInterval: z.number().optional(),\n});\n\nexport class GeminiVideoModel extends VideoModel<GeminiVideoModelInput, GeminiVideoModelOutput> {\n  constructor(public override options?: GeminiVideoModelOptions) {\n    super({\n      ...options,\n      description: options?.description ?? \"Generate videos using Google Gemini Veo models\",\n      inputSchema: geminiVideoModelInputSchema,\n    });\n\n    if (options) checkArguments(this.name, geminiVideoModelOptionsSchema, options);\n  }\n\n  /**\n   * @hidden\n   */\n  protected _client?: GoogleGenAI;\n\n  protected apiKeyEnvName = \"GEMINI_API_KEY\";\n\n  get client() {\n    const { apiKey } = this.credential;\n    if (!apiKey)\n      throw new Error(\n        `${this.name} requires an API key. Please provide it via \\`options.apiKey\\`, or set the \\`${this.apiKeyEnvName}\\` environment variable`,\n      );\n\n    this._client ??= new GoogleGenAI({ apiKey, ...this.options?.clientOptions });\n    return this._client;\n  }\n\n  override get credential() {\n    return {\n      url: this.options?.baseURL || process.env.GEMINI_BASE_URL,\n      apiKey: this.options?.apiKey || process.env[this.apiKeyEnvName],\n      model: this.options?.model || DEFAULT_MODEL,\n    };\n  }\n\n  get modelOptions() {\n    return this.options?.modelOptions;\n  }\n\n  async downloadToFile(\n    dir: string,\n    videoId: string,\n    videoFile: { uri?: string; videoBytes?: any },\n  ): Promise<string> {\n    logger.debug(\"Downloading video content...\");\n    const localPath = nodejs.path.join(dir, `${videoId}.mp4`);\n    await this.client.files.download({ file: videoFile, downloadPath: localPath });\n    logger.debug(`Generated video saved to ${localPath}`);\n\n    await waitFileSizeStable(localPath);\n\n    const buffer = await nodejs.fs.readFile(localPath);\n    return buffer.toString(\"base64\");\n  }\n\n  override async process(\n    input: GeminiVideoModelInput,\n    options: AgentInvokeOptions,\n  ): Promise<GeminiVideoModelOutput> {\n    const model = input.model ?? input.modelOptions?.model ?? this.credential.model;\n    const mergedInput = { ...this.modelOptions, ...input };\n\n    if (mergedInput.referenceImages && !model.includes(\"veo-3.1\")) {\n      throw new Error(\"referenceImages is only supported in Veo 3.1 models\");\n    }\n\n    const config: GenerateVideosParameters[\"config\"] = {};\n    if (mergedInput.negativePrompt) config.negativePrompt = mergedInput.negativePrompt;\n    if (mergedInput.aspectRatio) config.aspectRatio = mergedInput.aspectRatio;\n    if (mergedInput.size) config.resolution = mergedInput.size;\n    if (mergedInput.seconds) config.durationSeconds = parseInt(mergedInput.seconds, 10);\n    if (mergedInput.personGeneration) config.personGeneration = mergedInput.personGeneration;\n    if (mergedInput.lastFrame) {\n      config.lastFrame = await this.transformFileType(\"file\", mergedInput.lastFrame).then(\n        (file) => {\n          return {\n            imageBytes: file.data,\n            mimeType: file.mimeType,\n          };\n        },\n      );\n    }\n\n    if (mergedInput.referenceImages) {\n      config.referenceImages = await Promise.all(\n        mergedInput.referenceImages.map(async (image) => {\n          return await this.transformFileType(\"file\", image).then((file) => {\n            return {\n              image: {\n                imageBytes: file.data,\n                mimeType: file.mimeType,\n              },\n            };\n          });\n        }),\n      );\n    }\n\n    const params: GenerateVideosParameters = {\n      model,\n      prompt: mergedInput.prompt,\n      config,\n    };\n\n    if (mergedInput.image) {\n      params.image = await this.transformFileType(\"file\", mergedInput.image).then((file) => {\n        return {\n          imageBytes: file.data,\n          mimeType: file.mimeType,\n        };\n      });\n    }\n\n    // Start video generation\n    let operation = await this.client.models.generateVideos(params);\n    logger.debug(\"Video generation started...\");\n\n    // Poll operation status until complete\n    const pollingInterval = this.options?.pollingInterval ?? 10000;\n    while (!operation.done) {\n      logger.debug(\"Waiting for video generation to complete...\");\n      await new Promise((resolve) => setTimeout(resolve, pollingInterval));\n      operation = await this.client.operations.getVideosOperation({ operation });\n    }\n\n    if (!operation.response?.generatedVideos?.[0]?.video) {\n      throw new Error(\"Video generation failed: No video generated\");\n    }\n\n    // Download the generated video\n    const generatedVideo = operation.response.generatedVideos[0];\n    const videoFile = generatedVideo.video;\n\n    if (!videoFile) {\n      throw new Error(\"Video generation failed: No video file returned\");\n    }\n\n    // Save to temporary directory\n    const dir = nodejs.path.join(nodejs.os.tmpdir(), options?.context?.id || \"\");\n    await nodejs.fs.mkdir(dir, { recursive: true });\n\n    const videoId = Date.now().toString();\n\n    return {\n      videos: [\n        {\n          type: \"file\",\n          data: await this.downloadToFile(dir, videoId, videoFile),\n          mimeType: \"video/mp4\",\n          filename: `${videoId}.mp4`,\n        },\n      ],\n      usage: {\n        inputTokens: 0,\n        outputTokens: 0,\n      },\n      model,\n      seconds: mergedInput.seconds ? parseInt(mergedInput.seconds, 10) : DEFAULT_SECONDS,\n    };\n  }\n}\n"],"mappings":";;;;;;;;;AAeA,MAAM,gBAAgB;AACtB,MAAM,kBAAkB;AA0GxB,MAAM,8BAA8D,sBAAsB,OAAO;CAC/F,gBAAgB,EAAE,QAAQ,CAAC,UAAU;CACrC,aAAa,EAAE,KAAK,CAAC,QAAQ,OAAO,CAAC,CAAC,UAAU;CAChD,MAAM,EAAE,KAAK,CAAC,QAAQ,QAAQ,CAAC,CAAC,UAAU;CAC1C,SAAS,EAAE,KAAK;EAAC;EAAK;EAAK;EAAI,CAAC,CAAC,UAAU;CAC3C,kBAAkB,EAAE,QAAQ,CAAC,UAAU;CACvC,WAAW,uBAAuB,UAAU;CAC5C,iBAAiB,uBAAuB,OAAO,CAAC,UAAU;CAC3D,CAAC;AAEF,MAAM,gCAAgC,EAAE,OAAO;CAC7C,QAAQ,EAAE,QAAQ,CAAC,UAAU;CAC7B,SAAS,EAAE,QAAQ,CAAC,UAAU;CAC9B,OAAO,EAAE,QAAQ,CAAC,UAAU;CAC5B,cAAc,EAAE,OAAO,EAAE,CAAC,CAAC,UAAU;CACrC,eAAe,EAAE,OAAO,EAAE,CAAC,CAAC,UAAU;CACtC,iBAAiB,EAAE,QAAQ,CAAC,UAAU;CACvC,CAAC;AAEF,IAAa,mBAAb,cAAsC,WAA0D;CAC9F,YAAY,AAAgB,SAAmC;AAC7D,QAAM;GACJ,GAAG;GACH,aAAa,SAAS,eAAe;GACrC,aAAa;GACd,CAAC;EALwB;AAO1B,MAAI,QAAS,gBAAe,KAAK,MAAM,+BAA+B,QAAQ;;;;;CAMhF,AAAU;CAEV,AAAU,gBAAgB;CAE1B,IAAI,SAAS;EACX,MAAM,EAAE,WAAW,KAAK;AACxB,MAAI,CAAC,OACH,OAAM,IAAI,MACR,GAAG,KAAK,KAAK,+EAA+E,KAAK,cAAc,yBAChH;AAEH,OAAK,YAAY,IAAI,YAAY;GAAE;GAAQ,GAAG,KAAK,SAAS;GAAe,CAAC;AAC5E,SAAO,KAAK;;CAGd,IAAa,aAAa;AACxB,SAAO;GACL,KAAK,KAAK,SAAS,WAAW,QAAQ,IAAI;GAC1C,QAAQ,KAAK,SAAS,UAAU,QAAQ,IAAI,KAAK;GACjD,OAAO,KAAK,SAAS,SAAS;GAC/B;;CAGH,IAAI,eAAe;AACjB,SAAO,KAAK,SAAS;;CAGvB,MAAM,eACJ,KACA,SACA,WACiB;AACjB,SAAO,MAAM,+BAA+B;EAC5C,MAAM,YAAY,OAAO,KAAK,KAAK,KAAK,GAAG,QAAQ,MAAM;AACzD,QAAM,KAAK,OAAO,MAAM,SAAS;GAAE,MAAM;GAAW,cAAc;GAAW,CAAC;AAC9E,SAAO,MAAM,4BAA4B,YAAY;AAErD,QAAM,mBAAmB,UAAU;AAGnC,UADe,MAAM,OAAO,GAAG,SAAS,UAAU,EACpC,SAAS,SAAS;;CAGlC,MAAe,QACb,OACA,SACiC;EACjC,MAAM,QAAQ,MAAM,SAAS,MAAM,cAAc,SAAS,KAAK,WAAW;EAC1E,MAAM,cAAc;GAAE,GAAG,KAAK;GAAc,GAAG;GAAO;AAEtD,MAAI,YAAY,mBAAmB,CAAC,MAAM,SAAS,UAAU,CAC3D,OAAM,IAAI,MAAM,sDAAsD;EAGxE,MAAM,SAA6C,EAAE;AACrD,MAAI,YAAY,eAAgB,QAAO,iBAAiB,YAAY;AACpE,MAAI,YAAY,YAAa,QAAO,cAAc,YAAY;AAC9D,MAAI,YAAY,KAAM,QAAO,aAAa,YAAY;AACtD,MAAI,YAAY,QAAS,QAAO,kBAAkB,SAAS,YAAY,SAAS,GAAG;AACnF,MAAI,YAAY,iBAAkB,QAAO,mBAAmB,YAAY;AACxE,MAAI,YAAY,UACd,QAAO,YAAY,MAAM,KAAK,kBAAkB,QAAQ,YAAY,UAAU,CAAC,MAC5E,SAAS;AACR,UAAO;IACL,YAAY,KAAK;IACjB,UAAU,KAAK;IAChB;IAEJ;AAGH,MAAI,YAAY,gBACd,QAAO,kBAAkB,MAAM,QAAQ,IACrC,YAAY,gBAAgB,IAAI,OAAO,UAAU;AAC/C,UAAO,MAAM,KAAK,kBAAkB,QAAQ,MAAM,CAAC,MAAM,SAAS;AAChE,WAAO,EACL,OAAO;KACL,YAAY,KAAK;KACjB,UAAU,KAAK;KAChB,EACF;KACD;IACF,CACH;EAGH,MAAM,SAAmC;GACvC;GACA,QAAQ,YAAY;GACpB;GACD;AAED,MAAI,YAAY,MACd,QAAO,QAAQ,MAAM,KAAK,kBAAkB,QAAQ,YAAY,MAAM,CAAC,MAAM,SAAS;AACpF,UAAO;IACL,YAAY,KAAK;IACjB,UAAU,KAAK;IAChB;IACD;EAIJ,IAAI,YAAY,MAAM,KAAK,OAAO,OAAO,eAAe,OAAO;AAC/D,SAAO,MAAM,8BAA8B;EAG3C,MAAM,kBAAkB,KAAK,SAAS,mBAAmB;AACzD,SAAO,CAAC,UAAU,MAAM;AACtB,UAAO,MAAM,8CAA8C;AAC3D,SAAM,IAAI,SAAS,YAAY,WAAW,SAAS,gBAAgB,CAAC;AACpE,eAAY,MAAM,KAAK,OAAO,WAAW,mBAAmB,EAAE,WAAW,CAAC;;AAG5E,MAAI,CAAC,UAAU,UAAU,kBAAkB,IAAI,MAC7C,OAAM,IAAI,MAAM,8CAA8C;EAKhE,MAAM,YADiB,UAAU,SAAS,gBAAgB,GACzB;AAEjC,MAAI,CAAC,UACH,OAAM,IAAI,MAAM,kDAAkD;EAIpE,MAAM,MAAM,OAAO,KAAK,KAAK,OAAO,GAAG,QAAQ,EAAE,SAAS,SAAS,MAAM,GAAG;AAC5E,QAAM,OAAO,GAAG,MAAM,KAAK,EAAE,WAAW,MAAM,CAAC;EAE/C,MAAM,UAAU,KAAK,KAAK,CAAC,UAAU;AAErC,SAAO;GACL,QAAQ,CACN;IACE,MAAM;IACN,MAAM,MAAM,KAAK,eAAe,KAAK,SAAS,UAAU;IACxD,UAAU;IACV,UAAU,GAAG,QAAQ;IACtB,CACF;GACD,OAAO;IACL,aAAa;IACb,cAAc;IACf;GACD;GACA,SAAS,YAAY,UAAU,SAAS,YAAY,SAAS,GAAG,GAAG;GACpE"}

package/dist/index.cjs

@@ -0,0 +1,7 @@
+const require_gemini_chat_model = require('./gemini-chat-model.cjs');
+const require_gemini_image_model = require('./gemini-image-model.cjs');
+const require_gemini_video_model = require('./gemini-video-model.cjs');
+
+exports.GeminiChatModel = require_gemini_chat_model.GeminiChatModel;
+exports.GeminiImageModel = require_gemini_image_model.GeminiImageModel;
+exports.GeminiVideoModel = require_gemini_video_model.GeminiVideoModel;

package/dist/index.d.cts

@@ -0,0 +1,4 @@
+import { GeminiChatModel, GeminiChatModelOptions } from "./gemini-chat-model.cjs";
+import { GeminiImageModel, GeminiImageModelInput, GeminiImageModelOptions, GeminiImageModelOutput } from "./gemini-image-model.cjs";
+import { GeminiVideoModel, GeminiVideoModelInput, GeminiVideoModelOptions, GeminiVideoModelOutput } from "./gemini-video-model.cjs";
+export { GeminiChatModel, GeminiChatModelOptions, GeminiImageModel, GeminiImageModelInput, GeminiImageModelOptions, GeminiImageModelOutput, GeminiVideoModel, GeminiVideoModelInput, GeminiVideoModelOptions, GeminiVideoModelOutput };

package/dist/index.d.mts

@@ -0,0 +1,4 @@
+import { GeminiChatModel, GeminiChatModelOptions } from "./gemini-chat-model.mjs";
+import { GeminiImageModel, GeminiImageModelInput, GeminiImageModelOptions, GeminiImageModelOutput } from "./gemini-image-model.mjs";
+import { GeminiVideoModel, GeminiVideoModelInput, GeminiVideoModelOptions, GeminiVideoModelOutput } from "./gemini-video-model.mjs";
+export { GeminiChatModel, GeminiChatModelOptions, GeminiImageModel, GeminiImageModelInput, GeminiImageModelOptions, GeminiImageModelOutput, GeminiVideoModel, GeminiVideoModelInput, GeminiVideoModelOptions, GeminiVideoModelOutput };

package/dist/index.mjs

@@ -0,0 +1,5 @@
+import { GeminiChatModel } from "./gemini-chat-model.mjs";
+import { GeminiImageModel } from "./gemini-image-model.mjs";
+import { GeminiVideoModel } from "./gemini-video-model.mjs";
+
+export { GeminiChatModel, GeminiImageModel, GeminiVideoModel };

package/dist/utils.cjs

@@ -0,0 +1,34 @@
+let _aigne_utils_nodejs = require("@aigne/utils/nodejs");
+
+//#region src/utils.ts
+/**
+* Wait for file size to stabilize, ensuring the file download is complete.
+*
+* @param filePath - The path to the file to check
+* @param options - Configuration options
+* @param options.checkInterval - Check interval in milliseconds (default: 500ms)
+* @param options.stableCount - Number of consecutive checks with same size to consider stable (default: 3)
+* @param options.timeout - Timeout in milliseconds (default: 60000ms)
+* @throws Error when timeout is reached
+*/
+async function waitFileSizeStable(filePath, options) {
+	const checkInterval = options?.checkInterval ?? 500;
+	const requiredStableCount = options?.stableCount ?? 3;
+	const timeout = options?.timeout ?? 6e4;
+	const startTime = Date.now();
+	let previousSize = 0;
+	let stableCount = 0;
+	while (stableCount < requiredStableCount) {
+		if (Date.now() - startTime > timeout) throw new Error(`Timeout waiting for file to stabilize: ${filePath}`);
+		await new Promise((resolve) => setTimeout(resolve, checkInterval));
+		const currentSize = (await _aigne_utils_nodejs.nodejs.fs.stat(filePath)).size;
+		if (currentSize === previousSize && currentSize > 0) stableCount++;
+		else {
+			stableCount = 0;
+			previousSize = currentSize;
+		}
+	}
+}
+
+//#endregion
+exports.waitFileSizeStable = waitFileSizeStable;

package/dist/utils.mjs

@@ -0,0 +1,35 @@
+import { nodejs } from "@aigne/utils/nodejs";
+
+//#region src/utils.ts
+/**
+* Wait for file size to stabilize, ensuring the file download is complete.
+*
+* @param filePath - The path to the file to check
+* @param options - Configuration options
+* @param options.checkInterval - Check interval in milliseconds (default: 500ms)
+* @param options.stableCount - Number of consecutive checks with same size to consider stable (default: 3)
+* @param options.timeout - Timeout in milliseconds (default: 60000ms)
+* @throws Error when timeout is reached
+*/
+async function waitFileSizeStable(filePath, options) {
+	const checkInterval = options?.checkInterval ?? 500;
+	const requiredStableCount = options?.stableCount ?? 3;
+	const timeout = options?.timeout ?? 6e4;
+	const startTime = Date.now();
+	let previousSize = 0;
+	let stableCount = 0;
+	while (stableCount < requiredStableCount) {
+		if (Date.now() - startTime > timeout) throw new Error(`Timeout waiting for file to stabilize: ${filePath}`);
+		await new Promise((resolve) => setTimeout(resolve, checkInterval));
+		const currentSize = (await nodejs.fs.stat(filePath)).size;
+		if (currentSize === previousSize && currentSize > 0) stableCount++;
+		else {
+			stableCount = 0;
+			previousSize = currentSize;
+		}
+	}
+}
+
+//#endregion
+export { waitFileSizeStable };
+//# sourceMappingURL=utils.mjs.map

package/dist/utils.mjs.map

@@ -0,0 +1 @@
+{"version":3,"file":"utils.mjs","names":[],"sources":["../src/utils.ts"],"sourcesContent":["import { nodejs } from \"@aigne/utils/nodejs\";\n\n/**\n * Wait for file size to stabilize, ensuring the file download is complete.\n *\n * @param filePath - The path to the file to check\n * @param options - Configuration options\n * @param options.checkInterval - Check interval in milliseconds (default: 500ms)\n * @param options.stableCount - Number of consecutive checks with same size to consider stable (default: 3)\n * @param options.timeout - Timeout in milliseconds (default: 60000ms)\n * @throws Error when timeout is reached\n */\nexport async function waitFileSizeStable(\n  filePath: string,\n  options?: {\n    checkInterval?: number;\n    stableCount?: number;\n    timeout?: number;\n  },\n): Promise<void> {\n  const checkInterval = options?.checkInterval ?? 500;\n  const requiredStableCount = options?.stableCount ?? 3;\n  const timeout = options?.timeout ?? 60000;\n\n  const startTime = Date.now();\n  let previousSize = 0;\n  let stableCount = 0;\n\n  while (stableCount < requiredStableCount) {\n    if (Date.now() - startTime > timeout) {\n      throw new Error(`Timeout waiting for file to stabilize: ${filePath}`);\n    }\n\n    await new Promise((resolve) => setTimeout(resolve, checkInterval));\n\n    const stats = await nodejs.fs.stat(filePath);\n    const currentSize = stats.size;\n\n    if (currentSize === previousSize && currentSize > 0) {\n      stableCount++;\n    } else {\n      stableCount = 0;\n      previousSize = currentSize;\n    }\n  }\n}\n"],"mappings":";;;;;;;;;;;;;AAYA,eAAsB,mBACpB,UACA,SAKe;CACf,MAAM,gBAAgB,SAAS,iBAAiB;CAChD,MAAM,sBAAsB,SAAS,eAAe;CACpD,MAAM,UAAU,SAAS,WAAW;CAEpC,MAAM,YAAY,KAAK,KAAK;CAC5B,IAAI,eAAe;CACnB,IAAI,cAAc;AAElB,QAAO,cAAc,qBAAqB;AACxC,MAAI,KAAK,KAAK,GAAG,YAAY,QAC3B,OAAM,IAAI,MAAM,0CAA0C,WAAW;AAGvE,QAAM,IAAI,SAAS,YAAY,WAAW,SAAS,cAAc,CAAC;EAGlE,MAAM,eADQ,MAAM,OAAO,GAAG,KAAK,SAAS,EAClB;AAE1B,MAAI,gBAAgB,gBAAgB,cAAc,EAChD;OACK;AACL,iBAAc;AACd,kBAAe"}

package/package.json

@@ -1,62 +1,61 @@
 {
   "name": "@aigne/gemini",
-  "version": "0.14.16",
+  "version": "1.74.0-beta",
   "description": "AIGNE Gemini SDK for integrating with Google's Gemini AI models",
+  "license": "Elastic-2.0",
   "publishConfig": {
     "access": "public"
   },
   "author": "Arcblock <blocklet@arcblock.io> https://github.com/blocklet",
   "homepage": "https://www.aigne.io/framework",
-  "license": "Elastic-2.0",
   "repository": {
     "type": "git",
-    "url": "git+https://github.com/AIGNE-io/aigne-framework"
+    "url": "git+https://github.com/ArcBlock/aigne-framework"
   },
   "bugs": {
-    "url": "https://github.com/AIGNE-io/aigne-framework/issues"
+    "url": "https://github.com/ArcBlock/aigne-framework/issues"
+  },
+  "type": "module",
+  "main": "./dist/index.cjs",
+  "module": "./dist/index.mjs",
+  "types": "./dist/index.d.cts",
+  "exports": {
+    ".": {
+      "require": "./dist/index.cjs",
+      "import": "./dist/index.mjs"
+    },
+    "./*": "./*"
   },
   "files": [
-    "lib/cjs",
-    "lib/dts",
-    "lib/esm",
+    "dist",
     "LICENSE",
     "README.md",
     "CHANGELOG.md"
   ],
-  "type": "module",
-  "main": "./lib/cjs/index.js",
-  "module": "./lib/esm/index.js",
-  "types": "./lib/dts/index.d.ts",
-  "exports": {
-    ".": {
-      "import": "./lib/esm/index.js",
-      "require": "./lib/cjs/index.js",
-      "types": "./lib/dts/index.d.ts"
-    }
-  },
   "dependencies": {
     "@aigne/uuid": "^13.0.1",
     "@google/genai": "^1.30.0",
     "yaml": "^2.8.1",
     "zod": "^3.25.67",
     "zod-to-json-schema": "^3.24.6",
-    "@aigne/core": "^1.72.0",
-    "@aigne/platform-helpers": "^0.6.7"
+    "@aigne/core": "^1.74.0-beta",
+    "@aigne/utils": "^1.74.0-beta"
   },
   "devDependencies": {
-    "@types/bun": "^1.2.22",
-    "@types/node": "^24.5.1",
+    "@types/bun": "^1.3.6",
     "npm-run-all": "^4.1.5",
-    "rimraf": "^6.0.1",
-    "typescript": "^5.9.2",
-    "@aigne/test-utils": "^0.5.69"
+    "rimraf": "^6.1.2",
+    "tsdown": "0.20.0-beta.3",
+    "typescript": "5.9.2",
+    "@aigne/scripts": "0.0.0",
+    "@aigne/typescript-config": "0.0.0",
+    "@aigne/utils": "1.74.0-beta"
   },
   "scripts": {
-    "lint": "tsc --noEmit",
-    "build": "tsc --build scripts/tsconfig.build.json",
-    "clean": "rimraf lib test/coverage",
+    "build": "tsdown",
+    "check-types": "tsc --noEmit",
+    "clean": "rimraf dist coverage",
     "test": "bun test",
-    "test:coverage": "bun test --coverage --coverage-reporter=lcov --coverage-reporter=text",
-    "postbuild": "node ../../scripts/post-build-lib.mjs"
+    "test:coverage": "bun test --coverage --coverage-reporter=lcov --coverage-reporter=text"
   }
 }