@aigne/gemini 0.9.9 → 0.10.0

package/CHANGELOG.md

@@ -1,5 +1,32 @@
 # Changelog
 
+## [0.10.0](https://github.com/AIGNE-io/aigne-framework/compare/gemini-v0.9.10...gemini-v0.10.0) (2025-08-27)
+
+
+### Features
+
+* **models:** support gemini and ideogram images models ([#412](https://github.com/AIGNE-io/aigne-framework/issues/412)) ([6534fec](https://github.com/AIGNE-io/aigne-framework/commit/6534fecb0bdfb4b0a4440d44c0e563b9a029a68f))
+* **models:** support gemini and ideogram images models ([#412](https://github.com/AIGNE-io/aigne-framework/issues/412)) ([6534fec](https://github.com/AIGNE-io/aigne-framework/commit/6534fecb0bdfb4b0a4440d44c0e563b9a029a68f))
+
+
+### Dependencies
+
+* The following workspace dependencies were updated
+  * dependencies
+    * @aigne/openai bumped to 0.13.0
+
+## [0.9.10](https://github.com/AIGNE-io/aigne-framework/compare/gemini-v0.9.9...gemini-v0.9.10) (2025-08-26)
+
+
+### Dependencies
+
+* The following workspace dependencies were updated
+  * dependencies
+    * @aigne/openai bumped to 0.12.4
+  * devDependencies
+    * @aigne/core bumped to 1.55.1
+    * @aigne/test-utils bumped to 0.5.36
+
 ## [0.9.9](https://github.com/AIGNE-io/aigne-framework/compare/gemini-v0.9.8...gemini-v0.9.9) (2025-08-25)
 
 

package/lib/cjs/gemini-image-model.d.ts

@@ -0,0 +1,53 @@
+import { ImageModel, type ImageModelInput, type ImageModelOptions, type ImageModelOutput } from "@aigne/core";
+import { type GenerateImagesConfig, GoogleGenAI } from "@google/genai";
+export interface GeminiImageModelInput extends ImageModelInput, GenerateImagesConfig {
+}
+export interface GeminiImageModelOutput extends ImageModelOutput {
+}
+export interface GeminiImageModelOptions extends ImageModelOptions<GeminiImageModelInput, GeminiImageModelOutput> {
+    /**
+     * 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 'gemini-2.0-flash'
+     */
+    model?: string;
+    /**
+     * Additional model options to control behavior
+     */
+    modelOptions?: Omit<Partial<GeminiImageModelInput>, "model">;
+    /**
+     * Client options for Gemini API
+     */
+    clientOptions?: Record<string, any>;
+}
+export declare class GeminiImageModel extends ImageModel<GeminiImageModelInput, GeminiImageModelOutput> {
+    options?: GeminiImageModelOptions | undefined;
+    constructor(options?: GeminiImageModelOptions | undefined);
+    protected _client?: GoogleGenAI;
+    protected apiKeyEnvName: string;
+    get client(): GoogleGenAI;
+    get credential(): {
+        url: string | undefined;
+        apiKey: string | undefined;
+        model: string;
+    };
+    get modelOptions(): Omit<Partial<GeminiImageModelInput>, "model"> | undefined;
+    /**
+     * Process the input and generate a response
+     * @param input The input to process
+     * @returns The generated response
+     */
+    process(input: GeminiImageModelInput): Promise<ImageModelOutput>;
+}

package/lib/cjs/gemini-image-model.js

@@ -0,0 +1,95 @@
+"use strict";
+Object.defineProperty(exports, "__esModule", { value: true });
+exports.GeminiImageModel = void 0;
+const core_1 = require("@aigne/core");
+const type_utils_js_1 = require("@aigne/core/utils/type-utils.js");
+const genai_1 = require("@google/genai");
+const zod_1 = require("zod");
+const DEFAULT_MODEL = "imagen-4.0-generate-001";
+const geminiImageModelInputSchema = core_1.imageModelInputSchema.extend({});
+const geminiImageModelOptionsSchema = zod_1.z.object({
+    apiKey: zod_1.z.string().optional(),
+    baseURL: zod_1.z.string().optional(),
+    model: zod_1.z.string().optional(),
+    modelOptions: zod_1.z.object({}).optional(),
+    clientOptions: zod_1.z.object({}).optional(),
+});
+class GeminiImageModel extends core_1.ImageModel {
+    options;
+    constructor(options) {
+        super({
+            ...options,
+            inputSchema: geminiImageModelInputSchema,
+            description: options?.description ?? "Draw or edit image by Gemini image models",
+        });
+        this.options = options;
+        if (options)
+            (0, type_utils_js_1.checkArguments)(this.name, geminiImageModelOptionsSchema, options);
+    }
+    _client;
+    apiKeyEnvName = "GEMINI_API_KEY";
+    get client() {
+        if (this._client)
+            return this._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 genai_1.GoogleGenAI({ apiKey });
+        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;
+    }
+    /**
+     * Process the input and generate a response
+     * @param input The input to process
+     * @returns The generated response
+     */
+    async process(input) {
+        const model = input.model || this.credential.model;
+        const responseFormat = input.responseFormat || "base64";
+        if (responseFormat === "url") {
+            throw new Error("Gemini image models currently only support base64 format");
+        }
+        const mergedInput = { ...this.modelOptions, ...input };
+        const inputKeys = [
+            "seed",
+            "safetyFilterLevel",
+            "personGeneration",
+            "outputMimeType",
+            "outputGcsUri",
+            "outputCompressionQuality",
+            "negativePrompt",
+            "language",
+            "includeSafetyAttributes",
+            "includeRaiReason",
+            "imageSize",
+            "guidanceScale",
+            "aspectRatio",
+            "addWatermark",
+        ];
+        const response = await this.client.models.generateImages({
+            model: model,
+            prompt: mergedInput.prompt,
+            config: { numberOfImages: mergedInput.n || 1, ...(0, type_utils_js_1.pick)(mergedInput, inputKeys) },
+        });
+        return {
+            images: response.generatedImages
+                ?.filter((image) => image.image?.imageBytes)
+                .map((image) => ({ base64: image.image?.imageBytes })) || [],
+            usage: {
+                inputTokens: 0,
+                outputTokens: 0,
+            },
+            model,
+        };
+    }
+}
+exports.GeminiImageModel = GeminiImageModel;

package/lib/cjs/index.d.ts

@@ -1 +1,2 @@
 export * from "./gemini-chat-model.js";
+export * from "./gemini-image-model.js";

package/lib/cjs/index.js

@@ -15,3 +15,4 @@ var __exportStar = (this && this.__exportStar) || function(m, exports) {
 };
 Object.defineProperty(exports, "__esModule", { value: true });
 __exportStar(require("./gemini-chat-model.js"), exports);
+__exportStar(require("./gemini-image-model.js"), exports);

package/lib/dts/gemini-image-model.d.ts

@@ -0,0 +1,53 @@
+import { ImageModel, type ImageModelInput, type ImageModelOptions, type ImageModelOutput } from "@aigne/core";
+import { type GenerateImagesConfig, GoogleGenAI } from "@google/genai";
+export interface GeminiImageModelInput extends ImageModelInput, GenerateImagesConfig {
+}
+export interface GeminiImageModelOutput extends ImageModelOutput {
+}
+export interface GeminiImageModelOptions extends ImageModelOptions<GeminiImageModelInput, GeminiImageModelOutput> {
+    /**
+     * 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 'gemini-2.0-flash'
+     */
+    model?: string;
+    /**
+     * Additional model options to control behavior
+     */
+    modelOptions?: Omit<Partial<GeminiImageModelInput>, "model">;
+    /**
+     * Client options for Gemini API
+     */
+    clientOptions?: Record<string, any>;
+}
+export declare class GeminiImageModel extends ImageModel<GeminiImageModelInput, GeminiImageModelOutput> {
+    options?: GeminiImageModelOptions | undefined;
+    constructor(options?: GeminiImageModelOptions | undefined);
+    protected _client?: GoogleGenAI;
+    protected apiKeyEnvName: string;
+    get client(): GoogleGenAI;
+    get credential(): {
+        url: string | undefined;
+        apiKey: string | undefined;
+        model: string;
+    };
+    get modelOptions(): Omit<Partial<GeminiImageModelInput>, "model"> | undefined;
+    /**
+     * Process the input and generate a response
+     * @param input The input to process
+     * @returns The generated response
+     */
+    process(input: GeminiImageModelInput): Promise<ImageModelOutput>;
+}

package/lib/dts/index.d.ts

@@ -1 +1,2 @@
 export * from "./gemini-chat-model.js";
+export * from "./gemini-image-model.js";

package/lib/esm/gemini-image-model.d.ts

@@ -0,0 +1,53 @@
+import { ImageModel, type ImageModelInput, type ImageModelOptions, type ImageModelOutput } from "@aigne/core";
+import { type GenerateImagesConfig, GoogleGenAI } from "@google/genai";
+export interface GeminiImageModelInput extends ImageModelInput, GenerateImagesConfig {
+}
+export interface GeminiImageModelOutput extends ImageModelOutput {
+}
+export interface GeminiImageModelOptions extends ImageModelOptions<GeminiImageModelInput, GeminiImageModelOutput> {
+    /**
+     * 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 'gemini-2.0-flash'
+     */
+    model?: string;
+    /**
+     * Additional model options to control behavior
+     */
+    modelOptions?: Omit<Partial<GeminiImageModelInput>, "model">;
+    /**
+     * Client options for Gemini API
+     */
+    clientOptions?: Record<string, any>;
+}
+export declare class GeminiImageModel extends ImageModel<GeminiImageModelInput, GeminiImageModelOutput> {
+    options?: GeminiImageModelOptions | undefined;
+    constructor(options?: GeminiImageModelOptions | undefined);
+    protected _client?: GoogleGenAI;
+    protected apiKeyEnvName: string;
+    get client(): GoogleGenAI;
+    get credential(): {
+        url: string | undefined;
+        apiKey: string | undefined;
+        model: string;
+    };
+    get modelOptions(): Omit<Partial<GeminiImageModelInput>, "model"> | undefined;
+    /**
+     * Process the input and generate a response
+     * @param input The input to process
+     * @returns The generated response
+     */
+    process(input: GeminiImageModelInput): Promise<ImageModelOutput>;
+}

package/lib/esm/gemini-image-model.js

@@ -0,0 +1,91 @@
+import { ImageModel, imageModelInputSchema, } from "@aigne/core";
+import { checkArguments, pick } from "@aigne/core/utils/type-utils.js";
+import { GoogleGenAI } from "@google/genai";
+import { z } from "zod";
+const DEFAULT_MODEL = "imagen-4.0-generate-001";
+const geminiImageModelInputSchema = imageModelInputSchema.extend({});
+const geminiImageModelOptionsSchema = z.object({
+    apiKey: z.string().optional(),
+    baseURL: z.string().optional(),
+    model: z.string().optional(),
+    modelOptions: z.object({}).optional(),
+    clientOptions: z.object({}).optional(),
+});
+export class GeminiImageModel extends ImageModel {
+    options;
+    constructor(options) {
+        super({
+            ...options,
+            inputSchema: geminiImageModelInputSchema,
+            description: options?.description ?? "Draw or edit image by Gemini image models",
+        });
+        this.options = options;
+        if (options)
+            checkArguments(this.name, geminiImageModelOptionsSchema, options);
+    }
+    _client;
+    apiKeyEnvName = "GEMINI_API_KEY";
+    get client() {
+        if (this._client)
+            return this._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 });
+        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;
+    }
+    /**
+     * Process the input and generate a response
+     * @param input The input to process
+     * @returns The generated response
+     */
+    async process(input) {
+        const model = input.model || this.credential.model;
+        const responseFormat = input.responseFormat || "base64";
+        if (responseFormat === "url") {
+            throw new Error("Gemini image models currently only support base64 format");
+        }
+        const mergedInput = { ...this.modelOptions, ...input };
+        const inputKeys = [
+            "seed",
+            "safetyFilterLevel",
+            "personGeneration",
+            "outputMimeType",
+            "outputGcsUri",
+            "outputCompressionQuality",
+            "negativePrompt",
+            "language",
+            "includeSafetyAttributes",
+            "includeRaiReason",
+            "imageSize",
+            "guidanceScale",
+            "aspectRatio",
+            "addWatermark",
+        ];
+        const response = await this.client.models.generateImages({
+            model: model,
+            prompt: mergedInput.prompt,
+            config: { numberOfImages: mergedInput.n || 1, ...pick(mergedInput, inputKeys) },
+        });
+        return {
+            images: response.generatedImages
+                ?.filter((image) => image.image?.imageBytes)
+                .map((image) => ({ base64: image.image?.imageBytes })) || [],
+            usage: {
+                inputTokens: 0,
+                outputTokens: 0,
+            },
+            model,
+        };
+    }
+}

package/lib/esm/index.d.ts

@@ -1 +1,2 @@
 export * from "./gemini-chat-model.js";
+export * from "./gemini-image-model.js";

package/lib/esm/index.js

@@ -1 +1,2 @@
 export * from "./gemini-chat-model.js";
+export * from "./gemini-image-model.js";

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@aigne/gemini",
-  "version": "0.9.9",
+  "version": "0.10.0",
   "description": "AIGNE Gemini SDK for integrating with Google's Gemini AI models",
   "publishConfig": {
     "access": "public"
@@ -35,7 +35,9 @@
     }
   },
   "dependencies": {
-    "@aigne/openai": "^0.12.3"
+    "@google/genai": "^1.15.0",
+    "zod": "^3.25.67",
+    "@aigne/openai": "^0.13.0"
   },
   "devDependencies": {
     "@types/bun": "^1.2.18",
@@ -43,8 +45,8 @@
     "npm-run-all": "^4.1.5",
     "rimraf": "^6.0.1",
     "typescript": "^5.8.3",
-    "@aigne/test-utils": "^0.5.35",
-    "@aigne/core": "^1.55.0"
+    "@aigne/core": "^1.55.1",
+    "@aigne/test-utils": "^0.5.36"
   },
   "scripts": {
     "lint": "tsc --noEmit",