feat: 初步支持 Gemini 网页版

CHANGELOG.md

@@ -5,6 +5,12 @@ All notable changes to this project will be documented in this file.
 The format is based on [Keep a Changelog](https://keepachangelog.com/en/1.0.0/),
 and this project adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0.html).
 
+## [2.3.0] - 2025-12-12
+
+### Added
+- **支持新网站**
+  - 初步支持对 Gemini 网页版的支持
+
 ## [2.2.3] - 2025-12-12
 
 ### Added

README.md

@@ -10,6 +10,7 @@ LMArenaImagenAutomator 是一个基于 Playwright + Camoufox 的自动化图像
   - [Gemini Enterprise Business](https://business.gemini.google/)
   - [Nano Banana Free](https://nanobananafree.ai/)
   - [zAI](https://zai.is/)
+  - [Google Gemini](https://gemini.google.com/)
   - 未来可能支持更多网站。。。
 
 ### ✨ 主要特性

config.example.yaml

@@ -21,6 +21,7 @@ server:
 backend:
   # 适配器设置
   # - lmarena (LMArena)
+  # - gemini (Gemini 网页版)
   # - gemini_biz (Gemini Enterprise Business)
   # - nanobananafree_ai (Nano Banana Free)
   # - zai_is (zAI)

lib/backend/adapter/gemini.js

@@ -0,0 +1,185 @@
+import { initBrowserBase } from '../../browser/launcher.js';
+import {
+    sleep,
+    safeClick,
+    pasteImages
+} from '../../browser/utils.js';
+import {
+    fillPrompt,
+    normalizePageError,
+    moveMouseAway
+} from '../utils.js';
+import { logger } from '../../utils/logger.js';
+
+// --- 配置常量 ---
+const TARGET_URL = 'https://gemini.google.com/app?hl=en';
+
+/**
+ * 初始化浏览器会话
+ * @param {object} config - 全局配置对象
+ * @returns {Promise<{browser: object, page: object, config: object}>}
+ */
+async function initBrowser(config) {
+    // 输入框验证逻辑
+    const waitInputValidator = async (page) => {
+        await page.getByRole('textbox').waitFor({ timeout: 60000 });
+        await safeClick(page, page.getByRole('textbox'), { bias: 'input' });
+        await sleep(500, 1000);
+    };
+
+    const base = await initBrowserBase(config, {
+        userDataDir: config.paths.userDataDir,
+        targetUrl: TARGET_URL,
+        productName: 'Gemini',
+        waitInputValidator
+    });
+    return { ...base, config };
+}
+
+/**
+ * 执行生图任务
+ * @param {object} context - 浏览器上下文 { page, config }
+ * @param {string} prompt - 提示词
+ * @param {string[]} imgPaths - 图片路径数组
+ * @param {string} [modelId] - 模型 ID (此适配器未使用)
+ * @param {object} [meta={}] - 日志元数据
+ * @returns {Promise<{image?: string, error?: string}>}
+ */
+async function generateImage(context, prompt, imgPaths, modelId, meta = {}) {
+    const { page } = context;
+    const inputLocator = page.getByRole('textbox');
+    const sendBtnLocator = page.getByRole('button', { name: 'Send message' });
+
+    try {
+        logger.info('适配器', '开启新会话...', meta);
+        await page.goto(TARGET_URL, { waitUntil: 'domcontentloaded' });
+
+        // 1. 等待输入框加载
+        await inputLocator.waitFor({ timeout: 30000 });
+        await sleep(1500, 2500);
+
+        // 2. 上传图片
+        if (imgPaths && imgPaths.length > 0) {
+            const expectedUploads = imgPaths.length;
+            let uploadedCount = 0;
+
+            await pasteImages(page, inputLocator, imgPaths, {
+                uploadValidator: (response) => {
+                    const url = response.url();
+                    // 检测上传成功：google.com/upload/?upload_id= 的 POST 请求
+                    if (response.status() === 200 &&
+                        url.includes('google.com/upload/') &&
+                        url.includes('upload_id=')) {
+                        uploadedCount++;
+                        logger.info('适配器', `图片上传进度: ${uploadedCount}/${expectedUploads}`, meta);
+                        return uploadedCount >= expectedUploads;
+                    }
+                    return false;
+                }
+            });
+
+            await sleep(1000, 2000);
+        }
+
+        // 3. 填写提示词
+        await safeClick(page, inputLocator, { bias: 'input' });
+        await fillPrompt(page, inputLocator, prompt, meta);
+        await sleep(500, 1000);
+
+        // 4. 点击 Tools 按钮启用图片生成
+        logger.debug('适配器', '点击 Tools 按钮...', meta);
+        const toolsBtn = page.getByRole('button', { name: 'Tools' });
+        await safeClick(page, toolsBtn, { bias: 'button' });
+        await sleep(500, 1000);
+
+        // 5. 点击 Create images 按钮
+        logger.debug('适配器', '点击 Create images 按钮...', meta);
+        const createImagesBtn = page.getByRole('button', { name: 'Create images' });
+        await safeClick(page, createImagesBtn, { bias: 'button' });
+        await sleep(500, 1000);
+
+        // 6. 设置响应监听 - 等待 StreamGenerate 成功后捕获图片
+        let imageData = null;
+
+        const imagePromise = new Promise((resolve, reject) => {
+            const timeout = setTimeout(() => {
+                cleanup();
+                reject(new Error('等待图片响应超时 (120秒)'));
+            }, 120000);
+
+            let streamGenerateSuccess = false;
+
+            const onResponse = async (response) => {
+                const url = response.url();
+
+                // 先等待 StreamGenerate 成功
+                if (!streamGenerateSuccess &&
+                    url.includes('assistant.lamda.BardFrontendService/StreamGenerate') &&
+                    response.request().method() === 'POST' &&
+                    response.status() === 200) {
+                    streamGenerateSuccess = true;
+                    logger.info('适配器', '生成请求成功，等待图片...', meta);
+                }
+
+                // StreamGenerate 成功后，捕获图片响应
+                if (streamGenerateSuccess &&
+                    url.includes('googleusercontent.com/rd-gg-dl') &&
+                    url.includes('=s1024-rj') &&
+                    response.request().method() === 'GET' &&
+                    response.status() === 200) {
+                    try {
+                        // 直接获取图片二进制数据
+                        const buffer = await response.body();
+                        const base64 = buffer.toString('base64');
+
+                        // 根据 Content-Type 确定图片格式
+                        const contentType = response.headers()['content-type'] || 'image/jpeg';
+                        imageData = `data:${contentType};base64,${base64}`;
+
+                        logger.info('适配器', '已捕获图片数据', meta);
+                        cleanup();
+                        resolve(imageData);
+                    } catch (e) {
+                        logger.warn('适配器', `捕获图片失败: ${e.message}`, meta);
+                    }
+                }
+            };
+
+            const cleanup = () => {
+                clearTimeout(timeout);
+                page.off('response', onResponse);
+            };
+
+            page.on('response', onResponse);
+        });
+
+        // 7. 点击发送
+        logger.debug('适配器', '点击发送...', meta);
+        await safeClick(page, sendBtnLocator, { bias: 'button' });
+
+        logger.info('适配器', '等待生成结果...', meta);
+
+        // 7. 等待图片响应
+        const image = await imagePromise;
+
+        if (image) {
+            logger.info('适配器', '已获取图片，任务完成', meta);
+            return { image };
+        } else {
+            return { error: '未能获取图片' };
+        }
+
+    } catch (err) {
+        // 顶层错误处理
+        const pageError = normalizePageError(err, meta);
+        if (pageError) return pageError;
+
+        logger.error('适配器', '生成任务失败', { ...meta, error: err.message });
+        return { error: `生成任务失败: ${err.message}` };
+    } finally {
+        // 任务结束，将鼠标移至安全区域
+        await moveMouseAway(page);
+    }
+}
+
+export { initBrowser, generateImage };

lib/backend/index.js

@@ -8,6 +8,7 @@ import { logger } from '../utils/logger.js';
 // 导入适配器
 import * as lmarenaBackend from './adapter/lmarena.js';
 import * as geminiBackend from './adapter/gemini_biz.js';
+import * as geminiConsumerBackend from './adapter/gemini.js';
 import * as nanobananafreeBackend from './adapter/nanobananafree_ai.js';
 import * as zaiIsBackend from './adapter/zai_is.js';
 
@@ -34,6 +35,7 @@ config.paths = {
 // 适配器映射表
 const ADAPTER_MAP = {
     'gemini_biz': geminiBackend,
+    'gemini': geminiConsumerBackend,
     'nanobananafree_ai': nanobananafreeBackend,
     'zai_is': zaiIsBackend,
     'lmarena': lmarenaBackend

lib/backend/models.js

@@ -165,6 +165,13 @@ export const ZAI_IS_MODELS = {
     }
 };
 
+// Gemini 后端模型配置
+export const GEMINI_MODELS = {
+    "gemini-3-pro-image-preview": {
+        imagePolicy: IMAGE_POLICY.OPTIONAL
+    }
+};
+
 /**
  * 获取后端对应的模型配置表
  * @param {string} backendName - 后端名称 ('lmarena' 或 'gemini_biz' 或 'nanobananafree_ai')
@@ -177,13 +184,12 @@ function getModelsConfigForBackend(backendName) {
             return LMARENA_MODELS;
         case 'gemini_biz':
             return GEMINI_BIZ_MODELS;
+        case 'gemini':
+            return GEMINI_MODELS;
         case 'nanobananafree_ai':
             return NANOBANANAFREE_AI_MODELS;
         case 'zai_is':
             return ZAI_IS_MODELS;
-        // 将来新增其它后端：
-        // case 'foo_site':
-        //   return FOO_SITE_MODELS;
         default:
             return {};
     }

lib/browser/utils.js

@@ -131,9 +131,9 @@ export function getHumanClickPoint(box, type = 'random') {
 
 /**
  * 安全点击元素 (包含拟人化移动和点击)
- * 支持 CSS selector 和 ElementHandle 两种输入
+ * 支持 CSS selector、ElementHandle 和 Locator 三种输入
  * @param {import('playwright-core').Page} page - Playwright 页面对象
- * @param {string|import('playwright-core').ElementHandle} target - CSS 选择器或元素句柄
+ * @param {string|import('playwright-core').ElementHandle|import('playwright-core').Locator} target - CSS 选择器、元素句柄或 Locator
  * @param {object} [options] - 点击选项
  * @param {string} [options.bias='random'] - 偏移偏好: 'input' 或 'random'
  * @returns {Promise<void>}
@@ -142,11 +142,17 @@ export async function safeClick(page, target, options = {}) {
     try {
         let el;
 
-        // 判断是 selector 还是 ElementHandle
+        // 判断输入类型
         if (typeof target === 'string') {
+            // CSS selector
             el = await page.$(target);
             if (!el) throw new Error(`未找到: ${target}`);
+        } else if (typeof target.elementHandle === 'function') {
+            // Locator (来自 page.getByRole, page.getByText 等)
+            el = await target.elementHandle();
+            if (!el) throw new Error(`Locator 未匹配到元素`);
         } else {
+            // ElementHandle
             el = target;
             if (!el || !el.asElement()) throw new Error(`Element handle invalid`);
         }

lib/utils/config.js

@@ -47,6 +47,7 @@ server:
 backend:
   # 适配器设置
   # - lmarena (LMArena)
+  # - gemini (Gemini 网页版)
   # - gemini_biz (Gemini Enterprise Business)
   # - nanobananafree_ai (Nano Banana Free)
   # - zai_is (zAI)