chuan/WebAI2API
1
0
Fork 0
mirror of https://github.com/foxhui/WebAI2API.git synced 2026-06-16 21:03:59 +08:00
Code Issues Packages Projects Releases Wiki Activity

feat: 支持 Google Flow 图片生成

Browse Source
This commit is contained in:
foxhui
2025-12-25 04:00:34 +08:00
Unverified
parent f67eaa8cd5
commit 069d699b99
3 changed files with 273 additions and 0 deletions
Download Patch File Download Diff File Expand all files Collapse all files
CHANGELOG.md
+4
View File
@@ -7,6 +7,10 @@ and this project adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0
## [3.4.1] - 2025-12-24
### ✨ Added
- **新增适配器**
- 支持 Google Flow 图片生成适配器
### 🐛 Fixed
- **Gemini Business**：修复因懒加载导致的等待图片超时问题
README.md
+1
View File
@@ -44,6 +44,7 @@
| [**ChatGPT**](https://chatgpt.com/) | ✅ | ✅ | 🚫 |
| [**DeepSeek**](https://chat.deepseek.com/) | ✅ | 🚫 | 🚫 |
| [**Sora**](https://sora.chatgpt.com/) | 🚫 | 🚫 | ✅ |
| [**Google Flow**](https://labs.google/fx/zh/tools/flow) | 🚫 | ✅ | ❌ |
| 待续... | - | - | - |
> [!NOTE]
src/backend/adapter/google_flow.js
+268
View File
@@ -0,0 +1,268 @@
/**
* @fileoverview Google Flow 图片生成适配器
*/
import {
sleep,
safeClick,
uploadFilesViaChooser
} from '../engine/utils.js';
import {
normalizePageError,
moveMouseAway,
waitForInput,
gotoWithCheck,
waitApiResponse,
useContextDownload
} from '../utils/index.js';
import { logger } from '../../utils/logger.js';
// --- 配置常量 ---
const TARGET_URL = 'https://labs.google/fx/zh/tools/flow';
/**
* 执行图片生成任务
* @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
* @param {object} [meta={}] - 日志元数据
* @returns {Promise<{image?: string, error?: string}>}
*/
async function generate(context, prompt, imgPaths, modelId, meta = {}) {
const { page } = context;
// 获取模型配置
const modelConfig = manifest.models.find(m => m.id === modelId) || manifest.models[0];
const { codeName, imageSize } = modelConfig;
try {
// 1. 导航到入口页面
logger.info('适配器', '开启新会话...', meta);
await gotoWithCheck(page, TARGET_URL);
await sleep(1500, 2500);
// 2. 创建项目 - 点击 add_2 按钮
logger.debug('适配器', '创建新项目...', meta);
const addProjectBtn = page.getByRole('button', { name: /^add_2/ });
await addProjectBtn.waitFor({ state: 'visible', timeout: 30000 });
await safeClick(page, addProjectBtn, { bias: 'button' });
await sleep(1000, 1500);
// 3. 选择 Images 模式
logger.debug('适配器', '选择 Images 模式...', meta);
const imageRadio = page.getByRole('radio', { name: 'image Images' });
await imageRadio.waitFor({ state: 'visible', timeout: 10000 });
await safeClick(page, imageRadio, { bias: 'button' });
await sleep(1000, 1500);
// 4. 打开 Tune 菜单进行配置
logger.debug('适配器', '打开设置菜单...', meta);
const tuneBtn = page.getByRole('button', { name: /^tune/ });
await tuneBtn.waitFor({ state: 'visible', timeout: 10000 });
await safeClick(page, tuneBtn, { bias: 'button' });
await sleep(800, 1200);
// 获取所有 combobox
const allComboboxes = page.getByRole('combobox');
const comboboxCount = await allComboboxes.count();
logger.debug('适配器', `找到 ${comboboxCount} 个 combobox`, meta);
// 4.1 设置生成数量为 1 (排除法：找到不包含模型/尺寸关键词但包含数字1-4的 combobox)
logger.debug('适配器', '设置生成数量为 1...', meta);
let countFound = false;
for (let i = 0; i < comboboxCount; i++) {
const combobox = allComboboxes.nth(i);
const fullText = await combobox.textContent().catch(() => '');
logger.debug('适配器', `combobox[${i}] 完整内容: "${fullText}"`, meta);
// 排除模型和尺寸选择器，找到包含数字1-4但不包含其他关键词的
const isNotModel = !/Banana|Imagen/i.test(fullText);
const isNotSize = !/16:9|9:16|1:1|4:3|3:4/.test(fullText);
const hasNumber = /[1-4]/.test(fullText);
if (isNotModel && isNotSize && hasNumber) {
await safeClick(page, combobox, { bias: 'button' });
await sleep(300, 500);
await safeClick(page, page.getByRole('option', { name: '1' }), { bias: 'button' });
await sleep(300, 500);
logger.debug('适配器', '生成数量已设置为 1', meta);
countFound = true;
break;
}
}
if (!countFound) {
logger.warn('适配器', '未找到数量选择 combobox，跳过', meta);
}
// 4.2 选择模型 (查找包含模型名称的 combobox)
logger.debug('适配器', `选择模型: ${codeName}...`, meta);
for (let i = 0; i < comboboxCount; i++) {
const combobox = allComboboxes.nth(i);
const text = await combobox.textContent().catch(() => '');
if (/Nano Banana|Imagen 4/.test(text)) {
await safeClick(page, combobox, { bias: 'button' });
await sleep(300, 500);
await safeClick(page, page.getByRole('option', { name: codeName }), { bias: 'button' });
await sleep(300, 500);
logger.debug('适配器', `模型已设置为 ${codeName}`, meta);
break;
}
}
// 4.3 选择横竖版 (查找包含 16:9 或 9:16 的 combobox)
logger.debug('适配器', `选择尺寸: ${imageSize}...`, meta);
for (let i = 0; i < comboboxCount; i++) {
const combobox = allComboboxes.nth(i);
const text = await combobox.textContent().catch(() => '');
if (/16:9|9:16/.test(text)) {
await safeClick(page, combobox, { bias: 'button' });
await sleep(300, 500);
// 使用包含匹配，因为 option 名字中可能包含 16:9 或 9:16
const sizeOption = page.getByRole('option').filter({ hasText: imageSize });
await safeClick(page, sizeOption.first(), { bias: 'button' });
await sleep(300, 500);
logger.debug('适配器', `尺寸已设置为 ${imageSize}`, meta);
break;
}
}
// 关闭 Tune 菜单 (再次点击 tune 按钮)
await safeClick(page, tuneBtn, { bias: 'button' });
await sleep(500, 1000);
// 5. 上传图片 (如果有)
if (imgPaths && imgPaths.length > 0) {
logger.info('适配器', `开始上传 ${imgPaths.length} 张图片...`, meta);
for (let i = 0; i < imgPaths.length; i++) {
const imgPath = imgPaths[i];
logger.debug('适配器', `上传图片 ${i + 1}/${imgPaths.length}...`, meta);
// 5.1 点击 add 按钮
const addBtn = page.getByRole('button', { name: 'add' });
await addBtn.waitFor({ state: 'visible', timeout: 10000 });
await safeClick(page, addBtn, { bias: 'button' });
await sleep(500, 1000);
// 5.2 点击 upload 按钮并选择文件（不等待上传完成）
const uploadBtn = page.getByRole('button', { name: /^upload/ });
await uploadFilesViaChooser(page, uploadBtn, [imgPath]);
await sleep(500, 1000);
// 5.3 先启动上传监听，再点击 crop 按钮
const uploadResponsePromise = waitApiResponse(page, {
urlMatch: 'v1:uploadUserImage',
method: 'POST',
timeout: 60000
});
const cropBtn = page.getByRole('button', { name: /^crop/ });
await cropBtn.waitFor({ state: 'visible', timeout: 10000 });
await safeClick(page, cropBtn, { bias: 'button' });
// 5.4 等待上传完成
await uploadResponsePromise;
logger.info('适配器', `图片 ${i + 1} 上传完成`, meta);
await sleep(1000, 1500);
}
logger.info('适配器', '所有图片上传完成', meta);
}
// 6. 输入提示词
logger.info('适配器', '输入提示词...', meta);
const textarea = page.locator('textarea[placeholder]');
await waitForInput(page, textarea, { click: true });
await textarea.fill(prompt);
await sleep(500, 1000);
// 7. 先启动 API 监听，再点击发送
logger.debug('适配器', '启动 API 监听...', meta);
const apiResponsePromise = waitApiResponse(page, {
urlMatch: 'flowMedia:batchGenerateImages',
method: 'POST',
timeout: 120000,
meta
});
// 8. 点击发送按钮
logger.info('适配器', '点击发送...', meta);
const sendBtn = page.getByRole('button', { name: /^arrow_forward/ });
await sendBtn.waitFor({ state: 'visible', timeout: 10000 });
await safeClick(page, sendBtn, { bias: 'button' });
// 9. 等待 API 响应
logger.info('适配器', '等待生成结果...', meta);
const apiResponse = await apiResponsePromise;
// 10. 解析响应获取图片 URL
let imageUrl;
try {
const responseBody = await apiResponse.json();
imageUrl = responseBody?.media?.[0]?.image?.generatedImage?.fifeUrl;
if (!imageUrl) {
logger.error('适配器', '响应中没有图片 URL', meta);
return { error: '生成成功但响应中没有图片 URL' };
}
logger.info('适配器', '已获取图片链接', meta);
} catch (e) {
logger.error('适配器', '解析响应失败', { ...meta, error: e.message });
return { error: `解析响应失败: ${e.message}` };
}
// 11. 下载图片并转为 base64
logger.info('适配器', '正在下载图片...', meta);
const downloadResult = await useContextDownload(imageUrl, page);
if (downloadResult.error) {
logger.error('适配器', downloadResult.error, meta);
return downloadResult;
}
logger.info('适配器', '图片生成完成', meta);
return { image: downloadResult.image };
} 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);
}
}
/**
* 适配器 manifest
*/
export const manifest = {
id: 'google_flow',
displayName: 'Google Flow (图片生成)',
// 入口 URL
getTargetUrl(config, workerConfig) {
return TARGET_URL;
},
// 模型列表
models: [
{ id: 'gemini-3-pro-image-preview-landspace', codeName: '🍌 Nano Banana Pro', imageSize: '16:9', imagePolicy: 'optional' },
{ id: 'gemini-3-pro-image-preview-portrait', codeName: '🍌 Nano Banana Pro', imageSize: '9:16', imagePolicy: 'optional' },
{ id: 'gemini-2.5-flash-image-preview-landspace', codeName: '🍌 Nano Banana', imageSize: '16:9', imagePolicy: 'optional' },
{ id: 'gemini-2.5-flash-image-preview-portrait', codeName: '🍌 Nano Banana', imageSize: '9:16', imagePolicy: 'optional' },
{ id: 'imagen-4-landspace', codeName: 'Imagen 4', imageSize: '16:9', imagePolicy: 'optional' },
{ id: 'imagen-4-portrait', codeName: 'Imagen 4', imageSize: '9:16', imagePolicy: 'optional' }
],
// 无需导航处理器
navigationHandlers: [],
// 核心图片生成方法
generate
};