add: gemini-web-generate skill（整合 CLI + skill）
This commit is contained in:
@@ -0,0 +1,557 @@
|
||||
#!/usr/bin/env node
|
||||
import { connectBrowser, navigateToGemini, log, checkLogin, listSessions, closeSession, captureChatUrl, findSessionByChatUrl } from './src/browser.js';
|
||||
import { generate } from './src/generator.js';
|
||||
import { downloadViaButtons, takeScreenshot } from './src/screenshot.js';
|
||||
import { checkStatus } from './src/status.js';
|
||||
import { config } from './config.js';
|
||||
import fs from 'fs';
|
||||
import path from 'path';
|
||||
|
||||
const sleep = (ms) => new Promise(r => setTimeout(r, ms));
|
||||
|
||||
/**
|
||||
* NDJSON event emitter.
|
||||
* In JSON mode, emits structured progress events to stdout as newline-delimited JSON.
|
||||
* Human-readable logs always go to stderr.
|
||||
*/
|
||||
|
||||
let jsonMode = false;
|
||||
|
||||
/**
|
||||
* Emit an NDJSON event to stdout.
|
||||
* In JSON mode, only stdout is used (no stderr logging).
|
||||
* In human mode, only stderr logging is used (no JSON).
|
||||
* @param {string} type - "progress" | "success" | "error"
|
||||
* @param {object} data
|
||||
*/
|
||||
function emit(type, data) {
|
||||
// JSON mode: emit NDJSON to stdout only
|
||||
if (jsonMode) {
|
||||
process.stdout.write(JSON.stringify({ type, ...data }) + '\n');
|
||||
return;
|
||||
}
|
||||
|
||||
// Human mode: log to stderr, and output JSON for terminal events
|
||||
if (type === 'progress' && data.message) {
|
||||
log(data.message);
|
||||
} else if (type === 'success') {
|
||||
console.log(JSON.stringify(data, null, 2));
|
||||
} else if (type === 'error' && data.message) {
|
||||
log('ERROR:', data.message);
|
||||
console.log(JSON.stringify({ type, ...data }, null, 2));
|
||||
}
|
||||
}
|
||||
|
||||
/**
|
||||
* Output a single JSON object to stdout.
|
||||
* @param {object} data
|
||||
*/
|
||||
function output(data) {
|
||||
if (jsonMode) {
|
||||
process.stdout.write(JSON.stringify(data) + '\n');
|
||||
} else {
|
||||
console.log(JSON.stringify(data, null, 2));
|
||||
}
|
||||
}
|
||||
|
||||
/**
|
||||
* Output error and exit.
|
||||
* @param {string} message
|
||||
*/
|
||||
function error(message) {
|
||||
emit('error', { message });
|
||||
process.exit(1);
|
||||
}
|
||||
|
||||
/**
|
||||
* Parse command line arguments.
|
||||
* @returns {{command: string, args: object}}
|
||||
*/
|
||||
function parseArgs() {
|
||||
const args = process.argv.slice(2);
|
||||
const command = args[0];
|
||||
const parsed = { command, args: {} };
|
||||
|
||||
for (let i = 1; i < args.length; i++) {
|
||||
if (args[i].startsWith('--')) {
|
||||
const key = args[i].slice(2);
|
||||
const value = args[i + 1]?.startsWith('--') ? undefined : args[i + 1];
|
||||
parsed.args[key] = value === undefined ? true : value;
|
||||
if (value !== undefined) i++;
|
||||
}
|
||||
}
|
||||
|
||||
return parsed;
|
||||
}
|
||||
|
||||
function readStdin() {
|
||||
return new Promise((resolve, reject) => {
|
||||
let data = '';
|
||||
process.stdin.setEncoding('utf8');
|
||||
process.stdin.on('data', chunk => { data += chunk; });
|
||||
process.stdin.on('end', () => resolve(data.trim()));
|
||||
process.stdin.on('error', reject);
|
||||
// Timeout if no data within 1s — treat as empty input
|
||||
setTimeout(() => {
|
||||
if (!data) resolve('');
|
||||
}, 1000);
|
||||
});
|
||||
}
|
||||
|
||||
/**
|
||||
* Resolve image path - handle relative paths.
|
||||
* @param {string} imagePath
|
||||
* @returns {string}
|
||||
*/
|
||||
function resolveImagePath(imagePath) {
|
||||
if (path.isAbsolute(imagePath)) return imagePath;
|
||||
return path.resolve(process.cwd(), imagePath);
|
||||
}
|
||||
|
||||
/**
|
||||
* Check login and output error if not logged in.
|
||||
* @param {import('puppeteer-core').Page} page
|
||||
*/
|
||||
async function requireLogin(page) {
|
||||
const loginStatus = await checkLogin(page);
|
||||
if (!loginStatus.loggedIn) {
|
||||
emit('error', {
|
||||
type: 'not_logged_in',
|
||||
message: 'Not logged into Gemini. Please log in and try again.',
|
||||
action_required: 'open_browser_and_login',
|
||||
url: config.geminiUrl,
|
||||
});
|
||||
process.exit(0);
|
||||
}
|
||||
if (loginStatus.account) {
|
||||
emit('progress', { step: 'login', message: `Logged in as: ${loginStatus.account}` });
|
||||
}
|
||||
}
|
||||
|
||||
async function main() {
|
||||
const { command, args } = parseArgs();
|
||||
|
||||
// Check for JSON mode flag
|
||||
jsonMode = args.json === true || args.json === 'true';
|
||||
if (jsonMode) {
|
||||
process.env.GEMINI_JSON_MODE = '1';
|
||||
}
|
||||
|
||||
if (!command || command === 'help' || command === '--help' || command === '-h') {
|
||||
const helpText = `
|
||||
gemini-web-cli — 通过 CLI 驱动 Gemini 网页 AI 生图
|
||||
|
||||
用法:
|
||||
node cli.js <command> [options]
|
||||
|
||||
命令:
|
||||
|
||||
generate 向 Gemini 发送提示词（可选参考图），等待图片生成
|
||||
download 从当前 Gemini 标签页下载已生成的图片
|
||||
status 检查页面状态（空闲 / 生成中 / 完成 / 异常）
|
||||
sessions 列出所有活跃的 Gemini 会话（标签页）
|
||||
find_session 通过 Gemini 对话链接找回已丢失的 session
|
||||
close 关闭指定会话的标签页
|
||||
|
||||
generate 参数:
|
||||
--prompt "文本" 发送给 Gemini 的提示词（必填），支持 "stdin" 从管道读取
|
||||
--prompt-file <路径> 从文件读取提示词（支持换行）
|
||||
--image <路径> 单张参考图片路径
|
||||
--images <路径1,路径2,...> 多张参考图片，逗号分隔，最多 10 张
|
||||
--session <id> 复用已有会话的标签页
|
||||
--chatUrl <url> 打开指定的 Gemini 对话链接（创建新会话）
|
||||
--mode <single|multi> "single" 生成后关闭标签页，"multi" 保持标签页打开（默认 multi）
|
||||
--tool <工具名> 发送前选择指定工具（默认选择"制作图片"）
|
||||
--timeout <毫秒> 生成最大等待时间（默认 300000ms = 5分钟）
|
||||
--download-timeout <毫秒> 下载最大等待时间（默认 120000ms = 2分钟）
|
||||
--screenshot 出错或超时时截图（低质量 JPEG），保存路径写入输出
|
||||
--json 启用 NDJSON 模式：每行输出一个 JSON 事件
|
||||
|
||||
download 参数:
|
||||
--timeout <毫秒> 下载最大等待时间（默认 120000ms = 2分钟）
|
||||
--session <id> 复用已有会话的标签页
|
||||
|
||||
status 参数:
|
||||
--session <id> 检查指定会话的标签页
|
||||
--wait 持续轮询，直到完成 / 出错
|
||||
--timeout <毫秒> --wait 模式下的最大等待时间（默认 300000ms）
|
||||
--screenshot 状态异常时截图（低质量 JPEG），保存路径写入输出
|
||||
|
||||
sessions（无参数）
|
||||
|
||||
find_session 参数:
|
||||
--chatUrl <url> Gemini 对话链接（必填）
|
||||
--open 如果没找到匹配的标签页，自动打开新标签页并导航
|
||||
|
||||
close 参数:
|
||||
--session <id> 要关闭的会话 ID（必填）
|
||||
|
||||
示例:
|
||||
|
||||
# 提示词 + 参考图生图
|
||||
node cli.js generate --prompt "一只可爱的猫" --image ./cat.png
|
||||
|
||||
# NDJSON 模式（适合程序解析）
|
||||
node cli.js generate --prompt "一只可爱的猫" --image ./cat.png --json
|
||||
|
||||
# single 模式，生成后自动关闭标签页
|
||||
node cli.js generate --prompt "日落风景" --mode single
|
||||
|
||||
# 轮询状态，直到完成
|
||||
node cli.js status --session 7ac6cfff --wait
|
||||
|
||||
# 快速检查状态（立即返回）
|
||||
node cli.js status --session 7ac6cfff
|
||||
|
||||
# 出错/超时时自动截图（路径写入输出 JSON）
|
||||
node cli.js generate --prompt "日落风景" --screenshot
|
||||
node cli.js status --session 7ac6cfff --wait --screenshot
|
||||
|
||||
# 下载已生成的图片
|
||||
node cli.js download --session 7ac6cfff
|
||||
|
||||
# 列出所有活跃的会话
|
||||
node cli.js sessions
|
||||
|
||||
# 通过对话链接找回 session
|
||||
node cli.js find_session --chatUrl "https://gemini.google.com/app/4c089f364e1cf745"
|
||||
|
||||
# 没找到时自动打开新标签页
|
||||
node cli.js find_session --chatUrl "https://gemini.google.com/app/4c089f364e1cf745" --open
|
||||
|
||||
# 关闭指定会话的标签页
|
||||
node cli.js close --session 7ac6cfff
|
||||
|
||||
# 通过之前的对话链接继续生图
|
||||
node cli.js generate --chatUrl "https://gemini.google.com/app/4c089f364e1cf745" --prompt "换成晚上场景"
|
||||
|
||||
环境变量:
|
||||
CDP_URL Chrome DevTools Protocol 连接地址（默认 http://127.0.0.1:9222）
|
||||
`.trim();
|
||||
process.stdout.write(helpText + '\n');
|
||||
return;
|
||||
}
|
||||
|
||||
// Sessions command
|
||||
if (command === 'sessions') {
|
||||
const sessions = await listSessions();
|
||||
output({ type: 'sessions', sessions });
|
||||
return;
|
||||
}
|
||||
|
||||
// Find session command — recover a lost session by matching chatUrl
|
||||
if (command === 'find_session') {
|
||||
if (!args.chatUrl) {
|
||||
error('--chatUrl is required');
|
||||
}
|
||||
const openIfNotFound = args.open === true || args.open === 'true';
|
||||
const result = await findSessionByChatUrl({ chatUrl: args.chatUrl, openIfNotFound });
|
||||
output({ type: 'find_session', ...result });
|
||||
return;
|
||||
}
|
||||
|
||||
// Close command
|
||||
if (command === 'close') {
|
||||
if (!args.session) {
|
||||
error('--session is required to close a tab');
|
||||
}
|
||||
const result = await closeSession(args.session);
|
||||
output({ type: 'close', ...result });
|
||||
return;
|
||||
}
|
||||
|
||||
// Ensure output directories exist
|
||||
fs.mkdirSync(config.outputDir, { recursive: true });
|
||||
fs.mkdirSync(config.screenshotDir, { recursive: true });
|
||||
fs.mkdirSync(config.downloadDir, { recursive: true });
|
||||
|
||||
let browser;
|
||||
try {
|
||||
emit('progress', { step: 'start', message: `Command: ${command}` });
|
||||
|
||||
const sessionId = args.session || undefined;
|
||||
// Default to 'multi' mode (keep tab open for subsequent calls)
|
||||
const mode = (args.mode === 'multi' || args.mode === 'single') ? args.mode : 'multi';
|
||||
|
||||
switch (command) {
|
||||
case 'generate': {
|
||||
// Resolve prompt: --prompt-file > --prompt (no value = stdin) > --prompt literal
|
||||
let prompt = args.prompt;
|
||||
if (args.promptFile) {
|
||||
const p = resolveImagePath(args.promptFile);
|
||||
if (!fs.existsSync(p)) {
|
||||
error(`Prompt file not found: ${p}`);
|
||||
}
|
||||
prompt = fs.readFileSync(p, 'utf8').trim();
|
||||
} else if (prompt === true || prompt === 'stdin') {
|
||||
// --prompt 后面没有值，或显式写了 stdin，从管道/heredoc 读取
|
||||
prompt = await readStdin();
|
||||
}
|
||||
|
||||
if (!prompt) {
|
||||
error('--prompt is required');
|
||||
}
|
||||
|
||||
// Support --images path1,path2,... or --image single
|
||||
let imagePaths = [];
|
||||
const imagesArg = args.images;
|
||||
if (imagesArg) {
|
||||
imagePaths = imagesArg.split(',').map(p => resolveImagePath(p.trim()));
|
||||
} else if (args.image) {
|
||||
imagePaths = [resolveImagePath(args.image)];
|
||||
}
|
||||
if (imagePaths.length > 10) {
|
||||
imagePaths = imagePaths.slice(0, 10);
|
||||
}
|
||||
for (const p of imagePaths) {
|
||||
if (!fs.existsSync(p)) {
|
||||
error(`Image file not found: ${p}`);
|
||||
}
|
||||
}
|
||||
|
||||
const timeout = args.timeout ? parseInt(args.timeout, 10) : config.timeout;
|
||||
|
||||
emit('progress', { step: 'connect', message: 'Connecting to browser...' });
|
||||
const { browser: b, page, cdp, sessionId: sid, isNew } = await connectBrowser({ sessionId });
|
||||
browser = b;
|
||||
|
||||
let continuedSession = false;
|
||||
if (sessionId && !isNew) {
|
||||
emit('progress', { step: 'connect', message: `Continuing session: ${sessionId} (mode: ${mode})` });
|
||||
// Don't navigate — stay on the current chat page for multi-round
|
||||
continuedSession = true;
|
||||
} else if (args.chatUrl) {
|
||||
emit('progress', { step: 'navigate', message: `Navigating to chat URL: ${args.chatUrl}` });
|
||||
await page.goto(args.chatUrl, { waitUntil: 'domcontentloaded', timeout: 60000 });
|
||||
// Wait for conversation content to load — check for download buttons or existing images
|
||||
await page.waitForFunction(
|
||||
() => document.querySelectorAll('button.generated-image-button').length > 0 ||
|
||||
document.querySelectorAll('[contenteditable="true"]').length > 0,
|
||||
{ timeout: 30000 }
|
||||
).catch(() => {});
|
||||
await sleep(3000);
|
||||
// Chat URL opens an existing conversation, treat as continued
|
||||
continuedSession = true;
|
||||
} else {
|
||||
emit('progress', { step: 'navigate', message: `New session created (mode: ${mode})` });
|
||||
await navigateToGemini(page);
|
||||
}
|
||||
|
||||
emit('progress', { step: 'login', message: 'Checking login status...' });
|
||||
const loginStatus = await checkLogin(page);
|
||||
if (!loginStatus.loggedIn) {
|
||||
emit('error', {
|
||||
message: 'Not logged into Gemini. Please log in at ' + config.geminiUrl + ' and try again.',
|
||||
action_required: 'open_browser_and_login',
|
||||
sessionId: sid,
|
||||
});
|
||||
return;
|
||||
}
|
||||
if (loginStatus.account) {
|
||||
emit('progress', { step: 'login', message: `Logged in as: ${loginStatus.account}` });
|
||||
}
|
||||
|
||||
// generate() handles its own progress events
|
||||
const genStartTime = Date.now();
|
||||
|
||||
const result = await generate(page, {
|
||||
prompt: prompt,
|
||||
images: imagePaths,
|
||||
timeout,
|
||||
isContinuedSession: continuedSession,
|
||||
tool: args.tool,
|
||||
});
|
||||
|
||||
// Capture the Gemini chat URL for recovery
|
||||
const chatUrl = await captureChatUrl(sid, page);
|
||||
const elapsed = ((Date.now() - genStartTime) / 1000).toFixed(1);
|
||||
|
||||
// Add session info to result
|
||||
result.sessionId = sid;
|
||||
result.chatUrl = chatUrl;
|
||||
result.mode = mode;
|
||||
|
||||
if (result.status === 'success') {
|
||||
emit('progress', {
|
||||
step: 'generate',
|
||||
state: 'DONE',
|
||||
elapsed: parseFloat(elapsed),
|
||||
message: `Generation completed (${elapsed}s)`,
|
||||
});
|
||||
|
||||
// Only attempt download if this was an image generation
|
||||
try {
|
||||
const downloadTimeout = parseInt(args.downloadTimeout, 10) || 120000;
|
||||
const downloaded = await downloadViaButtons(page, cdp, {
|
||||
existingButtonCount: result.existingButtonCount || 0,
|
||||
timeout: downloadTimeout,
|
||||
});
|
||||
if (downloaded.length > 0) {
|
||||
result.images = downloaded;
|
||||
for (const img of downloaded) {
|
||||
const fileName = img.path.split('/').pop();
|
||||
emit('progress', { step: 'download', message: `Downloaded: ${fileName}` });
|
||||
}
|
||||
}
|
||||
} catch (e) {
|
||||
emit('progress', { step: 'warning', message: `Failed to download generated images: ${e.message}` });
|
||||
emit('progress', { step: 'download', message: `Download failed: ${e.message}` });
|
||||
}
|
||||
|
||||
// Remove internal fields from output
|
||||
delete result.existingButtonCount;
|
||||
|
||||
// Emit final success event
|
||||
emit('success', {
|
||||
sessionId: sid,
|
||||
chatUrl,
|
||||
mode,
|
||||
images: result.images?.map(i => i.path) || [],
|
||||
});
|
||||
} else if (result.status === 'text_only') {
|
||||
// Text-only response means image generation failed
|
||||
emit('error', {
|
||||
message: result.message,
|
||||
sessionId: sid,
|
||||
status: 'text_only',
|
||||
});
|
||||
} else {
|
||||
// Error or timeout — emit error as terminal event
|
||||
emit('error', {
|
||||
message: result.message,
|
||||
sessionId: sid,
|
||||
status: result.status,
|
||||
timeout: result.timeout,
|
||||
});
|
||||
}
|
||||
|
||||
// Capture screenshot on error or timeout
|
||||
if (args.screenshot && result.status && result.status !== 'success') {
|
||||
try {
|
||||
const ss = await takeScreenshot(page);
|
||||
result.screenshot = ss.path;
|
||||
} catch (e) {
|
||||
emit('progress', { step: 'warning', message: `Failed to take screenshot: ${e.message}` });
|
||||
}
|
||||
}
|
||||
|
||||
// Single mode: close the tab
|
||||
if (mode === 'single') {
|
||||
await closeSession(sid);
|
||||
}
|
||||
break;
|
||||
}
|
||||
|
||||
case 'download': {
|
||||
emit('progress', { step: 'connect', message: 'Connecting to browser...' });
|
||||
const { browser: b, page, cdp, sessionId: sid, isNew } = await connectBrowser({ sessionId });
|
||||
browser = b;
|
||||
|
||||
// Don't navigate — stay on the current chat page
|
||||
if (isNew) {
|
||||
emit('progress', { step: 'navigate', message: 'Navigating to Gemini...' });
|
||||
await navigateToGemini(page);
|
||||
}
|
||||
|
||||
emit('progress', { step: 'login', message: 'Checking login status...' });
|
||||
await requireLogin(page);
|
||||
|
||||
emit('progress', { step: 'download', message: 'Downloading generated images...' });
|
||||
const downloadTimeout = parseInt(args.downloadTimeout, 10) || parseInt(args.timeout, 10) || 120000;
|
||||
const downloadResult = await downloadViaButtons(page, cdp, { newestOnly: true, timeout: downloadTimeout });
|
||||
if (downloadResult.length === 0) {
|
||||
error('No new images found to download.');
|
||||
}
|
||||
|
||||
for (const img of downloadResult) {
|
||||
const fileName = img.path.split('/').pop();
|
||||
emit('progress', { step: 'download', message: `Downloaded: ${fileName}` });
|
||||
}
|
||||
|
||||
emit('success', {
|
||||
sessionId: sid,
|
||||
path: downloadResult[0].path,
|
||||
});
|
||||
|
||||
if (mode === 'single') {
|
||||
await closeSession(sid);
|
||||
}
|
||||
break;
|
||||
}
|
||||
|
||||
case 'status': {
|
||||
emit('progress', { step: 'connect', message: 'Connecting to browser...' });
|
||||
const { browser: b, page, sessionId: sid, isNew } = await connectBrowser({ sessionId });
|
||||
browser = b;
|
||||
|
||||
// Navigate if a new tab was created (either no session requested,
|
||||
// or requested session was lost and replaced)
|
||||
if (isNew) {
|
||||
emit('progress', { step: 'navigate', message: 'Navigating to Gemini...' });
|
||||
await navigateToGemini(page);
|
||||
}
|
||||
|
||||
emit('progress', { step: 'login', message: 'Checking login status...' });
|
||||
const loginStatus = await checkLogin(page);
|
||||
if (!loginStatus.loggedIn) {
|
||||
emit('error', {
|
||||
message: 'Not logged into Gemini',
|
||||
action_required: 'open_browser_and_login',
|
||||
sessionId: sid,
|
||||
});
|
||||
return;
|
||||
}
|
||||
|
||||
const waitForCompletion = args.wait === true || args.wait === 'true';
|
||||
const pollTimeout = args.timeout ? parseInt(args.timeout, 10) : undefined;
|
||||
|
||||
if (waitForCompletion) {
|
||||
emit('progress', { step: 'status', message: 'Waiting for generation to complete...' });
|
||||
}
|
||||
|
||||
const result = await checkStatus(page, {
|
||||
waitForCompletion,
|
||||
timeout: pollTimeout,
|
||||
onStateChange: (state, elapsed) => {
|
||||
emit('progress', { step: 'status', state, elapsed, message: `State: ${state} (${elapsed}s)` });
|
||||
},
|
||||
});
|
||||
result.sessionId = sid;
|
||||
|
||||
// Capture screenshot on error or unclear states
|
||||
if (args.screenshot && result.state && ['error', 'page_error', 'generating', 'idle'].includes(result.state)) {
|
||||
try {
|
||||
const ss = await takeScreenshot(page);
|
||||
result.screenshot = ss.path;
|
||||
} catch (e) {
|
||||
emit('progress', { step: 'warning', message: `Failed to take screenshot: ${e.message}` });
|
||||
}
|
||||
}
|
||||
|
||||
output({ type: 'status', ...result });
|
||||
break;
|
||||
}
|
||||
|
||||
default:
|
||||
error(`Unknown command: ${command}. Available commands: generate, download, status, sessions, close`);
|
||||
}
|
||||
} catch (err) {
|
||||
// CDP connection failures
|
||||
if (err.message.includes('ECONNREFUSED') || err.message.includes('connect')) {
|
||||
error(`Cannot connect to browser at ${config.cdpUrl}. Make sure Chrome is running with --remote-debugging-port.`);
|
||||
}
|
||||
error(err.message);
|
||||
} finally {
|
||||
// Don't close the browser - it's managed externally (OpenClaw, etc.)
|
||||
// Just release the CDP connection
|
||||
if (browser) {
|
||||
try {
|
||||
await browser.disconnect();
|
||||
} catch {
|
||||
// Already disconnected
|
||||
}
|
||||
}
|
||||
}
|
||||
}
|
||||
|
||||
main();
|
||||
Reference in New Issue
Block a user