521 lines
13 KiB
Go
521 lines
13 KiB
Go
package geminiweb
|
||
|
||
import (
|
||
"context"
|
||
"cursor-api-proxy/internal/apitypes"
|
||
"cursor-api-proxy/internal/config"
|
||
"fmt"
|
||
"os"
|
||
"path/filepath"
|
||
"strings"
|
||
"sync"
|
||
"time"
|
||
|
||
"github.com/playwright-community/playwright-go"
|
||
)
|
||
|
||
// PlaywrightProvider 使用 Playwright 的 Gemini Provider
|
||
type PlaywrightProvider struct {
|
||
cfg config.BridgeConfig
|
||
pw *playwright.Playwright
|
||
browser playwright.Browser
|
||
context playwright.BrowserContext
|
||
page playwright.Page
|
||
mu sync.Mutex
|
||
userDataDir string
|
||
}
|
||
|
||
var (
|
||
playwrightInstance *playwright.Playwright
|
||
playwrightOnce sync.Once
|
||
playwrightErr error
|
||
)
|
||
|
||
// NewPlaywrightProvider 建立新的 Playwright Provider
|
||
func NewPlaywrightProvider(cfg config.BridgeConfig) (*PlaywrightProvider, error) {
|
||
// 確保 Playwright 已初始化（單例）
|
||
playwrightOnce.Do(func() {
|
||
playwrightInstance, playwrightErr = playwright.Run()
|
||
if playwrightErr != nil {
|
||
playwrightErr = fmt.Errorf("failed to run playwright: %w", playwrightErr)
|
||
}
|
||
})
|
||
|
||
if playwrightErr != nil {
|
||
return nil, playwrightErr
|
||
}
|
||
|
||
// 清理 Chrome 鎖檔案
|
||
userDataDir := filepath.Join(cfg.GeminiAccountDir, "default-session")
|
||
cleanLockFiles(userDataDir)
|
||
|
||
// 確保目錄存在
|
||
if err := os.MkdirAll(userDataDir, 0755); err != nil {
|
||
return nil, fmt.Errorf("failed to create user data dir: %w", err)
|
||
}
|
||
|
||
return &PlaywrightProvider{
|
||
cfg: cfg,
|
||
pw: playwrightInstance,
|
||
userDataDir: userDataDir,
|
||
}, nil
|
||
}
|
||
|
||
// getName 返回 Provider 名稱
|
||
func (p *PlaywrightProvider) Name() string {
|
||
return "gemini-web"
|
||
}
|
||
|
||
// launchIfNeeded 如果需要則啟動瀏覽器
|
||
func (p *PlaywrightProvider) launchIfNeeded() error {
|
||
p.mu.Lock()
|
||
defer p.mu.Unlock()
|
||
|
||
if p.context != nil && p.page != nil {
|
||
return nil
|
||
}
|
||
|
||
fmt.Println("[GeminiWeb] Launching Chromium...")
|
||
|
||
// 使用 LaunchPersistentContext（自動保存 session）
|
||
context, err := p.pw.Chromium.LaunchPersistentContext(p.userDataDir,
|
||
playwright.BrowserTypeLaunchPersistentContextOptions{
|
||
Headless: playwright.Bool(!p.cfg.GeminiBrowserVisible),
|
||
Args: []string{
|
||
"--no-first-run",
|
||
"--no-default-browser-check",
|
||
"--disable-background-networking",
|
||
"--disable-extensions",
|
||
"--disable-plugins",
|
||
"--disable-sync",
|
||
},
|
||
})
|
||
if err != nil {
|
||
return fmt.Errorf("failed to launch persistent context: %w", err)
|
||
}
|
||
|
||
p.context = context
|
||
|
||
// 取得或建立頁面
|
||
pages := context.Pages()
|
||
if len(pages) > 0 {
|
||
p.page = pages[0]
|
||
} else {
|
||
page, err := context.NewPage()
|
||
if err != nil {
|
||
_ = context.Close()
|
||
return fmt.Errorf("failed to create page: %w", err)
|
||
}
|
||
p.page = page
|
||
}
|
||
|
||
fmt.Println("[GeminiWeb] Browser launched")
|
||
return nil
|
||
}
|
||
|
||
// Generate 生成回應
|
||
func (p *PlaywrightProvider) Generate(ctx context.Context, model string, messages []apitypes.Message, tools []apitypes.Tool, cb func(apitypes.StreamChunk)) error {
|
||
fmt.Printf("[GeminiWeb] Starting generation with model: %s\n", model)
|
||
|
||
// 1. 確保瀏覽器已啟動
|
||
if err := p.launchIfNeeded(); err != nil {
|
||
return fmt.Errorf("failed to launch browser: %w", err)
|
||
}
|
||
|
||
// 2. 導航到 Gemini（如果需要）
|
||
currentURL := p.page.URL()
|
||
if !strings.Contains(currentURL, "gemini.google.com") {
|
||
fmt.Println("[GeminiWeb] Navigating to Gemini...")
|
||
if _, err := p.page.Goto("https://gemini.google.com/app", playwright.PageGotoOptions{
|
||
WaitUntil: playwright.WaitUntilStateNetworkidle,
|
||
Timeout: playwright.Float(30000),
|
||
}); err != nil {
|
||
return fmt.Errorf("failed to navigate: %w", err)
|
||
}
|
||
}
|
||
|
||
// 3. 調試模式：如果可見，等待用戶確認
|
||
if p.cfg.GeminiBrowserVisible {
|
||
fmt.Println("\n" + strings.Repeat("=", 60))
|
||
fmt.Println("DEBUG MODE: Browser is visible")
|
||
fmt.Println("Please check the browser and press ENTER when ready...")
|
||
fmt.Println("If you see login page, please log in first")
|
||
fmt.Println(strings.Repeat("=", 60))
|
||
|
||
// 等待用戶按 Enter
|
||
var input string
|
||
fmt.Scanln(&input)
|
||
|
||
// 截圖保存
|
||
screenshotPath := "/tmp/gemini-debug.png"
|
||
if _, err := p.page.Screenshot(playwright.PageScreenshotOptions{
|
||
Path: playwright.String(screenshotPath),
|
||
}); err == nil {
|
||
fmt.Printf("[GeminiWeb] Screenshot saved to: %s\n", screenshotPath)
|
||
}
|
||
|
||
// 輸出頁面信息
|
||
fmt.Printf("[GeminiWeb] Current URL: %s\n", p.page.URL())
|
||
if title, err := p.page.Title(); err == nil {
|
||
fmt.Printf("[GeminiWeb] Page Title: %s\n", title)
|
||
}
|
||
}
|
||
|
||
// 4. 等待頁面完全載入（project-golem 策略）
|
||
fmt.Println("[GeminiWeb] Waiting for page to be ready...")
|
||
if err := p.waitForPageReady(); err != nil {
|
||
fmt.Printf("[GeminiWeb] Warning: %v\n", err)
|
||
|
||
// 額外調試：輸出頁面 HTML 結構
|
||
if p.cfg.GeminiBrowserVisible {
|
||
html, _ := p.page.Content()
|
||
debugPath := "/tmp/gemini-debug.html"
|
||
if err := os.WriteFile(debugPath, []byte(html), 0644); err == nil {
|
||
fmt.Printf("[GeminiWeb] HTML saved to: %s\n", debugPath)
|
||
}
|
||
}
|
||
}
|
||
|
||
// 4. 檢查登入狀態
|
||
fmt.Println("[GeminiWeb] Checking login status...")
|
||
loggedIn := p.isLoggedIn()
|
||
if !loggedIn {
|
||
fmt.Println("[GeminiWeb] Not logged in, continuing anyway")
|
||
if p.cfg.GeminiBrowserVisible {
|
||
fmt.Println("\n========================================")
|
||
fmt.Println("Browser is open. You can:")
|
||
fmt.Println("1. Log in to Gemini now")
|
||
fmt.Println("2. Continue without login")
|
||
fmt.Println("========================================\n")
|
||
}
|
||
} else {
|
||
fmt.Println("[GeminiWeb] Logged in")
|
||
}
|
||
|
||
// 5. 建構提示詞
|
||
prompt := buildPromptFromMessagesPlaywright(messages)
|
||
fmt.Printf("[GeminiWeb] Typing prompt (%d chars)...\n", len(prompt))
|
||
|
||
// 6. 輸入文字（使用 Playwright 的 Auto-wait）
|
||
if err := p.typeInput(prompt); err != nil {
|
||
return fmt.Errorf("failed to type: %w", err)
|
||
}
|
||
|
||
// 7. 發送訊息
|
||
fmt.Println("[GeminiWeb] Sending message...")
|
||
if err := p.sendMessage(); err != nil {
|
||
return fmt.Errorf("failed to send: %w", err)
|
||
}
|
||
|
||
// 8. 提取回應
|
||
fmt.Println("[GeminiWeb] Waiting for response...")
|
||
response, err := p.extractResponse()
|
||
if err != nil {
|
||
return fmt.Errorf("failed to extract response: %w", err)
|
||
}
|
||
|
||
// 9. 回調
|
||
cb(apitypes.StreamChunk{Type: apitypes.ChunkText, Text: response})
|
||
cb(apitypes.StreamChunk{Type: apitypes.ChunkDone, Done: true})
|
||
|
||
fmt.Printf("[GeminiWeb] Response complete (%d chars)\n", len(response))
|
||
return nil
|
||
}
|
||
|
||
// Close 關閉 Provider
|
||
func (p *PlaywrightProvider) Close() error {
|
||
p.mu.Lock()
|
||
defer p.mu.Unlock()
|
||
|
||
if p.context != nil {
|
||
if err := p.context.Close(); err != nil {
|
||
return err
|
||
}
|
||
p.context = nil
|
||
p.page = nil
|
||
}
|
||
return nil
|
||
}
|
||
|
||
// waitForPageReady 等待頁面完全就緒（project-golem 策略）
|
||
func (p *PlaywrightProvider) waitForPageReady() error {
|
||
fmt.Println("[GeminiWeb] Checking for ready state...")
|
||
|
||
// 1. 等待停止按鈕消失（如果存在）
|
||
_, _ = p.page.WaitForSelector("button[aria-label*='Stop'], button[aria-label*='停止']", playwright.PageWaitForSelectorOptions{
|
||
State: playwright.WaitForSelectorStateDetached,
|
||
Timeout: playwright.Float(5000),
|
||
})
|
||
|
||
// 2. 嘗試多種等待策略
|
||
inputSelectors := []string{
|
||
".ProseMirror",
|
||
"rich-textarea",
|
||
"div[role='textbox']",
|
||
"div[contenteditable='true']",
|
||
"textarea",
|
||
"input[type='text']",
|
||
}
|
||
|
||
// 策略 A: 等待任一輸入框出現
|
||
for i, sel := range inputSelectors {
|
||
fmt.Printf(" [%d/%d] Waiting for: %s\n", i+1, len(inputSelectors), sel)
|
||
locator := p.page.Locator(sel)
|
||
if err := locator.WaitFor(playwright.LocatorWaitForOptions{
|
||
Timeout: playwright.Float(5000),
|
||
State: playwright.WaitForSelectorStateVisible,
|
||
}); err == nil {
|
||
fmt.Printf(" ✓ Input field found: %s\n", sel)
|
||
return nil
|
||
}
|
||
}
|
||
|
||
// 策略 B: 等待頁面完全載入
|
||
fmt.Println("[GeminiWeb] Waiting for page load...")
|
||
time.Sleep(3 * time.Second)
|
||
|
||
// 策略 C: 使用 JavaScript 檢查
|
||
fmt.Println("[GeminiWeb] Checking with JavaScript...")
|
||
result, err := p.page.Evaluate(`
|
||
() => {
|
||
// 檢查所有可能的輸入元素
|
||
const selectors = [
|
||
'.ProseMirror',
|
||
'rich-textarea',
|
||
'div[role="textbox"]',
|
||
'div[contenteditable="true"]',
|
||
'textarea',
|
||
'input[type="text"]'
|
||
];
|
||
|
||
for (const sel of selectors) {
|
||
const el = document.querySelector(sel);
|
||
if (el) {
|
||
return {
|
||
found: true,
|
||
selector: sel,
|
||
tagName: el.tagName,
|
||
className: el.className,
|
||
visible: el.offsetParent !== null
|
||
};
|
||
}
|
||
}
|
||
|
||
return { found: false };
|
||
}
|
||
`)
|
||
|
||
if err == nil {
|
||
if m, ok := result.(map[string]interface{}); ok {
|
||
if found, _ := m["found"].(bool); found {
|
||
sel, _ := m["selector"].(string)
|
||
fmt.Printf(" ✓ JavaScript found: %s\n", sel)
|
||
return nil
|
||
}
|
||
}
|
||
}
|
||
|
||
// 策略 D: 調試模式 - 輸出頁面結構
|
||
if p.cfg.GeminiBrowserVisible {
|
||
fmt.Println("[GeminiWeb].dump: Page structure analysis")
|
||
_, _ = p.page.Evaluate(`
|
||
() => {
|
||
const allElements = document.querySelectorAll('*');
|
||
const inputLike = [];
|
||
for (const el of allElements) {
|
||
if (el.contentEditable === 'true' ||
|
||
el.role === 'textbox' ||
|
||
el.tagName === 'TEXTAREA' ||
|
||
el.tagName === 'INPUT') {
|
||
inputLike.push({
|
||
tag: el.tagName,
|
||
class: el.className,
|
||
id: el.id,
|
||
role: el.role,
|
||
contentEditable: el.contentEditable
|
||
});
|
||
}
|
||
}
|
||
console.log('Input-like elements:', inputLike);
|
||
}
|
||
`)
|
||
}
|
||
|
||
return fmt.Errorf("no input field found after all strategies")
|
||
}
|
||
|
||
// isLoggedIn 檢查是否已登入
|
||
func (p *PlaywrightProvider) isLoggedIn() bool {
|
||
// 嘗試找輸入框（登入狀態的主要特徵）
|
||
selectors := []string{
|
||
".ProseMirror",
|
||
"rich-textarea",
|
||
"div[role='textbox']",
|
||
"div[contenteditable='true']",
|
||
"textarea",
|
||
}
|
||
|
||
for _, sel := range selectors {
|
||
locator := p.page.Locator(sel)
|
||
if count, _ := locator.Count(); count > 0 {
|
||
return true
|
||
}
|
||
}
|
||
return false
|
||
}
|
||
|
||
// typeInput 輸入文字（使用 Playwright 的 Auto-wait）
|
||
func (p *PlaywrightProvider) typeInput(text string) error {
|
||
fmt.Println("[GeminiWeb] Looking for input field...")
|
||
|
||
selectors := []string{
|
||
".ProseMirror",
|
||
"rich-textarea",
|
||
"div[role='textbox'][contenteditable='true']",
|
||
"div[contenteditable='true']",
|
||
"textarea",
|
||
}
|
||
|
||
var inputLocator playwright.Locator
|
||
var found bool
|
||
|
||
for _, sel := range selectors {
|
||
fmt.Printf(" Trying: %s\n", sel)
|
||
locator := p.page.Locator(sel)
|
||
if err := locator.WaitFor(playwright.LocatorWaitForOptions{
|
||
Timeout: playwright.Float(3000),
|
||
}); err == nil {
|
||
inputLocator = locator
|
||
found = true
|
||
fmt.Printf(" ✓ Found with: %s\n", sel)
|
||
break
|
||
}
|
||
}
|
||
|
||
if !found {
|
||
// 顯示 debug 信息
|
||
url := p.page.URL()
|
||
title, _ := p.page.Title()
|
||
return fmt.Errorf("input field not found (URL=%s, Title=%s)", url, title)
|
||
}
|
||
|
||
// Focus 並填充（Playwright 自動等待）
|
||
fmt.Printf("[GeminiWeb] Typing %d chars...\n", len(text))
|
||
if err := inputLocator.Fill(text); err != nil {
|
||
return fmt.Errorf("failed to fill: %w", err)
|
||
}
|
||
|
||
fmt.Println("[GeminiWeb] Input complete")
|
||
return nil
|
||
}
|
||
|
||
// sendMessage 發送訊息
|
||
func (p *PlaywrightProvider) sendMessage() error {
|
||
// 方法 1: 按 Enter（最可靠）
|
||
if err := p.page.Keyboard().Press("Enter"); err != nil {
|
||
return fmt.Errorf("failed to press Enter: %w", err)
|
||
}
|
||
|
||
time.Sleep(200 * time.Millisecond)
|
||
|
||
// 方法 2: 嘗試點擊發送按鈕（補強）
|
||
_, _ = p.page.Evaluate(`
|
||
() => {
|
||
const keywords = ['發送', 'Send', '傳送'];
|
||
const buttons = Array.from(document.querySelectorAll('button, [role="button"]'));
|
||
|
||
for (const btn of buttons) {
|
||
const text = (btn.innerText || btn.textContent || '').trim();
|
||
const label = (btn.getAttribute('aria-label') || '').trim();
|
||
|
||
// 跳過停止按鈕
|
||
if (['停止', 'Stop', '中斷'].includes(text) || label.toLowerCase().includes('stop')) {
|
||
continue;
|
||
}
|
||
|
||
if (keywords.some(kw => text.includes(kw) || label.includes(kw))) {
|
||
btn.click();
|
||
return true;
|
||
}
|
||
}
|
||
return false;
|
||
}
|
||
`)
|
||
|
||
return nil
|
||
}
|
||
|
||
// extractResponse 提取回應
|
||
func (p *PlaywrightProvider) extractResponse() (string, error) {
|
||
selectors := []string{
|
||
".model-response-text",
|
||
".message-content",
|
||
".markdown",
|
||
".prose",
|
||
"model-response",
|
||
}
|
||
|
||
var lastText string
|
||
lastUpdate := time.Now()
|
||
timeout := 120 * time.Second
|
||
startTime := time.Now()
|
||
|
||
for time.Since(startTime) < timeout {
|
||
time.Sleep(500 * time.Millisecond)
|
||
|
||
// 嘗試所有選擇器
|
||
for _, sel := range selectors {
|
||
locator := p.page.Locator(sel)
|
||
count, _ := locator.Count()
|
||
|
||
if count > 0 {
|
||
// 取最後一個元素
|
||
lastEl := locator.Last()
|
||
text, err := lastEl.TextContent()
|
||
if err != nil {
|
||
continue
|
||
}
|
||
|
||
text = strings.TrimSpace(text)
|
||
if text != "" && len(text) > len(lastText) {
|
||
lastText = text
|
||
lastUpdate = time.Now()
|
||
fmt.Printf("[GeminiWeb] Response length: %d\n", len(text))
|
||
}
|
||
}
|
||
}
|
||
|
||
// 檢查是否完成（2秒內無新內容）
|
||
if time.Since(lastUpdate) > 2*time.Second && lastText != "" {
|
||
// 最終檢查：停止按鈕是否還存在
|
||
stopBtn := p.page.Locator("button[aria-label*='Stop'], button[aria-label*='停止']")
|
||
count, _ := stopBtn.Count()
|
||
|
||
if count == 0 {
|
||
return lastText, nil
|
||
}
|
||
}
|
||
}
|
||
|
||
if lastText != "" {
|
||
return lastText, nil
|
||
}
|
||
return "", fmt.Errorf("response timeout")
|
||
}
|
||
|
||
// buildPromptFromMessages 從訊息列表建構提示詞
|
||
func buildPromptFromMessagesPlaywright(messages []apitypes.Message) string {
|
||
var prompt string
|
||
for _, m := range messages {
|
||
switch m.Role {
|
||
case "system":
|
||
prompt += "System: " + m.Content + "\n\n"
|
||
case "user":
|
||
prompt += m.Content + "\n\n"
|
||
case "assistant":
|
||
prompt += "Assistant: " + m.Content + "\n\n"
|
||
}
|
||
}
|
||
return prompt
|
||
}
|