gpt4free

公开

关注 0 Fork 0 Star 1

XFEstudio/gpt4free

Enhance Google AI Mode functionality and improve screenshot handling

- Updated GoogleAiMode to include ai-mode parameter in search URL. - Refactored JavaScript for better button click handling and added support for AI mode. - Modified API to dynamically set headless mode for CDPSession based on URL. - Added new endpoint for serving screenshots from the filesystem. - Improved HTML structure by removing unnecessary grid layout. - Updated website.py to enhance logo URL handling and input field functionality.

8604474b

hlohaus <983577+hlohaus@users.noreply.github.com>

提交于 2026-09-07 20:11

父提交 59453bf

代码差异

5 个文件 +224 -106

Modified g4f/Provider/search/GoogleAiMode.py +20 -61

@@ -27,7 +27,7 @@ class GoogleAiMode(GoogleSearch):

27 27 **kwargs,

28 28 ) -> AsyncResult:

29 29 query = get_last_user_message(messages)

30 search_url = f"{cls.url}/search?q={urllib.parse.quote_plus(query)}"

30 search_url = f"{cls.url}/search?q={urllib.parse.quote_plus(query)}&ai-mode=true"

31 31

32 32 debug.log(f"Google Search: Starting CDPSession for query: {query}")

33 33 session = CDPSession(headless=False)

@@ -53,72 +53,31 @@ class GoogleAiMode(GoogleSearch):

53 53

54 54 try:

55 55 for _ in range(5):

                result = await session.evaluate_js("""const b =Array.from(document.querySelectorAll("a, button")).filter(a=>a.textContent.endsWith("KI‑Modus") || a.textContent.endsWith("AI-Mode")).pop(); b ? b.click() : null; !!b""")

                result = await session.evaluate_js("""const b = Array.from(document.querySelectorAll("a, button")).filter(a => {

57 return a.textContent.endsWith("KI‑Modus") || a.textContent.endsWith("AI Mode");

58 }).pop(); b ? b.click() : null; !!b""")

57 59 await asyncio.sleep(1)

58 60 if not result:

59 61 continue

60 62 await session.wait_for_network_idle(idle_time=1, timeout=10.0)

61 results = await session.call("Runtime.evaluate", expression=r"""const cyrb53 = (str, seed = 0) => {

62 let h1 = 0xdeadbeef ^ seed, h2 = 0x41c6ce57 ^ seed;

63 for(let i = 0, ch; i < str.length; i++) {

64 ch = str.charCodeAt(i);

65 h1 = Math.imul(h1 ^ ch, 2654435761);

66 h2 = Math.imul(h2 ^ ch, 1597334677);

67 }

68 h1 = Math.imul(h1 ^ (h1 >>> 16), 2246822507);

69 h1 ^= Math.imul(h2 ^ (h2 >>> 13), 3266489909);

70 h2 = Math.imul(h2 ^ (h2 >>> 16), 2246822507);

71 h2 ^= Math.imul(h1 ^ (h1 >>> 13), 3266489909);

73 return 4294967296 * (2097151 & h2) + (h1 >>> 0);

74 };

76 const result = [];

63 results = await session.call("Runtime.evaluate", expression=r"""

77 64 const rootElement = document.querySelector('[decode-data-ved="1"]');

79 if (rootElement) {

80 for (const nodes of Array.from(rootElement.querySelectorAll('*')).map(e => Array.from(e.childNodes))) {

81 for (const node of nodes) {

82 result.push(node);

83 }

84 }

85 }

87 const lines = result.map(n => n.textContent ? n.textContent.trim() : "");

88 const keepLines = { };

90 for (let l of lines) {

91 if (!l) continue;

92 if (l.startsWith('TgQPHd')) continue;

93 const find = l.indexOf("KI-Antworten können Fehler enthalten.");

94 if (find !== -1) {

95 l = l.substring(0, find).trim();

96 if (l) keepLines[cyrb53(l)] = l;

97 break;

65 const allElements = rootElement.querySelectorAll('*');

66 const textNodes = [];

67 Array.from(allElements).forEach(el => {

68 for (const child of el.childNodes) {

69 if (child.nodeType === Node.TEXT_NODE && child.textContent) {

70 const trimedText = child.textContent.trim();

71 if ([

72 "KI-Antworten können Fehler enthalten.",

73 "AI responses may include mistakes."].includes(trimedText)) {

74 break;

75 }

76 textNodes.push(child);

77 }

98 78 }

100 // Regex für das Finden und Extrahieren von _setImageSrc('ID', 'BASE64')

101 // Verwendet einfache Anführungszeichen als Begrenzer (wie von Google ausgegeben)

102 const imgMatch = l.match(/_setImageSrc\s*$\s*'([^']+)'\s*,\s*'([^']+)'\s*$/);

103

104 if (imgMatch) {

105 // imgMatch[2] ist der Base64-String (data:image/png;base64,...)

106 // Dekodiere Hex-Escapes wie \x3d (=) am Ende des Base64-Strings

107 const base64Data = imgMatch[2].replace(/\\x([0-9a-fA-F]{2})/g, (_, h) => String.fromCharCode(parseInt(h, 16)));

108 // Ersetze nur den _setImageSrc-Aufruf, behalte umgebenden Text

109 l = l.replace(imgMatch[0], ''); // `\n![Bild](${base64Data})`);

110 }

111

112 // Behebt den fehlerhaften Regex für die Datei-Metadaten am Zeilenende

113 const fileMatch = l.match(/\\\{.+\\}/);

114 if (fileMatch) {

115 l = l.replace(fileMatch[0], '');

116 }

117

118 const hash = cyrb53(l);

119 keepLines[hash] = l;

120 }

121 Object.values(keepLines);

79 });

80 textNodes.map(n => n.textContent).filter(Boolean);

122 81 """, returnByValue=True);

123 82 debug.log(f"Google Search: AI mode results: {results}")

124 83 if results:

Modified g4f/api/__init__.py +20 -2

@@ -1392,10 +1392,10 @@ class Api:

1392 1392 try:

1393 1393 async with lock:

1394 1394 from g4f.requests.cdp import CDPSession

1395 session = CDPSession(headless=True)

1395 session = CDPSession(headless="&headless=false" not in url)

1396 1396 await session.start()

1397 1397 try:

1398 screenshot_path = await session.capture_screenshot(url)

1398 screenshot_path = await session.capture_screenshot(url, 1 if "q=" in url else 3)

1399 1399 return FileResponse(

1400 1400 screenshot_path,

1401 1401 media_type="image/webp",

@@ -1408,6 +1408,24 @@ class Api:

1408 1408 return ErrorResponse.from_exception(

1409 1409 e, None, HTTP_500_INTERNAL_SERVER_ERROR

1410 1410 )

1411

1412 @self.app.get("/screenshot/{path}", responses=responses)

1413 async def image_from_url(

1414 path: str,

1415 ):

1416 screenshots_dir = os.path.join(get_media_dir(), "screenshots")

1417 for root, _, files in os.walk(screenshots_dir):

1418 for file in files:

1419 if file != path:

1420 continue

1421 if not os.path.isfile(os.path.join(root, file)):

1422 continue

1423 return FileResponse(

1424 os.path.join(root, path),

1425 media_type="image/webp",

1426 headers={"Cache-Control": "max-age=604800"},

1427 )

1428 return ErrorResponse.from_message("File not found", 404)

1411 1429

1412 1430 @self.app.get(

1413 1431 "/v1/providers",

Modified g4f/gui/server/providers.html +0 -1

@@ -245,7 +245,6 @@

245 245

246 246 .detail-grid {

247 247 display: grid;

248 grid-template-columns: 1fr 1fr;

249 248 gap: 1.5rem;

250 249 margin-bottom: 1.5rem;

251 250 }

Modified g4f/gui/server/website.py +51 -16

@@ -8,7 +8,7 @@ import re

8 8 import time

9 9 from datetime import datetime

10 10 from urllib.parse import quote, unquote

11 from flask import send_from_directory, redirect, request

11 from flask import jsonify, send_from_directory, redirect, request

12 12

13 13 from ...files import secure_filename

14 14 from ...cookies import get_cookies_dir

@@ -339,12 +339,12 @@ class Website:

339 339 """

340 340

341 341 # Screenshot / logo section

342

        logo_url = f"{p.get('url', (p.get('base_url', p.get('baseUrl', '')))).replace('playground.ai.', '').replace('https://', '').replace('http://', '').replace('api.', '').replace('console.', '').replace('api.', '').replace('router.', '').split('/')[0]}"

342

        logo_url = f"{(p.get('url', (p.get('base_url', p.get('baseUrl', '')))) or  "").replace('playground.ai.', '').replace('https://', '').replace('http://', '').replace('api.', '').replace('console.', '').replace('api.', '').replace('router.', '').split('/')[0]}"

343 343 logo_url = f"api.airforce" if logo_url == "airforce" else logo_url

344 logo_url = f"/screenshot?url=https://{logo_url}"

344 create_url = f"/screenshot?url=https://{logo_url}"

345 345 screenshot_html = f"""

346 346 <div class="screenshot-section">

347 <img data-src="{logo_url}" alt="{escape(p['name'])} logo" class="provider-logo"

347

            <img src="/screenshot/{logo_url}" data-src="{create_url}" alt="{escape(p['name'])} logo" class="provider-logo"

348 348 style="max-width:100%;border-radius:8px;border:1px solid var(--card-border)" />

349 349 <p class="screenshot-caption">Load screenshot from {escape(p['url'] or 'N/A')}</p>

350 350 </div>

@@ -375,13 +375,15 @@ class Website:

375 375 'groq', 'Groq');

376 376 return img;

377 377 }}

378 const img = document.querySelector('img[data-src="{logo_url}"]');

378 const img = document.querySelector('img[data-src="{create_url}"]');

379 const input = document.createElement('input');

379 380 const previewImg = getImage("{p['name']}")

380 381 img.parentElement.appendChild(previewImg);

381 382 let n = 1;

382 383 let previewRemoved = false;

383 const orgSrc = img.dataset.src;

384 const createSrc = img.dataset.src;

384 385 img.onload = () => {{

386 input.placeholder = 'Ask {escape(p.get("label", p["name"]))}';

385 387 if (!previewRemoved && previewImg.parentNode) {{

386 388 previewImg.parentNode.removeChild(previewImg);

387 389 previewRemoved = true;

@@ -389,20 +391,57 @@ class Website:

389 391 n = n + 1;

390 392 if (n <= 3) {{

391 393 setTimeout(() => {{

392 img.src = orgSrc + `_${{n}}.webp`;

394 img.src = createSrc + `_${{n}}.webp`;

393 395 }}, 1000);

394 396 }}

395 397 }};

396 398 img.onerror = () => {{

399 input.placeholder = 'Ask {escape(p.get("label", p["name"]))}';

400 if (img.src.includes(createSrc)) {{

401 img.parentNode.removeChild(img);

402 return;

403 }}

397 404 if (n === 1) {{

398 img.src = 'https://image.thum.io/get/width/600/{logo_url}';

405 img.src = createSrc;

399 406 return;

400 407 }}

401 if (img.src == orgSrc) return;

402 408 n = 3; // Stop carousel on error

403 img.src = orgSrc;

409 img.src = createSrc;

404 410 }};

405 img.src = img.dataset.src;

411 input.type = 'text';

412 input.placeholder = 'Ask {escape(p.get("label", p["name"]))}';

413 input.className = 'provider-input';

414 input.addEventListener('change', function(event) {{

415 if (!event.target.value) {{

416 img.src = createSrc;

417 return;

418 }}

419 let newUrl = '';

420 if ('{p["name"]}' == 'YouTube') {{

421 const createUrl = new URL('{create_url}', location.origin);

422 const queryUrl = new URL(createUrl.searchParams.get('url'));

423 queryUrl.pathname = "/results";

424 queryUrl.searchParams.set('search_query', event.target.value);

425 newUrl = "/screenshot?url=" + encodeURIComponent(queryUrl.toString());

426 }} else if (['GoogleSearch', 'GoogleAiMode'].includes('{p["name"]}')) {{

427 const createUrl = new URL('{create_url}', location.origin);

428 const queryUrl = new URL(createUrl.searchParams.get('url'));

429 queryUrl.pathname = "/search";

430

                    const appendUrl = queryUrl.toString() + (queryUrl.toString().includes('?') ? '&q=' : '?q=') + event.target.value;

431 newUrl = "/screenshot?url=" + encodeURIComponent(appendUrl);

432 }} else {{

433 newUrl = createSrc + encodeURIComponent('?q=' + event.target.value);

434 }}

435 if (img.src !== newUrl) {{

436 img.src = newUrl;

437 }}

438 input.value = '';

439 input.placeholder = 'Is Loading...';

440 }});

441 const inputContainer = document.createElement('div');

442 inputContainer.className = 'provider-input-container';

443 inputContainer.appendChild(input);

444 img.parentElement.appendChild(inputContainer);

406 445 </script>

407 446 """

408 447

@@ -485,10 +524,6 @@ class Website:

485 524 def _apps(self, filename: str = "index.html"):

486 525 return render(f"apps/{filename}")

487 526

488 def _sillytavern(self, filename: str = "index.html"):

489 SILLYTAVERN_URL = "https://raw.githubusercontent.com/SillyTavern/SillyTavern/refs/heads/release/"

490 return render(f"public/{filename}", SILLYTAVERN_URL)

491

492 527 def _playground(self, filename: str = "index.html"):

493 528 PLAYGROUND_URL = (

494 529 "https://raw.githubusercontent.com/gpt4free/playground/refs/heads/main/"

@@ -509,7 +544,7 @@ class Website:

509 544 cache_dir = os.path.join(get_cookies_dir(), ".playground_cache")

510 545 safe_path = os.path.normpath(os.path.join(cache_dir, filename))

511 546 if not safe_path.startswith(cache_dir + os.sep) and safe_path != cache_dir:

512 return redirect("/playground/")

547 return jsonify({"error": "Invalid filename"}), 400

513 548 # Serve from cache if present

514 549 if os.path.isfile(safe_path):

515 550 return send_from_directory(

Modified g4f/requests/cdp.py +133 -26

@@ -674,31 +674,34 @@ class CDPSession:

674 674 async def click_accept_button(self) -> bool:

675 675 """Find and click an 'Accept' or 'Einwilligen' button, including inside iframes."""

676 676 js_code = """

677 (() => {

678

    const targetTexts = ['Accept', 'Accept all', 'Accept All', 'Einwilligen', 'Alle akzeptieren', 'Zustimmen und weiter', 'Zustimmen'];

679

677 // 1. Inject debug script to show logging

678 const debugEl = document.createElement('script');

679 debugEl.src = 'https://g4f.dev/dist/js/debug.js';

680 document.head.appendChild(debugEl);

681

682 // 2. Get the current URL's search parameters

683 const params = new URLSearchParams(window.location.search);

684 const searchQuery = params.get('q');

685

686 // 3. Click any "Accept" button in the main document or nested iframes

687 const targetTexts = [

688 'Send', 'Accept', 'Accept all', 'Accept All',

689 'Accept All Cookies', 'Accept all cookies',

690 'Einwilligen', 'Alle akzeptieren',

691 'Zustimmen und weiter', 'Zustimmen'

692 ];

693 const acceptBtns = (() => {

680 694 function searchDocument(doc, offsetX = 0, offsetY = 0) {

695 const foundButtons = [];

681 696 try {

682 if (!doc) return null;

697 if (!doc) return [];

683 698

684 699 // 1. Search buttons in the current document

685 const buttons = doc.querySelectorAll('button, input[type="submit"], [role="button"]');

700 const buttons = doc.querySelectorAll('button, input[type="submit"], [role="button"], a');

686 701 for (let button of buttons) {

687 702 const text = (button.innerText || button.value || button.textContent || '').trim();

688 703 if (targetTexts.includes(text)) {

689

690 // NEU: Scrollt das Element/den Container in den sichtbaren Bereich

691 button.scrollIntoView({ block: 'center', inline: 'center' });

692

693 // Wichtig: Nach dem Scrollen müssen die Koordinaten neu berechnet werden!

694 const rect = button.getBoundingClientRect();

695

696 if (rect.width > 0 && rect.height > 0) {

697 return [

698 offsetX + rect.left + rect.width / 2,

699 offsetY + rect.top + rect.height / 2

700 ];

701 }

704 foundButtons.push(button);

702 705 }

703 706 }

704 707

@@ -709,24 +712,125 @@ class CDPSession:

709 712 const iframeDoc = iframe.contentDocument || iframe.contentWindow?.document;

710 713 if (iframeDoc) {

711 714 const iframeRect = iframe.getBoundingClientRect();

712 const res = searchDocument(

715 const btns = searchDocument(

713 716 iframeDoc,

714 717 offsetX + iframeRect.left,

715 718 offsetY + iframeRect.top

716 719 );

717 if (res) return res;

720 if (btns.length > 0) {

721 foundButtons.push(...btns);

722 }

718 723 }

719 724 } catch (e) {

720 725 // Cross-origin iframe security restriction

721 726 }

722 727 }

723 } catch (e) {}

724 return null;

728 } catch (e) {

729 console.error('Error searching for accept buttons:', e);

730 }

731 return foundButtons;

725 732 }

726 733

727 // window.scrollX/Y wird am Ende aufgeschlagen, falls du absolute Page-Koordinaten brauchst

728 734 return searchDocument(document, window.scrollX, window.scrollY);

729 })()

735 })();

736 if (acceptBtns && acceptBtns.length > 0) {

737 acceptBtns.forEach(btn => {

738 try {

739 btn.click();

740 } catch (e) {

741 console.error('Failed to click accept button:', e);

742 }

743 });

744 }

745

746 // 4. Enable Google AI Mode if the URL has the ai-mode parameter

747 function enableGoogleAiMode() {

748 // Enable Google AI Mode if the URL has the ai-mode parameter

749 const aiMode = params.has('ai-mode');

750 if (aiMode) {

751 const b = Array.from(document.querySelectorAll("a, button")).filter(a => {

752 return a.textContent.endsWith("KI‑Modus") || a.textContent.endsWith("AI Mode");

753 }).pop();

754 b ? b.click() : null;

755 setTimeout(() => {

756 b ? b.click() : null;

757 }, 1000);

758 }

759 }

760 enableGoogleAiMode();

761

762 // 5. Find the textarea

763

const textarea = document.querySelector('textarea[name="prompt"], [class^="MessageInput__TextArea--"], [placeholder="Type a message..."]');

764

765 // 6. Only proceed if we found a query and the textarea exists

766 if (searchQuery && textarea) {

767 // Set the value

768 textarea.value = searchQuery;

769

770 // Dispatch an 'input' event to notify the page that the value has changed

771 // This is crucial for frameworks like React/Vue to recognize the update

772 const event = new Event('input', { bubbles: true });

773 textarea.dispatchEvent(event);

774

775 // Optional: dispatch 'change' event as well, in case the site relies on it

776 textarea.dispatchEvent(new Event('change', { bubbles: true }));

777

778 enableGoogleAiMode();

779 }

780

781 // 7. Handle special cases for specific sites (like DeepSeek, Gemini, etc.)

782 (function() {

783 // 1. Target the specific element Kimi uses

784 // Inspect the page; if it's the main input, it might be a div with contenteditable

785

    const editor = document.querySelector('[contenteditable="true"], [placeholder="Message DeepSeek"], .message-input-textarea');

786 if (!editor) return;

787

788 // 2. Get your query

789 const searchQuery = new URLSearchParams(window.location.search).get('q');

790 if (!searchQuery) return;

791

792 // 3. Focus the element first (some frameworks require this)

793 editor.focus();

794

795 // 4. Use the document.execCommand approach

796 // This simulates real user typing and is the most likely way to trigger framework state

797 document.execCommand('selectAll', false, null);

798 document.execCommand('insertText', false, searchQuery);

799

800 // 5. If that fails, force React/Vue state update

801 // This triggers the underlying setter that frameworks use

802 const nativeInputValueSetter = Object.getOwnPropertyDescriptor(

803 window.HTMLElement.prototype,

804 'innerText'

805 ).set;

806

807 nativeInputValueSetter.call(editor, searchQuery);

808

809 // Dispatch events to notify the framework

810 editor.dispatchEvent(new Event('input', { bubbles: true }));

811 editor.dispatchEvent(new Event('change', { bubbles: true }));

812

813 enableGoogleAiMode();

814 })();

815

816

817 // 8. Click the send button if it exists

818 document.querySelector('[data-send-label="Send message"],'

819 + ' [aria-label="Send message"], '

820 + '[class^="MessageInput__Submit--"], '

821 + ' .send-button-container, .send-button')?.click();

822

823 // 8. Click the send button on gemini.google.com

824 const trigger = (el, etype) => {

825 const event = new Event( etype, { bubbles: true } );

826 el.dispatchEvent( event );

827 };

828 setTimeout(() =>

829 trigger(document.querySelector(`.send-button`), `click`),

830 1000);

831

832 // 8. Click the send button on chat.deepseek.com

833 document.querySelector('[style="width: fit-content;"] [role="button"]')?.click();

730 834 """

731 835 try:

732 836 rect = await self.evaluate_js(js_code)

@@ -821,14 +925,17 @@ class CDPSession:

821 925 # Wait for network activity to settle before capturing

822 926 await self.wait_for_network_idle(idle_time=5, timeout=15.0)

823 927 # Try to click any "Accept" or "Einwilligen" cookie consent buttons

824 if n != 1:

928 if n < 3:

825 929 for _ in range(2):

826 930 debug.log("Attempting to click accept button...")

827 931 await asyncio.sleep(1)

828 932 if await self.click_accept_button():

829 933 debug.log("Clicked accept button.")

830 await asyncio.sleep(1)

831 934 break

935 print(url_without_suffix)

936

        if ("&headless=false" in url_without_suffix or "&sleep=" in url_without_suffix or "&wait=" in url_without_suffix) and n == 3:

937 debug.log("Waiting 5 seconds for page to settle due to sleep/wait parameter...")

938 await asyncio.sleep(30)

832 939 await self.wait_for_network_idle(idle_time=5, timeout=15.0)

833 940 result = await self.call("Page.captureScreenshot")

834 941 image_bytes = base64.b64decode(result["data"])