lumivor 0.1.7__py3-none-any.whl
This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.
- lumivor/README.md +51 -0
- lumivor/__init__.py +25 -0
- lumivor/agent/message_manager/service.py +252 -0
- lumivor/agent/message_manager/tests.py +246 -0
- lumivor/agent/message_manager/views.py +37 -0
- lumivor/agent/prompts.py +208 -0
- lumivor/agent/service.py +1017 -0
- lumivor/agent/tests.py +204 -0
- lumivor/agent/views.py +272 -0
- lumivor/browser/browser.py +208 -0
- lumivor/browser/context.py +993 -0
- lumivor/browser/tests/screenshot_test.py +38 -0
- lumivor/browser/tests/test_clicks.py +77 -0
- lumivor/browser/views.py +48 -0
- lumivor/controller/registry/service.py +140 -0
- lumivor/controller/registry/views.py +71 -0
- lumivor/controller/service.py +557 -0
- lumivor/controller/views.py +47 -0
- lumivor/dom/__init__.py +0 -0
- lumivor/dom/buildDomTree.js +428 -0
- lumivor/dom/history_tree_processor/service.py +112 -0
- lumivor/dom/history_tree_processor/view.py +33 -0
- lumivor/dom/service.py +100 -0
- lumivor/dom/tests/extraction_test.py +44 -0
- lumivor/dom/tests/process_dom_test.py +40 -0
- lumivor/dom/views.py +187 -0
- lumivor/logging_config.py +128 -0
- lumivor/telemetry/service.py +114 -0
- lumivor/telemetry/views.py +51 -0
- lumivor/utils.py +54 -0
- lumivor-0.1.7.dist-info/METADATA +100 -0
- lumivor-0.1.7.dist-info/RECORD +34 -0
- lumivor-0.1.7.dist-info/WHEEL +4 -0
- lumivor-0.1.7.dist-info/licenses/LICENSE +21 -0
@@ -0,0 +1,993 @@
|
|
1
|
+
"""
|
2
|
+
Playwright browser on steroids.
|
3
|
+
"""
|
4
|
+
|
5
|
+
import asyncio
|
6
|
+
import base64
|
7
|
+
import json
|
8
|
+
import logging
|
9
|
+
import os
|
10
|
+
import re
|
11
|
+
import time
|
12
|
+
import uuid
|
13
|
+
from dataclasses import dataclass, field
|
14
|
+
from typing import TYPE_CHECKING, Optional, TypedDict
|
15
|
+
|
16
|
+
from playwright.async_api import Browser as PlaywrightBrowser
|
17
|
+
from playwright.async_api import (
|
18
|
+
BrowserContext as PlaywrightBrowserContext,
|
19
|
+
)
|
20
|
+
from playwright.async_api import (
|
21
|
+
ElementHandle,
|
22
|
+
FrameLocator,
|
23
|
+
Page,
|
24
|
+
)
|
25
|
+
|
26
|
+
from lumivor.browser.views import BrowserError, BrowserState, TabInfo
|
27
|
+
from lumivor.dom.service import DomService
|
28
|
+
from lumivor.dom.views import DOMElementNode, SelectorMap
|
29
|
+
from lumivor.utils import time_execution_sync
|
30
|
+
|
31
|
+
if TYPE_CHECKING:
|
32
|
+
from lumivor.browser.browser import Browser
|
33
|
+
|
34
|
+
logger = logging.getLogger(__name__)
|
35
|
+
|
36
|
+
|
37
|
+
class BrowserContextWindowSize(TypedDict):
|
38
|
+
width: int
|
39
|
+
height: int
|
40
|
+
|
41
|
+
|
42
|
+
@dataclass
|
43
|
+
class BrowserContextConfig:
|
44
|
+
"""
|
45
|
+
Configuration for the BrowserContext.
|
46
|
+
|
47
|
+
Default values:
|
48
|
+
cookies_file: None
|
49
|
+
Path to cookies file for persistence
|
50
|
+
|
51
|
+
disable_security: False
|
52
|
+
Disable browser security features
|
53
|
+
|
54
|
+
minimum_wait_page_load_time: 0.5
|
55
|
+
Minimum time to wait before getting page state for LLM input
|
56
|
+
|
57
|
+
wait_for_network_idle_page_load_time: 1.0
|
58
|
+
Time to wait for network requests to finish before getting page state.
|
59
|
+
Lower values may result in incomplete page loads.
|
60
|
+
|
61
|
+
maximum_wait_page_load_time: 5.0
|
62
|
+
Maximum time to wait for page load before proceeding anyway
|
63
|
+
|
64
|
+
wait_between_actions: 1.0
|
65
|
+
Time to wait between multiple per step actions
|
66
|
+
|
67
|
+
browser_window_size: {
|
68
|
+
'width': 1280,
|
69
|
+
'height': 1100,
|
70
|
+
}
|
71
|
+
Default browser window size
|
72
|
+
|
73
|
+
no_viewport: False
|
74
|
+
Disable viewport
|
75
|
+
save_recording_path: None
|
76
|
+
Path to save video recordings
|
77
|
+
|
78
|
+
trace_path: None
|
79
|
+
Path to save trace files. It will auto name the file with the TRACE_PATH/{context_id}.zip
|
80
|
+
"""
|
81
|
+
|
82
|
+
cookies_file: str | None = None
|
83
|
+
minimum_wait_page_load_time: float = 0.5
|
84
|
+
wait_for_network_idle_page_load_time: float = 1
|
85
|
+
maximum_wait_page_load_time: float = 5
|
86
|
+
wait_between_actions: float = 1
|
87
|
+
|
88
|
+
disable_security: bool = False
|
89
|
+
|
90
|
+
browser_window_size: BrowserContextWindowSize = field(
|
91
|
+
default_factory=lambda: {'width': 1280, 'height': 1100}
|
92
|
+
)
|
93
|
+
no_viewport: Optional[bool] = None
|
94
|
+
|
95
|
+
save_recording_path: str | None = None
|
96
|
+
trace_path: str | None = None
|
97
|
+
|
98
|
+
|
99
|
+
@dataclass
|
100
|
+
class BrowserSession:
|
101
|
+
context: PlaywrightBrowserContext
|
102
|
+
current_page: Page
|
103
|
+
cached_state: BrowserState
|
104
|
+
|
105
|
+
|
106
|
+
class BrowserContext:
|
107
|
+
def __init__(
|
108
|
+
self,
|
109
|
+
browser: 'Browser',
|
110
|
+
config: BrowserContextConfig = BrowserContextConfig(),
|
111
|
+
):
|
112
|
+
self.context_id = str(uuid.uuid4())
|
113
|
+
logger.debug(f'Initializing new browser context with id: {
|
114
|
+
self.context_id}')
|
115
|
+
|
116
|
+
self.config = config
|
117
|
+
self.browser = browser
|
118
|
+
|
119
|
+
# Initialize these as None - they'll be set up when needed
|
120
|
+
self.session: BrowserSession | None = None
|
121
|
+
|
122
|
+
async def __aenter__(self):
|
123
|
+
"""Async context manager entry"""
|
124
|
+
await self._initialize_session()
|
125
|
+
return self
|
126
|
+
|
127
|
+
async def __aexit__(self, exc_type, exc_val, exc_tb):
|
128
|
+
"""Async context manager exit"""
|
129
|
+
await self.close()
|
130
|
+
|
131
|
+
async def close(self):
|
132
|
+
"""Close the browser instance"""
|
133
|
+
logger.debug('Closing browser context')
|
134
|
+
|
135
|
+
try:
|
136
|
+
# check if already closed
|
137
|
+
if self.session is None:
|
138
|
+
return
|
139
|
+
|
140
|
+
await self.save_cookies()
|
141
|
+
|
142
|
+
if self.config.trace_path:
|
143
|
+
try:
|
144
|
+
await self.session.context.tracing.stop(
|
145
|
+
path=os.path.join(self.config.trace_path, f'{
|
146
|
+
self.context_id}.zip')
|
147
|
+
)
|
148
|
+
except Exception as e:
|
149
|
+
logger.debug(f'Failed to stop tracing: {e}')
|
150
|
+
|
151
|
+
try:
|
152
|
+
await self.session.context.close()
|
153
|
+
except Exception as e:
|
154
|
+
logger.debug(f'Failed to close context: {e}')
|
155
|
+
finally:
|
156
|
+
self.session = None
|
157
|
+
|
158
|
+
def __del__(self):
|
159
|
+
"""Cleanup when object is destroyed"""
|
160
|
+
if self.session is not None:
|
161
|
+
logger.debug(
|
162
|
+
'BrowserContext was not properly closed before destruction')
|
163
|
+
try:
|
164
|
+
# Use sync Playwright method for force cleanup
|
165
|
+
if hasattr(self.session.context, '_impl_obj'):
|
166
|
+
asyncio.run(self.session.context._impl_obj.close())
|
167
|
+
self.session = None
|
168
|
+
except Exception as e:
|
169
|
+
logger.warning(f'Failed to force close browser context: {e}')
|
170
|
+
|
171
|
+
async def _initialize_session(self):
|
172
|
+
"""Initialize the browser session"""
|
173
|
+
logger.debug('Initializing browser context')
|
174
|
+
|
175
|
+
playwright_browser = await self.browser.get_playwright_browser()
|
176
|
+
|
177
|
+
context = await self._create_context(playwright_browser)
|
178
|
+
page = await context.new_page()
|
179
|
+
|
180
|
+
# Instead of calling _update_state(), create an empty initial state
|
181
|
+
initial_state = BrowserState(
|
182
|
+
element_tree=DOMElementNode(
|
183
|
+
tag_name='root',
|
184
|
+
is_visible=True,
|
185
|
+
parent=None,
|
186
|
+
xpath='',
|
187
|
+
attributes={},
|
188
|
+
children=[],
|
189
|
+
),
|
190
|
+
selector_map={},
|
191
|
+
url=page.url,
|
192
|
+
title=await page.title(),
|
193
|
+
screenshot=None,
|
194
|
+
tabs=[],
|
195
|
+
)
|
196
|
+
|
197
|
+
self.session = BrowserSession(
|
198
|
+
context=context,
|
199
|
+
current_page=page,
|
200
|
+
cached_state=initial_state,
|
201
|
+
)
|
202
|
+
|
203
|
+
await self._add_new_page_listener(context)
|
204
|
+
|
205
|
+
return self.session
|
206
|
+
|
207
|
+
async def _add_new_page_listener(self, context: PlaywrightBrowserContext):
|
208
|
+
async def on_page(page: Page):
|
209
|
+
await page.wait_for_load_state()
|
210
|
+
logger.debug(f'New page opened: {page.url}')
|
211
|
+
if self.session is not None:
|
212
|
+
self.session.current_page = page
|
213
|
+
|
214
|
+
context.on('page', on_page)
|
215
|
+
|
216
|
+
async def get_session(self) -> BrowserSession:
|
217
|
+
"""Lazy initialization of the browser and related components"""
|
218
|
+
if self.session is None:
|
219
|
+
return await self._initialize_session()
|
220
|
+
return self.session
|
221
|
+
|
222
|
+
async def get_current_page(self) -> Page:
|
223
|
+
"""Get the current page"""
|
224
|
+
session = await self.get_session()
|
225
|
+
return session.current_page
|
226
|
+
|
227
|
+
async def _create_context(self, browser: PlaywrightBrowser):
|
228
|
+
"""Creates a new browser context with anti-detection measures and loads cookies if available."""
|
229
|
+
if self.browser.config.chrome_instance_path and len(browser.contexts) > 0:
|
230
|
+
# Connect to existing Chrome instance instead of creating new one
|
231
|
+
context = browser.contexts[0]
|
232
|
+
else:
|
233
|
+
# Original code for creating new context
|
234
|
+
context = await browser.new_context(
|
235
|
+
viewport=self.config.browser_window_size,
|
236
|
+
no_viewport=False,
|
237
|
+
user_agent=(
|
238
|
+
'Mozilla/5.0 (Windows NT 10.0; Win64; x64) AppleWebKit/537.36 '
|
239
|
+
'(KHTML, like Gecko) Chrome/85.0.4183.102 Safari/537.36'
|
240
|
+
),
|
241
|
+
java_script_enabled=True,
|
242
|
+
bypass_csp=self.config.disable_security,
|
243
|
+
ignore_https_errors=self.config.disable_security,
|
244
|
+
record_video_dir=self.config.save_recording_path,
|
245
|
+
)
|
246
|
+
|
247
|
+
if self.config.trace_path:
|
248
|
+
await context.tracing.start(screenshots=True, snapshots=True, sources=True)
|
249
|
+
|
250
|
+
# Load cookies if they exist
|
251
|
+
if self.config.cookies_file and os.path.exists(self.config.cookies_file):
|
252
|
+
with open(self.config.cookies_file, 'r') as f:
|
253
|
+
cookies = json.load(f)
|
254
|
+
logger.info(f'Loaded {len(cookies)} cookies from {
|
255
|
+
self.config.cookies_file}')
|
256
|
+
await context.add_cookies(cookies)
|
257
|
+
|
258
|
+
# Expose anti-detection scripts
|
259
|
+
await context.add_init_script(
|
260
|
+
"""
|
261
|
+
// Webdriver property
|
262
|
+
Object.defineProperty(navigator, 'webdriver', {
|
263
|
+
get: () => undefined
|
264
|
+
});
|
265
|
+
|
266
|
+
// Languages
|
267
|
+
Object.defineProperty(navigator, 'languages', {
|
268
|
+
get: () => ['en-US', 'en']
|
269
|
+
});
|
270
|
+
|
271
|
+
// Plugins
|
272
|
+
Object.defineProperty(navigator, 'plugins', {
|
273
|
+
get: () => [1, 2, 3, 4, 5]
|
274
|
+
});
|
275
|
+
|
276
|
+
// Chrome runtime
|
277
|
+
window.chrome = { runtime: {} };
|
278
|
+
|
279
|
+
// Permissions
|
280
|
+
const originalQuery = window.navigator.permissions.query;
|
281
|
+
window.navigator.permissions.query = (parameters) => (
|
282
|
+
parameters.name === 'notifications' ?
|
283
|
+
Promise.resolve({ state: Notification.permission }) :
|
284
|
+
originalQuery(parameters)
|
285
|
+
);
|
286
|
+
"""
|
287
|
+
)
|
288
|
+
|
289
|
+
return context
|
290
|
+
|
291
|
+
async def _wait_for_stable_network(self):
|
292
|
+
page = await self.get_current_page()
|
293
|
+
|
294
|
+
pending_requests = set()
|
295
|
+
last_activity = asyncio.get_event_loop().time()
|
296
|
+
|
297
|
+
# Define relevant resource types and content types
|
298
|
+
RELEVANT_RESOURCE_TYPES = {
|
299
|
+
'document',
|
300
|
+
'stylesheet',
|
301
|
+
'image',
|
302
|
+
'font',
|
303
|
+
'script',
|
304
|
+
'iframe',
|
305
|
+
}
|
306
|
+
|
307
|
+
RELEVANT_CONTENT_TYPES = {
|
308
|
+
'text/html',
|
309
|
+
'text/css',
|
310
|
+
'application/javascript',
|
311
|
+
'image/',
|
312
|
+
'font/',
|
313
|
+
'application/json',
|
314
|
+
}
|
315
|
+
|
316
|
+
# Additional patterns to filter out
|
317
|
+
IGNORED_URL_PATTERNS = {
|
318
|
+
# Analytics and tracking
|
319
|
+
'analytics',
|
320
|
+
'tracking',
|
321
|
+
'telemetry',
|
322
|
+
'beacon',
|
323
|
+
'metrics',
|
324
|
+
# Ad-related
|
325
|
+
'doubleclick',
|
326
|
+
'adsystem',
|
327
|
+
'adserver',
|
328
|
+
'advertising',
|
329
|
+
# Social media widgets
|
330
|
+
'facebook.com/plugins',
|
331
|
+
'platform.twitter',
|
332
|
+
'linkedin.com/embed',
|
333
|
+
# Live chat and support
|
334
|
+
'livechat',
|
335
|
+
'zendesk',
|
336
|
+
'intercom',
|
337
|
+
'crisp.chat',
|
338
|
+
'hotjar',
|
339
|
+
# Push notifications
|
340
|
+
'push-notifications',
|
341
|
+
'onesignal',
|
342
|
+
'pushwoosh',
|
343
|
+
# Background sync/heartbeat
|
344
|
+
'heartbeat',
|
345
|
+
'ping',
|
346
|
+
'alive',
|
347
|
+
# WebRTC and streaming
|
348
|
+
'webrtc',
|
349
|
+
'rtmp://',
|
350
|
+
'wss://',
|
351
|
+
# Common CDNs for dynamic content
|
352
|
+
'cloudfront.net',
|
353
|
+
'fastly.net',
|
354
|
+
}
|
355
|
+
|
356
|
+
async def on_request(request):
|
357
|
+
# Filter by resource type
|
358
|
+
if request.resource_type not in RELEVANT_RESOURCE_TYPES:
|
359
|
+
return
|
360
|
+
|
361
|
+
# Filter out streaming, websocket, and other real-time requests
|
362
|
+
if request.resource_type in {
|
363
|
+
'websocket',
|
364
|
+
'media',
|
365
|
+
'eventsource',
|
366
|
+
'manifest',
|
367
|
+
'other',
|
368
|
+
}:
|
369
|
+
return
|
370
|
+
|
371
|
+
# Filter out by URL patterns
|
372
|
+
url = request.url.lower()
|
373
|
+
if any(pattern in url for pattern in IGNORED_URL_PATTERNS):
|
374
|
+
return
|
375
|
+
|
376
|
+
# Filter out data URLs and blob URLs
|
377
|
+
if url.startswith(('data:', 'blob:')):
|
378
|
+
return
|
379
|
+
|
380
|
+
# Filter out requests with certain headers
|
381
|
+
headers = request.headers
|
382
|
+
if headers.get('purpose') == 'prefetch' or headers.get('sec-fetch-dest') in [
|
383
|
+
'video',
|
384
|
+
'audio',
|
385
|
+
]:
|
386
|
+
return
|
387
|
+
|
388
|
+
nonlocal last_activity
|
389
|
+
pending_requests.add(request)
|
390
|
+
last_activity = asyncio.get_event_loop().time()
|
391
|
+
# logger.debug(f'Request started: {request.url} ({request.resource_type})')
|
392
|
+
|
393
|
+
async def on_response(response):
|
394
|
+
request = response.request
|
395
|
+
if request not in pending_requests:
|
396
|
+
return
|
397
|
+
|
398
|
+
# Filter by content type if available
|
399
|
+
content_type = response.headers.get('content-type', '').lower()
|
400
|
+
|
401
|
+
# Skip if content type indicates streaming or real-time data
|
402
|
+
if any(
|
403
|
+
t in content_type
|
404
|
+
for t in [
|
405
|
+
'streaming',
|
406
|
+
'video',
|
407
|
+
'audio',
|
408
|
+
'webm',
|
409
|
+
'mp4',
|
410
|
+
'event-stream',
|
411
|
+
'websocket',
|
412
|
+
'protobuf',
|
413
|
+
]
|
414
|
+
):
|
415
|
+
pending_requests.remove(request)
|
416
|
+
return
|
417
|
+
|
418
|
+
# Only process relevant content types
|
419
|
+
if not any(ct in content_type for ct in RELEVANT_CONTENT_TYPES):
|
420
|
+
pending_requests.remove(request)
|
421
|
+
return
|
422
|
+
|
423
|
+
# Skip if response is too large (likely not essential for page load)
|
424
|
+
content_length = response.headers.get('content-length')
|
425
|
+
if content_length and int(content_length) > 5 * 1024 * 1024: # 5MB
|
426
|
+
pending_requests.remove(request)
|
427
|
+
return
|
428
|
+
|
429
|
+
nonlocal last_activity
|
430
|
+
pending_requests.remove(request)
|
431
|
+
last_activity = asyncio.get_event_loop().time()
|
432
|
+
# logger.debug(f'Request resolved: {request.url} ({content_type})')
|
433
|
+
|
434
|
+
# Attach event listeners
|
435
|
+
page.on('request', on_request)
|
436
|
+
page.on('response', on_response)
|
437
|
+
|
438
|
+
try:
|
439
|
+
# Wait for idle time
|
440
|
+
start_time = asyncio.get_event_loop().time()
|
441
|
+
while True:
|
442
|
+
await asyncio.sleep(0.1)
|
443
|
+
now = asyncio.get_event_loop().time()
|
444
|
+
if (
|
445
|
+
len(pending_requests) == 0
|
446
|
+
and (now - last_activity) >= self.config.wait_for_network_idle_page_load_time
|
447
|
+
):
|
448
|
+
break
|
449
|
+
if now - start_time > self.config.maximum_wait_page_load_time:
|
450
|
+
logger.debug(
|
451
|
+
f'Network timeout after {self.config.maximum_wait_page_load_time}s with {
|
452
|
+
len(pending_requests)} '
|
453
|
+
f'pending requests: {
|
454
|
+
[r.url for r in pending_requests]}'
|
455
|
+
)
|
456
|
+
break
|
457
|
+
|
458
|
+
finally:
|
459
|
+
# Clean up event listeners
|
460
|
+
page.remove_listener('request', on_request)
|
461
|
+
page.remove_listener('response', on_response)
|
462
|
+
|
463
|
+
logger.debug(
|
464
|
+
f'Network stabilized for {
|
465
|
+
self.config.wait_for_network_idle_page_load_time} seconds'
|
466
|
+
)
|
467
|
+
|
468
|
+
async def _wait_for_page_and_frames_load(self, timeout_overwrite: float | None = None):
|
469
|
+
"""
|
470
|
+
Ensures page is fully loaded before continuing.
|
471
|
+
Waits for either network to be idle or minimum WAIT_TIME, whichever is longer.
|
472
|
+
"""
|
473
|
+
# Start timing
|
474
|
+
start_time = time.time()
|
475
|
+
|
476
|
+
# await asyncio.sleep(self.minimum_wait_page_load_time)
|
477
|
+
|
478
|
+
# Wait for page load
|
479
|
+
try:
|
480
|
+
await self._wait_for_stable_network()
|
481
|
+
except Exception:
|
482
|
+
logger.warning('Page load failed, continuing...')
|
483
|
+
pass
|
484
|
+
|
485
|
+
# Calculate remaining time to meet minimum WAIT_TIME
|
486
|
+
elapsed = time.time() - start_time
|
487
|
+
remaining = max(
|
488
|
+
(timeout_overwrite or self.config.minimum_wait_page_load_time) - elapsed, 0)
|
489
|
+
|
490
|
+
logger.debug(
|
491
|
+
f'--Page loaded in {elapsed:.2f} seconds, waiting for additional {
|
492
|
+
remaining:.2f} seconds'
|
493
|
+
)
|
494
|
+
|
495
|
+
# Sleep remaining time if needed
|
496
|
+
if remaining > 0:
|
497
|
+
await asyncio.sleep(remaining)
|
498
|
+
|
499
|
+
async def navigate_to(self, url: str):
|
500
|
+
"""Navigate to a URL"""
|
501
|
+
page = await self.get_current_page()
|
502
|
+
await page.goto(url)
|
503
|
+
await page.wait_for_load_state()
|
504
|
+
|
505
|
+
async def refresh_page(self):
|
506
|
+
"""Refresh the current page"""
|
507
|
+
page = await self.get_current_page()
|
508
|
+
await page.reload()
|
509
|
+
await page.wait_for_load_state()
|
510
|
+
|
511
|
+
async def go_back(self):
|
512
|
+
"""Navigate back in history"""
|
513
|
+
page = await self.get_current_page()
|
514
|
+
await page.go_back()
|
515
|
+
await page.wait_for_load_state()
|
516
|
+
|
517
|
+
async def go_forward(self):
|
518
|
+
"""Navigate forward in history"""
|
519
|
+
page = await self.get_current_page()
|
520
|
+
await page.go_forward()
|
521
|
+
await page.wait_for_load_state()
|
522
|
+
|
523
|
+
async def close_current_tab(self):
|
524
|
+
"""Close the current tab"""
|
525
|
+
session = await self.get_session()
|
526
|
+
page = session.current_page
|
527
|
+
await page.close()
|
528
|
+
|
529
|
+
# Switch to the first available tab if any exist
|
530
|
+
if session.context.pages:
|
531
|
+
await self.switch_to_tab(0)
|
532
|
+
|
533
|
+
# otherwise the browser will be closed
|
534
|
+
|
535
|
+
async def get_page_html(self) -> str:
|
536
|
+
"""Get the current page HTML content"""
|
537
|
+
page = await self.get_current_page()
|
538
|
+
return await page.content()
|
539
|
+
|
540
|
+
async def execute_javascript(self, script: str):
|
541
|
+
"""Execute JavaScript code on the page"""
|
542
|
+
page = await self.get_current_page()
|
543
|
+
return await page.evaluate(script)
|
544
|
+
|
545
|
+
# This decorator might need to be updated to handle async
|
546
|
+
@time_execution_sync('--get_state')
|
547
|
+
async def get_state(self, use_vision: bool = False) -> BrowserState:
|
548
|
+
"""Get the current state of the browser"""
|
549
|
+
await self._wait_for_page_and_frames_load()
|
550
|
+
session = await self.get_session()
|
551
|
+
session.cached_state = await self._update_state(use_vision=use_vision)
|
552
|
+
|
553
|
+
# Save cookies if a file is specified
|
554
|
+
if self.config.cookies_file:
|
555
|
+
asyncio.create_task(self.save_cookies())
|
556
|
+
|
557
|
+
return session.cached_state
|
558
|
+
|
559
|
+
async def _update_state(self, use_vision: bool = False) -> BrowserState:
|
560
|
+
"""Update and return state."""
|
561
|
+
session = await self.get_session()
|
562
|
+
|
563
|
+
# Check if current page is still valid, if not switch to another available page
|
564
|
+
try:
|
565
|
+
page = await self.get_current_page()
|
566
|
+
# Test if page is still accessible
|
567
|
+
await page.evaluate('1')
|
568
|
+
except Exception as e:
|
569
|
+
logger.debug(f'Current page is no longer accessible: {str(e)}')
|
570
|
+
# Get all available pages
|
571
|
+
pages = session.context.pages
|
572
|
+
if pages:
|
573
|
+
session.current_page = pages[-1]
|
574
|
+
page = session.current_page
|
575
|
+
logger.debug(f'Switched to page: {await page.title()}')
|
576
|
+
else:
|
577
|
+
raise BrowserError('Browser closed: no valid pages available')
|
578
|
+
|
579
|
+
try:
|
580
|
+
await self.remove_highlights()
|
581
|
+
dom_service = DomService(page)
|
582
|
+
content = await dom_service.get_clickable_elements()
|
583
|
+
|
584
|
+
screenshot_b64 = None
|
585
|
+
if use_vision:
|
586
|
+
screenshot_b64 = await self.take_screenshot()
|
587
|
+
|
588
|
+
self.current_state = BrowserState(
|
589
|
+
element_tree=content.element_tree,
|
590
|
+
selector_map=content.selector_map,
|
591
|
+
url=page.url,
|
592
|
+
title=await page.title(),
|
593
|
+
tabs=await self.get_tabs_info(),
|
594
|
+
screenshot=screenshot_b64,
|
595
|
+
)
|
596
|
+
|
597
|
+
return self.current_state
|
598
|
+
except Exception as e:
|
599
|
+
logger.error(f'Failed to update state: {str(e)}')
|
600
|
+
# Return last known good state if available
|
601
|
+
if hasattr(self, 'current_state'):
|
602
|
+
return self.current_state
|
603
|
+
raise
|
604
|
+
|
605
|
+
# region - Browser Actions
|
606
|
+
|
607
|
+
async def take_screenshot(self, full_page: bool = False) -> str:
|
608
|
+
"""
|
609
|
+
Returns a base64 encoded screenshot of the current page.
|
610
|
+
"""
|
611
|
+
page = await self.get_current_page()
|
612
|
+
|
613
|
+
screenshot = await page.screenshot(
|
614
|
+
full_page=full_page,
|
615
|
+
animations='disabled',
|
616
|
+
)
|
617
|
+
|
618
|
+
screenshot_b64 = base64.b64encode(screenshot).decode('utf-8')
|
619
|
+
|
620
|
+
# await self.remove_highlights()
|
621
|
+
|
622
|
+
return screenshot_b64
|
623
|
+
|
624
|
+
async def remove_highlights(self):
|
625
|
+
"""
|
626
|
+
Removes all highlight overlays and labels created by the highlightElement function.
|
627
|
+
Handles cases where the page might be closed or inaccessible.
|
628
|
+
"""
|
629
|
+
try:
|
630
|
+
page = await self.get_current_page()
|
631
|
+
await page.evaluate(
|
632
|
+
"""
|
633
|
+
try {
|
634
|
+
// Remove the highlight container and all its contents
|
635
|
+
const container = document.getElementById('playwright-highlight-container');
|
636
|
+
if (container) {
|
637
|
+
container.remove();
|
638
|
+
}
|
639
|
+
|
640
|
+
// Remove highlight attributes from elements
|
641
|
+
const highlightedElements = document.querySelectorAll('[browser-user-highlight-id^="playwright-highlight-"]');
|
642
|
+
highlightedElements.forEach(el => {
|
643
|
+
el.removeAttribute('browser-user-highlight-id');
|
644
|
+
});
|
645
|
+
} catch (e) {
|
646
|
+
console.error('Failed to remove highlights:', e);
|
647
|
+
}
|
648
|
+
"""
|
649
|
+
)
|
650
|
+
except Exception as e:
|
651
|
+
logger.debug(
|
652
|
+
f'Failed to remove highlights (this is usually ok): {str(e)}')
|
653
|
+
# Don't raise the error since this is not critical functionality
|
654
|
+
pass
|
655
|
+
|
656
|
+
# endregion
|
657
|
+
|
658
|
+
# region - User Actions
|
659
|
+
def _convert_simple_xpath_to_css_selector(self, xpath: str) -> str:
|
660
|
+
"""Converts simple XPath expressions to CSS selectors."""
|
661
|
+
if not xpath:
|
662
|
+
return ''
|
663
|
+
|
664
|
+
# Remove leading slash if present
|
665
|
+
xpath = xpath.lstrip('/')
|
666
|
+
|
667
|
+
# Split into parts
|
668
|
+
parts = xpath.split('/')
|
669
|
+
css_parts = []
|
670
|
+
|
671
|
+
for part in parts:
|
672
|
+
if not part:
|
673
|
+
continue
|
674
|
+
|
675
|
+
# Handle index notation [n]
|
676
|
+
if '[' in part:
|
677
|
+
base_part = part[: part.find('[')]
|
678
|
+
index_part = part[part.find('['):]
|
679
|
+
|
680
|
+
# Handle multiple indices
|
681
|
+
indices = [i.strip('[]') for i in index_part.split(']')[:-1]]
|
682
|
+
|
683
|
+
for idx in indices:
|
684
|
+
try:
|
685
|
+
# Handle numeric indices
|
686
|
+
if idx.isdigit():
|
687
|
+
index = int(idx) - 1
|
688
|
+
base_part += f':nth-of-type({index+1})'
|
689
|
+
# Handle last() function
|
690
|
+
elif idx == 'last()':
|
691
|
+
base_part += ':last-of-type'
|
692
|
+
# Handle position() functions
|
693
|
+
elif 'position()' in idx:
|
694
|
+
if '>1' in idx:
|
695
|
+
base_part += ':nth-of-type(n+2)'
|
696
|
+
except ValueError:
|
697
|
+
continue
|
698
|
+
|
699
|
+
css_parts.append(base_part)
|
700
|
+
else:
|
701
|
+
css_parts.append(part)
|
702
|
+
|
703
|
+
base_selector = ' > '.join(css_parts)
|
704
|
+
return base_selector
|
705
|
+
|
706
|
+
def _enhanced_css_selector_for_element(self, element: DOMElementNode) -> str:
|
707
|
+
"""
|
708
|
+
Creates a CSS selector for a DOM element, handling various edge cases and special characters.
|
709
|
+
|
710
|
+
Args:
|
711
|
+
element: The DOM element to create a selector for
|
712
|
+
|
713
|
+
Returns:
|
714
|
+
A valid CSS selector string
|
715
|
+
"""
|
716
|
+
try:
|
717
|
+
# Get base selector from XPath
|
718
|
+
css_selector = self._convert_simple_xpath_to_css_selector(
|
719
|
+
element.xpath)
|
720
|
+
|
721
|
+
# Handle class attributes
|
722
|
+
if 'class' in element.attributes and element.attributes['class']:
|
723
|
+
# Define a regex pattern for valid class names in CSS
|
724
|
+
valid_class_name_pattern = re.compile(
|
725
|
+
r'^[a-zA-Z_][a-zA-Z0-9_-]*$')
|
726
|
+
|
727
|
+
# Iterate through the class attribute values
|
728
|
+
classes = element.attributes['class'].split()
|
729
|
+
for class_name in classes:
|
730
|
+
# Skip empty class names
|
731
|
+
if not class_name.strip():
|
732
|
+
continue
|
733
|
+
|
734
|
+
# Check if the class name is valid
|
735
|
+
if valid_class_name_pattern.match(class_name):
|
736
|
+
# Append the valid class name to the CSS selector
|
737
|
+
css_selector += f'.{class_name}'
|
738
|
+
else:
|
739
|
+
# Skip invalid class names
|
740
|
+
continue
|
741
|
+
|
742
|
+
# Expanded set of safe attributes that are stable and useful for selection
|
743
|
+
SAFE_ATTRIBUTES = {
|
744
|
+
# Standard HTML attributes
|
745
|
+
'id',
|
746
|
+
'name',
|
747
|
+
'type',
|
748
|
+
'value',
|
749
|
+
'placeholder',
|
750
|
+
# Accessibility attributes
|
751
|
+
'aria-label',
|
752
|
+
'aria-labelledby',
|
753
|
+
'aria-describedby',
|
754
|
+
'role',
|
755
|
+
# Common form attributes
|
756
|
+
'for',
|
757
|
+
'autocomplete',
|
758
|
+
'required',
|
759
|
+
'readonly',
|
760
|
+
# Media attributes
|
761
|
+
'alt',
|
762
|
+
'title',
|
763
|
+
'src',
|
764
|
+
# Data attributes (if they're stable in your application)
|
765
|
+
'data-testid',
|
766
|
+
'data-id',
|
767
|
+
'data-qa',
|
768
|
+
'data-cy',
|
769
|
+
# Custom stable attributes (add any application-specific ones)
|
770
|
+
'href',
|
771
|
+
'target',
|
772
|
+
}
|
773
|
+
|
774
|
+
# Handle other attributes
|
775
|
+
for attribute, value in element.attributes.items():
|
776
|
+
if attribute == 'class':
|
777
|
+
continue
|
778
|
+
|
779
|
+
# Skip invalid attribute names
|
780
|
+
if not attribute.strip():
|
781
|
+
continue
|
782
|
+
|
783
|
+
if attribute not in SAFE_ATTRIBUTES:
|
784
|
+
continue
|
785
|
+
|
786
|
+
# Escape special characters in attribute names
|
787
|
+
safe_attribute = attribute.replace(':', r'\:')
|
788
|
+
|
789
|
+
# Handle different value cases
|
790
|
+
if value == '':
|
791
|
+
css_selector += f'[{safe_attribute}]'
|
792
|
+
elif any(char in value for char in '"\'<>`'):
|
793
|
+
# Use contains for values with special characters
|
794
|
+
safe_value = value.replace('"', '\\"')
|
795
|
+
css_selector += f'[{safe_attribute}*="{safe_value}"]'
|
796
|
+
else:
|
797
|
+
css_selector += f'[{safe_attribute}="{value}"]'
|
798
|
+
|
799
|
+
return css_selector
|
800
|
+
|
801
|
+
except Exception:
|
802
|
+
# Fallback to a more basic selector if something goes wrong
|
803
|
+
tag_name = element.tag_name or '*'
|
804
|
+
return f"{tag_name}[highlight_index='{element.highlight_index}']"
|
805
|
+
|
806
|
+
async def get_locate_element(self, element: DOMElementNode) -> ElementHandle | None:
|
807
|
+
current_frame = await self.get_current_page()
|
808
|
+
|
809
|
+
# Start with the target element and collect all parents
|
810
|
+
parents: list[DOMElementNode] = []
|
811
|
+
current = element
|
812
|
+
while current.parent is not None:
|
813
|
+
parent = current.parent
|
814
|
+
parents.append(parent)
|
815
|
+
current = parent
|
816
|
+
if parent.tag_name == 'iframe':
|
817
|
+
break
|
818
|
+
|
819
|
+
# There can be only one iframe parent (by design of the loop above)
|
820
|
+
iframe_parent = [item for item in parents if item.tag_name == 'iframe']
|
821
|
+
if iframe_parent:
|
822
|
+
parent = iframe_parent[0]
|
823
|
+
css_selector = self._enhanced_css_selector_for_element(parent)
|
824
|
+
current_frame = current_frame.frame_locator(css_selector)
|
825
|
+
|
826
|
+
css_selector = self._enhanced_css_selector_for_element(element)
|
827
|
+
|
828
|
+
try:
|
829
|
+
if isinstance(current_frame, FrameLocator):
|
830
|
+
return await current_frame.locator(css_selector).element_handle()
|
831
|
+
else:
|
832
|
+
# Try to scroll into view if hidden
|
833
|
+
element_handle = await current_frame.query_selector(css_selector)
|
834
|
+
if element_handle:
|
835
|
+
await element_handle.scroll_into_view_if_needed()
|
836
|
+
return element_handle
|
837
|
+
except Exception as e:
|
838
|
+
logger.error(f'Failed to locate element: {str(e)}')
|
839
|
+
return None
|
840
|
+
|
841
|
+
async def _input_text_element_node(self, element_node: DOMElementNode, text: str):
|
842
|
+
try:
|
843
|
+
page = await self.get_current_page()
|
844
|
+
element = await self.get_locate_element(element_node)
|
845
|
+
|
846
|
+
if element is None:
|
847
|
+
raise Exception(f'Element: {repr(element_node)} not found')
|
848
|
+
|
849
|
+
await element.scroll_into_view_if_needed(timeout=2500)
|
850
|
+
await element.fill('')
|
851
|
+
await element.type(text)
|
852
|
+
await page.wait_for_load_state()
|
853
|
+
|
854
|
+
except Exception as e:
|
855
|
+
raise Exception(
|
856
|
+
f'Failed to input text into element: {
|
857
|
+
repr(element_node)}. Error: {str(e)}'
|
858
|
+
)
|
859
|
+
|
860
|
+
async def _click_element_node(self, element_node: DOMElementNode):
|
861
|
+
"""
|
862
|
+
Optimized method to click an element using xpath.
|
863
|
+
"""
|
864
|
+
page = await self.get_current_page()
|
865
|
+
|
866
|
+
try:
|
867
|
+
element = await self.get_locate_element(element_node)
|
868
|
+
|
869
|
+
if element is None:
|
870
|
+
raise Exception(f'Element: {repr(element_node)} not found')
|
871
|
+
|
872
|
+
# await element.scroll_into_view_if_needed()
|
873
|
+
|
874
|
+
try:
|
875
|
+
await element.click(timeout=1500)
|
876
|
+
await page.wait_for_load_state()
|
877
|
+
except Exception:
|
878
|
+
try:
|
879
|
+
await page.evaluate('(el) => el.click()', element)
|
880
|
+
await page.wait_for_load_state()
|
881
|
+
except Exception as e:
|
882
|
+
raise Exception(f'Failed to click element: {str(e)}')
|
883
|
+
|
884
|
+
except Exception as e:
|
885
|
+
raise Exception(f'Failed to click element: {
|
886
|
+
repr(element_node)}. Error: {str(e)}')
|
887
|
+
|
888
|
+
async def get_tabs_info(self) -> list[TabInfo]:
|
889
|
+
"""Get information about all tabs"""
|
890
|
+
session = await self.get_session()
|
891
|
+
|
892
|
+
tabs_info = []
|
893
|
+
for page_id, page in enumerate(session.context.pages):
|
894
|
+
tab_info = TabInfo(page_id=page_id, url=page.url, title=await page.title())
|
895
|
+
tabs_info.append(tab_info)
|
896
|
+
|
897
|
+
return tabs_info
|
898
|
+
|
899
|
+
async def switch_to_tab(self, page_id: int) -> None:
|
900
|
+
"""Switch to a specific tab by its page_id
|
901
|
+
|
902
|
+
@You can also use negative indices to switch to tabs from the end (Pure pythonic way)
|
903
|
+
"""
|
904
|
+
session = await self.get_session()
|
905
|
+
pages = session.context.pages
|
906
|
+
|
907
|
+
if page_id >= len(pages):
|
908
|
+
raise BrowserError(f'No tab found with page_id: {page_id}')
|
909
|
+
|
910
|
+
page = pages[page_id]
|
911
|
+
session.current_page = page
|
912
|
+
|
913
|
+
await page.bring_to_front()
|
914
|
+
await page.wait_for_load_state()
|
915
|
+
|
916
|
+
async def create_new_tab(self, url: str | None = None) -> None:
|
917
|
+
"""Create a new tab and optionally navigate to a URL"""
|
918
|
+
session = await self.get_session()
|
919
|
+
new_page = await session.context.new_page()
|
920
|
+
session.current_page = new_page
|
921
|
+
|
922
|
+
await new_page.wait_for_load_state()
|
923
|
+
|
924
|
+
page = await self.get_current_page()
|
925
|
+
|
926
|
+
if url:
|
927
|
+
await page.goto(url)
|
928
|
+
await self._wait_for_page_and_frames_load(timeout_overwrite=1)
|
929
|
+
|
930
|
+
# endregion
|
931
|
+
|
932
|
+
# region - Helper methods for easier access to the DOM
|
933
|
+
async def get_selector_map(self) -> SelectorMap:
|
934
|
+
session = await self.get_session()
|
935
|
+
return session.cached_state.selector_map
|
936
|
+
|
937
|
+
async def get_element_by_index(self, index: int) -> ElementHandle | None:
|
938
|
+
selector_map = await self.get_selector_map()
|
939
|
+
return await self.get_locate_element(selector_map[index])
|
940
|
+
|
941
|
+
async def get_dom_element_by_index(self, index: int) -> DOMElementNode | None:
|
942
|
+
selector_map = await self.get_selector_map()
|
943
|
+
return selector_map[index]
|
944
|
+
|
945
|
+
async def save_cookies(self):
|
946
|
+
"""Save current cookies to file"""
|
947
|
+
if self.session and self.session.context and self.config.cookies_file:
|
948
|
+
try:
|
949
|
+
cookies = await self.session.context.cookies()
|
950
|
+
logger.info(f'Saving {len(cookies)} cookies to {
|
951
|
+
self.config.cookies_file}')
|
952
|
+
|
953
|
+
# Check if the path is a directory and create it if necessary
|
954
|
+
dirname = os.path.dirname(self.config.cookies_file)
|
955
|
+
if dirname:
|
956
|
+
os.makedirs(dirname, exist_ok=True)
|
957
|
+
|
958
|
+
with open(self.config.cookies_file, 'w') as f:
|
959
|
+
json.dump(cookies, f)
|
960
|
+
except Exception as e:
|
961
|
+
logger.warning(f'Failed to save cookies: {str(e)}')
|
962
|
+
|
963
|
+
async def is_file_uploader(
|
964
|
+
self, element_node: DOMElementNode, max_depth: int = 3, current_depth: int = 0
|
965
|
+
) -> bool:
|
966
|
+
"""Check if element or its children are file uploaders"""
|
967
|
+
if current_depth > max_depth:
|
968
|
+
return False
|
969
|
+
|
970
|
+
# Check current element
|
971
|
+
is_uploader = False
|
972
|
+
|
973
|
+
if not isinstance(element_node, DOMElementNode):
|
974
|
+
return False
|
975
|
+
|
976
|
+
# Check for file input attributes
|
977
|
+
if element_node.tag_name == 'input':
|
978
|
+
is_uploader = (
|
979
|
+
element_node.attributes.get('type') == 'file'
|
980
|
+
or element_node.attributes.get('accept') is not None
|
981
|
+
)
|
982
|
+
|
983
|
+
if is_uploader:
|
984
|
+
return True
|
985
|
+
|
986
|
+
# Recursively check children
|
987
|
+
if element_node.children and current_depth < max_depth:
|
988
|
+
for child in element_node.children:
|
989
|
+
if isinstance(child, DOMElementNode):
|
990
|
+
if await self.is_file_uploader(child, max_depth, current_depth + 1):
|
991
|
+
return True
|
992
|
+
|
993
|
+
return False
|