Spaces:
Running
Running
Commit
·
b3b4203
1
Parent(s):
958ee21
update
Browse files- app/services/websearch.py +258 -51
app/services/websearch.py
CHANGED
@@ -5,11 +5,12 @@ from bs4 import BeautifulSoup
|
|
5 |
import urllib.parse
|
6 |
import time
|
7 |
import random
|
|
|
8 |
|
9 |
warnings.simplefilter('ignore', requests.packages.urllib3.exceptions.InsecureRequestWarning)
|
10 |
|
11 |
class WebSearch:
|
12 |
-
def __init__(self, num_results=4, max_chars_per_page=6000
|
13 |
self.num_results = num_results
|
14 |
self.max_chars_per_page = max_chars_per_page
|
15 |
self.reference = []
|
@@ -23,13 +24,138 @@ class WebSearch:
|
|
23 |
'DNT': '1',
|
24 |
'Connection': 'keep-alive',
|
25 |
}
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
26 |
|
27 |
def extract_text_from_webpage(self, html_content):
|
28 |
soup = BeautifulSoup(html_content, "html.parser")
|
29 |
-
|
|
|
|
|
|
|
30 |
tag.extract()
|
31 |
-
|
32 |
-
|
|
|
|
|
|
|
|
|
33 |
|
34 |
def search(self, query):
|
35 |
results = []
|
@@ -42,32 +168,86 @@ class WebSearch:
|
|
42 |
|
43 |
response = session.get(url, timeout=10)
|
44 |
soup = BeautifulSoup(response.text, 'html.parser')
|
45 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
46 |
|
47 |
from concurrent.futures import ThreadPoolExecutor, as_completed
|
48 |
|
49 |
def fetch_page(link):
|
50 |
try:
|
51 |
-
|
52 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
53 |
page_soup = BeautifulSoup(page_response.text, 'lxml')
|
54 |
-
|
55 |
-
|
56 |
-
|
|
|
|
|
|
|
57 |
text = ' '.join(page_soup.stripped_strings)
|
|
|
|
|
|
|
|
|
58 |
return {
|
59 |
'link': link,
|
|
|
60 |
'text': text[:self.max_chars_per_page]
|
61 |
}
|
62 |
except Exception as e:
|
|
|
63 |
return None
|
64 |
|
65 |
-
|
66 |
-
|
67 |
-
if result.find('a', class_='result__a')]
|
68 |
-
|
69 |
-
with ThreadPoolExecutor(max_workers=min(len(links), 4)) as executor:
|
70 |
-
future_to_url = {executor.submit(fetch_page, link): link for link in links}
|
71 |
|
72 |
for future in as_completed(future_to_url):
|
73 |
result = future.result()
|
@@ -77,10 +257,12 @@ class WebSearch:
|
|
77 |
return results
|
78 |
|
79 |
except Exception as e:
|
|
|
80 |
return []
|
81 |
|
82 |
def search_images(self, query):
|
83 |
images = []
|
|
|
84 |
|
85 |
headers = {
|
86 |
'User-Agent': 'Mozilla/5.0 (Windows NT 10.0; Win64; x64) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/91.0.4472.124 Safari/537.36',
|
@@ -92,45 +274,70 @@ class WebSearch:
|
|
92 |
'Upgrade-Insecure-Requests': '1'
|
93 |
}
|
94 |
|
95 |
-
|
96 |
-
|
97 |
-
|
98 |
-
|
99 |
-
|
100 |
-
|
101 |
-
|
102 |
-
|
103 |
-
|
104 |
-
|
105 |
-
|
106 |
-
|
107 |
-
|
108 |
-
|
109 |
-
|
110 |
-
|
111 |
-
|
112 |
-
|
113 |
-
|
114 |
-
|
115 |
-
|
116 |
-
|
117 |
-
|
118 |
-
|
119 |
-
|
120 |
-
|
121 |
-
|
122 |
-
|
123 |
-
|
|
|
|
|
|
|
|
|
124 |
|
|
|
125 |
seen = set()
|
126 |
-
|
127 |
-
|
128 |
-
|
|
|
|
|
|
|
129 |
|
130 |
def is_image_url(self, url):
|
|
|
131 |
image_extensions = ('.jpg', '.jpeg', '.png', '.gif', '.bmp', '.webp')
|
132 |
return any(url.lower().endswith(ext) for ext in image_extensions)
|
133 |
|
134 |
-
def
|
135 |
-
|
136 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
5 |
import urllib.parse
|
6 |
import time
|
7 |
import random
|
8 |
+
from urllib.parse import urlparse, parse_qs
|
9 |
|
10 |
warnings.simplefilter('ignore', requests.packages.urllib3.exceptions.InsecureRequestWarning)
|
11 |
|
12 |
class WebSearch:
|
13 |
+
def __init__(self, num_results=4, max_chars_per_page=6000, max_images=10):
|
14 |
self.num_results = num_results
|
15 |
self.max_chars_per_page = max_chars_per_page
|
16 |
self.reference = []
|
|
|
24 |
'DNT': '1',
|
25 |
'Connection': 'keep-alive',
|
26 |
}
|
27 |
+
# Common domains for direct content
|
28 |
+
self.content_domains = [
|
29 |
+
"wikipedia.org", "webmd.com", "mayoclinic.org", "healthline.com", "nih.gov",
|
30 |
+
"clevelandclinic.org", "nhs.uk", "cdc.gov", "medlineplus.gov", "hopkinsmedicine.org"
|
31 |
+
]
|
32 |
+
# Ad and tracking domains to filter out
|
33 |
+
self.blocked_domains = [
|
34 |
+
"ad.doubleclick.net", "googleadservices.com", "bing.com/aclick", "duckduckgo.com/y.js",
|
35 |
+
"amazon.com/s", "ads.google.com", "analytics", "tracker", "pixel", "adservice"
|
36 |
+
]
|
37 |
+
|
38 |
+
def is_valid_url(self, url):
|
39 |
+
"""Check if URL is valid and not an ad/tracking URL"""
|
40 |
+
if not url or len(url) < 10:
|
41 |
+
return False
|
42 |
+
|
43 |
+
try:
|
44 |
+
parsed = urlparse(url)
|
45 |
+
|
46 |
+
# Check if URL has a valid scheme and netloc
|
47 |
+
if not all([parsed.scheme, parsed.netloc]):
|
48 |
+
return False
|
49 |
+
|
50 |
+
# Filter out ad/tracking URLs
|
51 |
+
domain = parsed.netloc.lower()
|
52 |
+
path = parsed.path.lower()
|
53 |
+
query = parsed.query.lower()
|
54 |
+
|
55 |
+
# Block URLs containing ad-related indicators
|
56 |
+
for blocked in self.blocked_domains:
|
57 |
+
if blocked in domain or blocked in path:
|
58 |
+
return False
|
59 |
+
|
60 |
+
# Block URLs with ad-related query parameters
|
61 |
+
if any(param in query for param in ["ad", "click", "track", "clkid", "msclkid"]):
|
62 |
+
return False
|
63 |
+
|
64 |
+
# Extra check for redirect URLs
|
65 |
+
if "redirect" in path or "goto" in path or "go.php" in path:
|
66 |
+
return False
|
67 |
+
|
68 |
+
# Reject extremely long URLs (often tracking)
|
69 |
+
if len(url) > 500:
|
70 |
+
return False
|
71 |
+
|
72 |
+
return True
|
73 |
+
|
74 |
+
except Exception:
|
75 |
+
return False
|
76 |
+
|
77 |
+
def clean_url(self, url):
|
78 |
+
"""Clean the URL by removing tracking parameters"""
|
79 |
+
try:
|
80 |
+
parsed = urlparse(url)
|
81 |
+
|
82 |
+
# List of known tracking parameters to remove
|
83 |
+
tracking_params = [
|
84 |
+
'utm_', 'ref_', 'ref=', 'refid', 'fbclid', 'gclid', 'msclkid', 'dclid',
|
85 |
+
'zanpid', 'icid', 'igshid', 'mc_eid', '_hsenc', 'mkt_tok', 'yclid'
|
86 |
+
]
|
87 |
+
|
88 |
+
# Parse query parameters
|
89 |
+
query_params = parse_qs(parsed.query)
|
90 |
+
|
91 |
+
# Remove tracking parameters
|
92 |
+
filtered_params = {
|
93 |
+
k: v for k, v in query_params.items()
|
94 |
+
if not any(tracker in k.lower() for tracker in tracking_params)
|
95 |
+
}
|
96 |
+
|
97 |
+
# Rebuild query string
|
98 |
+
clean_query = urllib.parse.urlencode(filtered_params, doseq=True) if filtered_params else ""
|
99 |
+
|
100 |
+
# Reconstruct URL
|
101 |
+
clean_url = urllib.parse.urlunparse((
|
102 |
+
parsed.scheme,
|
103 |
+
parsed.netloc,
|
104 |
+
parsed.path,
|
105 |
+
parsed.params,
|
106 |
+
clean_query,
|
107 |
+
"" # Remove fragment
|
108 |
+
))
|
109 |
+
|
110 |
+
return clean_url
|
111 |
+
|
112 |
+
except Exception:
|
113 |
+
# If any error occurs, return the original URL
|
114 |
+
return url
|
115 |
+
|
116 |
+
def extract_real_url_from_redirect(self, url):
|
117 |
+
"""Extract the actual URL from a redirect URL"""
|
118 |
+
try:
|
119 |
+
parsed = urlparse(url)
|
120 |
+
|
121 |
+
# Handle DuckDuckGo redirects
|
122 |
+
if "duckduckgo.com" in parsed.netloc and "u3=" in parsed.query:
|
123 |
+
params = parse_qs(parsed.query)
|
124 |
+
if "u3" in params and params["u3"]:
|
125 |
+
redirect_url = params["u3"][0]
|
126 |
+
# Handle nested redirects (like Bing redirects inside DuckDuckGo)
|
127 |
+
if "bing.com/aclick" in redirect_url:
|
128 |
+
bing_parsed = urlparse(redirect_url)
|
129 |
+
bing_params = parse_qs(bing_parsed.query)
|
130 |
+
if "u" in bing_params and bing_params["u"]:
|
131 |
+
decoded_url = urllib.parse.unquote(bing_params["u"][0])
|
132 |
+
return self.clean_url(decoded_url)
|
133 |
+
return self.clean_url(redirect_url)
|
134 |
+
|
135 |
+
# Handle Bing redirects
|
136 |
+
if "bing.com/aclick" in url:
|
137 |
+
params = parse_qs(parsed.query)
|
138 |
+
if "u" in params and params["u"]:
|
139 |
+
return self.clean_url(urllib.parse.unquote(params["u"][0]))
|
140 |
+
|
141 |
+
return url
|
142 |
+
|
143 |
+
except Exception:
|
144 |
+
return url
|
145 |
|
146 |
def extract_text_from_webpage(self, html_content):
|
147 |
soup = BeautifulSoup(html_content, "html.parser")
|
148 |
+
|
149 |
+
# Remove non-content elements
|
150 |
+
for tag in soup(["script", "style", "header", "footer", "nav", "form", "svg",
|
151 |
+
"aside", "iframe", "noscript", "img", "figure", "button"]):
|
152 |
tag.extract()
|
153 |
+
|
154 |
+
# Extract text and normalize spacing
|
155 |
+
text = ' '.join(soup.stripped_strings)
|
156 |
+
text = re.sub(r'\s+', ' ', text).strip()
|
157 |
+
|
158 |
+
return text
|
159 |
|
160 |
def search(self, query):
|
161 |
results = []
|
|
|
168 |
|
169 |
response = session.get(url, timeout=10)
|
170 |
soup = BeautifulSoup(response.text, 'html.parser')
|
171 |
+
|
172 |
+
# Getting more results than needed to account for filtering
|
173 |
+
search_results = soup.find_all('div', class_='result')[:self.num_results * 2]
|
174 |
+
links = []
|
175 |
+
|
176 |
+
# Extract and process links
|
177 |
+
for result in search_results:
|
178 |
+
link_tag = result.find('a', class_='result__a')
|
179 |
+
if not link_tag or not link_tag.get('href'):
|
180 |
+
continue
|
181 |
+
|
182 |
+
original_link = link_tag['href']
|
183 |
+
|
184 |
+
# Process link to get the actual URL
|
185 |
+
clean_link = self.extract_real_url_from_redirect(original_link)
|
186 |
+
|
187 |
+
# Validate the URL
|
188 |
+
if self.is_valid_url(clean_link):
|
189 |
+
links.append(clean_link)
|
190 |
+
|
191 |
+
# Prioritize content domains
|
192 |
+
prioritized_links = []
|
193 |
+
other_links = []
|
194 |
+
|
195 |
+
for link in links:
|
196 |
+
if any(domain in link for domain in self.content_domains):
|
197 |
+
prioritized_links.append(link)
|
198 |
+
else:
|
199 |
+
other_links.append(link)
|
200 |
+
|
201 |
+
# Combine prioritized links first, then others
|
202 |
+
final_links = prioritized_links + other_links
|
203 |
+
|
204 |
+
# Limit to unique links up to num_results
|
205 |
+
unique_links = []
|
206 |
+
seen_domains = set()
|
207 |
+
|
208 |
+
for link in final_links:
|
209 |
+
domain = urlparse(link).netloc
|
210 |
+
if domain not in seen_domains and len(unique_links) < self.num_results:
|
211 |
+
unique_links.append(link)
|
212 |
+
seen_domains.add(domain)
|
213 |
|
214 |
from concurrent.futures import ThreadPoolExecutor, as_completed
|
215 |
|
216 |
def fetch_page(link):
|
217 |
try:
|
218 |
+
# Random delay to avoid being blocked
|
219 |
+
time.sleep(random.uniform(0.5, 1.5))
|
220 |
+
|
221 |
+
# Set a longer timeout for reliable fetching
|
222 |
+
page_response = session.get(link, timeout=10, verify=False)
|
223 |
+
|
224 |
+
# Only process HTML content
|
225 |
+
if 'text/html' not in page_response.headers.get('Content-Type', ''):
|
226 |
+
return None
|
227 |
+
|
228 |
page_soup = BeautifulSoup(page_response.text, 'lxml')
|
229 |
+
|
230 |
+
# Remove non-content elements
|
231 |
+
[tag.decompose() for tag in page_soup(['script', 'style', 'header', 'footer',
|
232 |
+
'nav', 'form', 'iframe', 'noscript'])]
|
233 |
+
|
234 |
+
# Extract text with better formatting
|
235 |
text = ' '.join(page_soup.stripped_strings)
|
236 |
+
text = re.sub(r'\s+', ' ', text).strip()
|
237 |
+
|
238 |
+
title = page_soup.title.string if page_soup.title else "Untitled Page"
|
239 |
+
|
240 |
return {
|
241 |
'link': link,
|
242 |
+
'title': title,
|
243 |
'text': text[:self.max_chars_per_page]
|
244 |
}
|
245 |
except Exception as e:
|
246 |
+
print(f"Error fetching {link}: {str(e)}")
|
247 |
return None
|
248 |
|
249 |
+
with ThreadPoolExecutor(max_workers=min(len(unique_links), 4)) as executor:
|
250 |
+
future_to_url = {executor.submit(fetch_page, link): link for link in unique_links}
|
|
|
|
|
|
|
|
|
251 |
|
252 |
for future in as_completed(future_to_url):
|
253 |
result = future.result()
|
|
|
257 |
return results
|
258 |
|
259 |
except Exception as e:
|
260 |
+
print(f"Search error: {str(e)}")
|
261 |
return []
|
262 |
|
263 |
def search_images(self, query):
|
264 |
images = []
|
265 |
+
encoded_query = urllib.parse.quote(query)
|
266 |
|
267 |
headers = {
|
268 |
'User-Agent': 'Mozilla/5.0 (Windows NT 10.0; Win64; x64) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/91.0.4472.124 Safari/537.36',
|
|
|
274 |
'Upgrade-Insecure-Requests': '1'
|
275 |
}
|
276 |
|
277 |
+
# Try multiple sources for better results
|
278 |
+
image_sources = [
|
279 |
+
f"https://www.google.com/search?q={encoded_query}&tbm=isch&hl=en",
|
280 |
+
f"https://www.bing.com/images/search?q={encoded_query}&form=HDRSC2&first=1",
|
281 |
+
f"https://duckduckgo.com/?q={encoded_query}&iar=images&iax=images&ia=images"
|
282 |
+
]
|
283 |
+
|
284 |
+
for source_url in image_sources:
|
285 |
+
try:
|
286 |
+
time.sleep(random.uniform(0.5, 1.0)) # Polite delay
|
287 |
+
response = requests.get(source_url, headers=headers, verify=False, timeout=10)
|
288 |
+
soup = BeautifulSoup(response.text, 'html.parser')
|
289 |
+
|
290 |
+
# Extract image URLs from img tags
|
291 |
+
for img in soup.find_all('img'):
|
292 |
+
src = img.get('src', '')
|
293 |
+
if src and src.startswith('http') and self.is_image_url(src):
|
294 |
+
cleaned_url = self.clean_url(src)
|
295 |
+
if self.is_valid_image(cleaned_url):
|
296 |
+
images.append(cleaned_url)
|
297 |
+
|
298 |
+
# Extract image URLs from scripts (useful for Google Images)
|
299 |
+
for script in soup.find_all('script'):
|
300 |
+
if script.string:
|
301 |
+
urls = re.findall(r'https?://[^\s<>"\']+?(?:\.(?:jpg|jpeg|png|gif|bmp|webp))', script.string)
|
302 |
+
for url in urls:
|
303 |
+
cleaned_url = self.clean_url(url)
|
304 |
+
if self.is_valid_image(cleaned_url):
|
305 |
+
images.append(cleaned_url)
|
306 |
+
|
307 |
+
except Exception as e:
|
308 |
+
print(f"Error searching images at {source_url}: {str(e)}")
|
309 |
+
continue
|
310 |
|
311 |
+
# Remove duplicates while preserving order
|
312 |
seen = set()
|
313 |
+
unique_images = [x for x in images if not (x in seen or seen.add(x))]
|
314 |
+
|
315 |
+
# Filter out small images and suspicious URLs
|
316 |
+
filtered_images = [img for img in unique_images if self.is_valid_image(img)]
|
317 |
+
|
318 |
+
return filtered_images[:self.max_images]
|
319 |
|
320 |
def is_image_url(self, url):
|
321 |
+
"""Check if URL points to an image file"""
|
322 |
image_extensions = ('.jpg', '.jpeg', '.png', '.gif', '.bmp', '.webp')
|
323 |
return any(url.lower().endswith(ext) for ext in image_extensions)
|
324 |
|
325 |
+
def is_valid_image(self, url):
|
326 |
+
"""Additional validation for image URLs"""
|
327 |
+
try:
|
328 |
+
# Reject tiny images (often icons) and tracking pixels
|
329 |
+
if re.search(r'(?:icon|pixel|tracker|thumb|logo|button)\d*\.(?:jpg|png|gif)', url.lower()):
|
330 |
+
return False
|
331 |
+
|
332 |
+
# Avoid suspicious domains for images
|
333 |
+
parsed = urlparse(url)
|
334 |
+
if any(bad in parsed.netloc.lower() for bad in ["tracker", "pixel", "counter", "ad."]):
|
335 |
+
return False
|
336 |
+
|
337 |
+
# Avoid very short URLs (likely not valid images)
|
338 |
+
if len(url) < 30:
|
339 |
+
return False
|
340 |
+
|
341 |
+
return True
|
342 |
+
except:
|
343 |
+
return False
|