aiws / search_engine.py
fikird
Enhance content processing with better summarization and topic extraction
8e83c5f
raw
history blame
13.6 kB
from typing import Dict, List, Any
import requests
from bs4 import BeautifulSoup
from transformers import pipeline
from langchain_community.embeddings import HuggingFaceEmbeddings
import time
import json
import os
from urllib.parse import urlparse, quote_plus
import logging
import random
logger = logging.getLogger(__name__)
class SearchResult:
def __init__(self, title: str, link: str, snippet: str):
self.title = title
self.link = link
self.snippet = snippet
class ModelManager:
"""Manages different AI models for specific tasks"""
def __init__(self):
self.device = "cpu"
self.models = {}
self.load_models()
def load_models(self):
# Use smaller models for CPU deployment
self.models['summarizer'] = pipeline(
"summarization",
model="facebook/bart-base",
device=self.device
)
self.models['embeddings'] = HuggingFaceEmbeddings(
model_name="sentence-transformers/all-MiniLM-L6-v2",
model_kwargs={"device": self.device}
)
class ContentProcessor:
"""Processes and analyzes different types of content"""
def __init__(self):
self.model_manager = ModelManager()
def clean_text(self, text: str) -> str:
"""Clean and normalize text content"""
# Remove extra whitespace and normalize
lines = [line.strip() for line in text.splitlines()]
text = ' '.join(line for line in lines if line)
# Remove redundant spaces
text = ' '.join(text.split())
# Remove common navigation elements
nav_patterns = [
"skip to content",
"search",
"menu",
"navigation",
"subscribe",
"sign in",
"log in"
]
for pattern in nav_patterns:
text = text.replace(pattern, "")
return text
def extract_key_points(self, text: str, max_points: int = 5) -> List[str]:
"""Extract key points from text using the summarizer"""
try:
# Split text into chunks of ~1000 characters
chunks = [text[i:i + 1000] for i in range(0, len(text), 1000)]
all_points = []
for chunk in chunks[:3]: # Process first 3 chunks only
summary = self.model_manager.models['summarizer'](
chunk,
max_length=100,
min_length=30,
do_sample=False
)[0]['summary_text']
# Split into sentences and add as points
sentences = [s.strip() for s in summary.split('.') if s.strip()]
all_points.extend(sentences)
# Return unique points, limited to max_points
unique_points = list(dict.fromkeys(all_points))
return unique_points[:max_points]
except Exception as e:
logger.error(f"Error extracting key points: {str(e)}")
return []
def process_content(self, content: str) -> Dict:
"""Process content and generate insights"""
try:
# Clean the text
cleaned_text = self.clean_text(content)
# Extract key points
key_points = self.extract_key_points(cleaned_text)
# Generate a concise summary
summary = self.model_manager.models['summarizer'](
cleaned_text[:1024],
max_length=150,
min_length=50,
do_sample=False
)[0]['summary_text']
# Extract potential topics/keywords
topics = []
common_topics = [
"quantum computing", "quantum processors", "quantum bits",
"quantum algorithms", "quantum supremacy", "quantum advantage",
"error correction", "quantum hardware", "quantum software",
"quantum research", "quantum applications"
]
for topic in common_topics:
if topic.lower() in cleaned_text.lower():
topics.append(topic)
return {
'summary': summary,
'key_points': key_points,
'topics': topics[:5], # Limit to top 5 topics
'content': cleaned_text
}
except Exception as e:
return {
'summary': f"Error processing content: {str(e)}",
'key_points': [],
'topics': [],
'content': content
}
class WebSearchEngine:
"""Main search engine class"""
def __init__(self):
self.processor = ContentProcessor()
self.session = requests.Session()
self.request_delay = 2.0
self.last_request_time = 0
self.max_retries = 3
self.headers = {
'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',
'Accept': 'text/html,application/xhtml+xml,application/xml;q=0.9,image/webp,*/*;q=0.8',
'Accept-Language': 'en-US,en;q=0.5',
'DNT': '1',
'Connection': 'keep-alive',
'Upgrade-Insecure-Requests': '1'
}
def safe_get(self, url: str, max_retries: int = 3) -> requests.Response:
"""Make a GET request with retries and error handling"""
for i in range(max_retries):
try:
# Add delay between requests
current_time = time.time()
time_since_last = current_time - self.last_request_time
if time_since_last < self.request_delay:
time.sleep(self.request_delay - time_since_last + random.uniform(0.5, 1.5))
response = self.session.get(url, headers=self.headers, timeout=10)
self.last_request_time = time.time()
if response.status_code == 200:
return response
elif response.status_code == 429: # Rate limit
wait_time = (i + 1) * 5
time.sleep(wait_time)
continue
else:
response.raise_for_status()
except Exception as e:
if i == max_retries - 1:
raise
time.sleep((i + 1) * 2)
raise Exception(f"Failed to fetch URL after {max_retries} attempts")
def is_valid_url(self, url: str) -> bool:
"""Check if URL is valid for crawling"""
try:
parsed = urlparse(url)
return bool(parsed.netloc and parsed.scheme)
except:
return False
def get_metadata(self, soup: BeautifulSoup) -> Dict:
"""Extract metadata from page"""
title = soup.title.string if soup.title else "No title"
description = ""
if soup.find("meta", attrs={"name": "description"}):
description = soup.find("meta", attrs={"name": "description"}).get("content", "")
return {
'title': title,
'description': description
}
def process_url(self, url: str) -> Dict:
"""Process a single URL"""
if not self.is_valid_url(url):
return {'error': f"Invalid URL: {url}"}
try:
response = self.safe_get(url)
soup = BeautifulSoup(response.text, 'lxml')
# Extract text content
for script in soup(["script", "style"]):
script.decompose()
text = soup.get_text()
lines = (line.strip() for line in text.splitlines())
chunks = (phrase.strip() for line in lines for phrase in line.split(" "))
content = ' '.join(chunk for chunk in chunks if chunk)
# Get metadata
metadata = self.get_metadata(soup)
# Process content
processed = self.processor.process_content(content)
return {
'url': url,
'title': metadata['title'],
'description': metadata['description'],
'summary': processed['summary'],
'key_points': processed['key_points'],
'topics': processed['topics'],
'content': processed['content']
}
except Exception as e:
return {'error': f"Error processing {url}: {str(e)}"}
def search_duckduckgo(self, query: str, max_results: int = 5) -> List[Dict]:
"""Search DuckDuckGo and parse HTML results"""
search_results = []
try:
# Encode query for URL
encoded_query = quote_plus(query)
# DuckDuckGo HTML search URL
search_url = f'https://html.duckduckgo.com/html/?q={encoded_query}'
# Get search results page
response = self.safe_get(search_url)
soup = BeautifulSoup(response.text, 'lxml')
# Find all result elements
results = soup.find_all('div', {'class': 'result'})
for result in results[:max_results]:
try:
# Extract link
link_elem = result.find('a', {'class': 'result__a'})
if not link_elem:
continue
link = link_elem.get('href', '')
if not link or not self.is_valid_url(link):
continue
# Extract title
title = link_elem.get_text(strip=True)
# Extract snippet
snippet_elem = result.find('a', {'class': 'result__snippet'})
snippet = snippet_elem.get_text(strip=True) if snippet_elem else ""
search_results.append({
'link': link,
'title': title,
'snippet': snippet
})
# Add delay between processing results
time.sleep(random.uniform(0.2, 0.5))
except Exception as e:
logger.warning(f"Error processing search result: {str(e)}")
continue
return search_results
except Exception as e:
logger.error(f"Error during DuckDuckGo search: {str(e)}")
return []
def search(self, query: str, max_results: int = 5) -> Dict:
"""Perform search and process results"""
try:
# Search using DuckDuckGo HTML
search_results = self.search_duckduckgo(query, max_results)
if not search_results:
return {'error': 'No results found'}
results = []
all_key_points = []
all_topics = set()
for result in search_results:
if 'link' in result:
processed = self.process_url(result['link'])
if 'error' not in processed:
results.append(processed)
# Collect key points and topics
if 'key_points' in processed:
all_key_points.extend(processed['key_points'])
if 'topics' in processed:
all_topics.update(processed.get('topics', []))
time.sleep(random.uniform(0.5, 1.0))
if not results:
return {'error': 'Failed to process any search results'}
# Combine all summaries
all_summaries = " ".join([r['summary'] for r in results if 'summary' in r])
# Generate a meta-summary of all content
meta_summary = self.processor.model_manager.models['summarizer'](
all_summaries[:1024],
max_length=200,
min_length=100,
do_sample=False
)[0]['summary_text']
# Get unique key points
unique_key_points = list(dict.fromkeys(all_key_points))
return {
'results': results,
'insights': {
'summary': meta_summary,
'key_points': unique_key_points[:7], # Top 7 key points
'topics': list(all_topics)[:5] # Top 5 topics
},
'follow_up_questions': [
f"What are the recent breakthroughs in {', '.join(list(all_topics)[:2])}?",
f"How do these developments impact the future of quantum computing?",
f"What are the practical applications of these quantum computing advances?"
]
}
except Exception as e:
return {'error': f"Search failed: {str(e)}"}
# Main search function
def search(query: str, max_results: int = 5) -> Dict:
"""Main search function"""
engine = WebSearchEngine()
return engine.search(query, max_results)