From the uDemy course on LLM engineering.
https://www.udemy.com/course/llm-engineering-master-ai-and-large-language-models
You can not select more than 25 topics
Topics must start with a letter or number, can include dashes ('-') and can be up to 35 characters long.
300 lines
12 KiB
300 lines
12 KiB
import asyncio |
|
from playwright.async_api import async_playwright |
|
from openai import OpenAI |
|
import logging |
|
import random |
|
import time |
|
import os |
|
from prometheus_client import start_http_server, Counter, Histogram |
|
from diskcache import Cache |
|
from dotenv import load_dotenv |
|
|
|
load_dotenv() |
|
|
|
# Setting up Prometheus metrics |
|
SCRAPE_ATTEMPTS = Counter('scrape_attempts', 'Total scraping attempts') |
|
SCRAPE_DURATION = Histogram( |
|
'scrape_duration', 'Scraping duration distribution') |
|
|
|
# Setting up cache |
|
cache = Cache('./scraper_cache') |
|
|
|
|
|
class ScrapingError(Exception): |
|
pass |
|
|
|
|
|
class ContentAnalysisError(Exception): |
|
pass |
|
|
|
|
|
class EnhancedOpenAIScraper: |
|
API_KEY = os.getenv("OPENAI_API_KEY") |
|
BROWSER_EXECUTABLE = os.getenv( |
|
"BROWSER_PATH", "/usr/bin/chromium-browser") |
|
MAX_CONTENT_LENGTH = int(os.getenv("MAX_CONTENT_LENGTH", 30000)) |
|
|
|
def __init__(self, headless=True): |
|
self.user_agents = [ |
|
"Mozilla/5.0 (Windows NT 10.0; Win64; x64) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/91.0.4472.124 Safari/537.36", |
|
"Mozilla/5.0 (Macintosh; Intel Mac OS X 10_15_7) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/91.0.4472.124 Safari/537.36" |
|
] |
|
self.timeout = 45000 # 45 seconds |
|
self.retry_count = int(os.getenv("RETRY_COUNT", 2)) |
|
self.headless = headless |
|
self.mouse_velocity_range = (100, 500) # px/ms |
|
self.interaction_delays = { |
|
'scroll': (int(os.getenv("SCROLL_DELAY_MIN", 500)), int(os.getenv("SCROLL_DELAY_MAX", 2000))), |
|
'click': (int(os.getenv("CLICK_DELAY_MIN", 100)), int(os.getenv("CLICK_DELAY_MAX", 300))), |
|
'movement': (int(os.getenv("MOVEMENT_DELAY_MIN", 50)), int(os.getenv("MOVEMENT_DELAY_MAX", 200))) |
|
} |
|
self.proxy_servers = [server.strip() for server in os.getenv( |
|
"PROXY_SERVERS", "").split(',') if server.strip()] |
|
|
|
async def human_interaction(self, page): |
|
"""Advanced simulation of user behavior""" |
|
# Random mouse movement path |
|
for _ in range(random.randint(2, 5)): |
|
x = random.randint(0, 1366) |
|
y = random.randint(0, 768) |
|
await page.mouse.move(x, y, steps=random.randint(5, 20)) |
|
await page.wait_for_timeout(random.randint(*self.interaction_delays['movement'])) |
|
|
|
# Simulating typing |
|
if random.random() < 0.3: |
|
await page.keyboard.press('Tab') |
|
await page.keyboard.type(' ', delay=random.randint(50, 200)) |
|
|
|
# More realistic scrolling |
|
scroll_distance = random.choice([300, 600, 900]) |
|
await page.mouse.wheel(0, scroll_distance) |
|
await page.wait_for_timeout(random.randint(*self.interaction_delays['scroll'])) |
|
|
|
async def load_page(self, page, url): |
|
"""Smarter page loading with dynamic waiting""" |
|
start_time = time.time() |
|
try: |
|
await page.goto(url, wait_until="domcontentloaded", timeout=self.timeout) |
|
|
|
# Smarter content extraction selectors |
|
selectors = [ |
|
'main article', |
|
'#main-content', |
|
'section:first-of-type', |
|
'div[class*="content"]', |
|
'body' # Fallback |
|
] |
|
|
|
for selector in selectors: |
|
try: |
|
element = await page.query_selector(selector) |
|
if element: |
|
return True |
|
except Exception: |
|
continue |
|
|
|
# Fallback if no selector is found within a certain time |
|
if time.time() - start_time < 30: # If we haven't used the full timeout |
|
await page.wait_for_timeout(30000 - int(time.time() - start_time)) |
|
|
|
return True # Page likely loaded |
|
except Exception as e: |
|
logging.error(f"Error loading page {url}: {e}") |
|
return False |
|
|
|
@SCRAPE_DURATION.time() |
|
async def scrape_with_retry(self): |
|
"""Main function with retry mechanism and browser reuse""" |
|
SCRAPE_ATTEMPTS.inc() |
|
last_error = None |
|
browser = None |
|
context = None |
|
page = None |
|
|
|
try: |
|
async with async_playwright() as p: |
|
launch_args = { |
|
"headless": self.headless, |
|
"args": [ |
|
"--disable-blink-features=AutomationControlled", |
|
"--single-process", |
|
"--no-sandbox", |
|
f"--user-agent={random.choice(self.user_agents)}" |
|
], |
|
"executable_path": self.BROWSER_EXECUTABLE |
|
} |
|
if self.proxy_servers: |
|
proxy_url = random.choice(self.proxy_servers) |
|
proxy_config = {"server": proxy_url} |
|
proxy_username = os.getenv('PROXY_USER') |
|
proxy_password = os.getenv('PROXY_PASS') |
|
if proxy_username and proxy_password: |
|
proxy_config['username'] = proxy_username |
|
proxy_config['password'] = proxy_password |
|
launch_args['proxy'] = proxy_config |
|
|
|
browser = await p.chromium.launch(**launch_args) |
|
context = await browser.new_context( |
|
user_agent=random.choice(self.user_agents), |
|
viewport={"width": 1366, "height": 768}, |
|
locale=os.getenv("BROWSER_LOCALE", "en-US") |
|
) |
|
await context.route("**/*", lambda route: route.continue_()) |
|
page = await context.new_page() |
|
await page.add_init_script(""" |
|
Object.defineProperty(navigator, 'webdriver', { get: () => false }); |
|
window.navigator.chrome = { runtime: {}, app: { isInstalled: false } }; |
|
""") |
|
|
|
for attempt in range(self.retry_count): |
|
try: |
|
logging.info( |
|
f"Attempt {attempt + 1}: Loading OpenAI...") |
|
if not await self.load_page(page, "https://openai.com"): |
|
raise ScrapingError( |
|
"Failed to load key content on OpenAI website.") |
|
await self.human_interaction(page) |
|
await page.screenshot(path=f"openai_debug_{attempt}.png") |
|
content = await page.evaluate("""() => { |
|
const selectors = [ |
|
'main article', |
|
'#main-content', |
|
'section:first-of-type', |
|
'div[class*="content"]' |
|
]; |
|
|
|
let content = ''; |
|
for (const selector of selectors) { |
|
const element = document.querySelector(selector); |
|
if (element) { |
|
content += element.innerText + '\\n\\n'; |
|
} |
|
} |
|
return content.trim() || document.body.innerText; |
|
}""") |
|
if not content.strip(): |
|
raise ContentAnalysisError( |
|
"No content extracted from the page.") |
|
return content[:self.MAX_CONTENT_LENGTH] |
|
|
|
except (ScrapingError, ContentAnalysisError) as e: |
|
last_error = e |
|
logging.warning( |
|
f"Attempt {attempt + 1} failed: {str(e)}") |
|
if attempt < self.retry_count - 1: |
|
await asyncio.sleep(5) |
|
else: |
|
if browser: |
|
await browser.close() |
|
browser = None |
|
raise |
|
except Exception as e: |
|
last_error = e |
|
logging.exception(f"Unexpected error on attempt { |
|
attempt + 1}: {str(e)}") |
|
if attempt < self.retry_count - 1: |
|
await asyncio.sleep(5) |
|
else: |
|
if browser: |
|
await browser.close() |
|
browser = None |
|
raise |
|
|
|
except Exception as e: |
|
last_error = e |
|
finally: |
|
if browser: |
|
await browser.close() |
|
|
|
raise last_error if last_error else Exception( |
|
"All scraping attempts failed.") |
|
|
|
async def get_cached_content(self): |
|
key = 'openai_content_cache_key' |
|
content = cache.get(key) |
|
if content is None: |
|
content = await self.scrape_with_retry() |
|
cache.set(key, content, expire=int( |
|
os.getenv("CACHE_EXPIRY", 3600))) |
|
return content |
|
|
|
|
|
async def analyze_content(headless=True): |
|
try: |
|
scraper = EnhancedOpenAIScraper(headless=headless) |
|
content = await scraper.get_cached_content() |
|
|
|
client = OpenAI(api_key=EnhancedOpenAIScraper.API_KEY) |
|
if not client.api_key: |
|
raise ContentAnalysisError( |
|
"OpenAI API key not configured (check environment variables).") |
|
|
|
prompt_template = """ |
|
Analyze the following website content and extract the following information if present: |
|
|
|
1. **Overall Summary of the Website:** Provide a concise overview of the website's purpose and the main topics discussed. |
|
2. **Key Individuals or Entities:** Identify and briefly describe any prominent individuals, companies, or organizations mentioned. |
|
3. **Recent Announcements or Updates:** List any recent announcements, news, or updates found on the website, including dates if available. |
|
4. **Main Topics or Themes:** Identify the primary subjects or themes explored on the website. |
|
5. **Any Noteworthy Features or Projects:** Highlight any significant features, projects, or initiatives mentioned. |
|
|
|
Format the output clearly under each of these headings. If a particular piece of information is not found, indicate that it is not present. |
|
|
|
Content: |
|
{content} |
|
""" |
|
|
|
formatted_prompt = prompt_template.format(content=content) |
|
model_name = os.getenv("OPENAI_MODEL", "gpt-4-turbo") |
|
temperature = float(os.getenv("MODEL_TEMPERATURE", 0.3)) |
|
max_tokens = int(os.getenv("MAX_TOKENS", 1500)) |
|
top_p = float(os.getenv("MODEL_TOP_P", 0.9)) |
|
|
|
response = client.chat.completions.create( |
|
model=model_name, |
|
messages=[ |
|
{"role": "system", "content": "You are a helpful assistant that analyzes website content and extracts key information in a structured format."}, |
|
{"role": "user", "content": formatted_prompt} |
|
], |
|
temperature=temperature, |
|
max_tokens=max_tokens, |
|
top_p=top_p |
|
) |
|
|
|
if not response.choices: |
|
raise ContentAnalysisError("Empty response from GPT.") |
|
|
|
return response.choices[0].message.content |
|
|
|
except (ScrapingError, ContentAnalysisError) as e: |
|
logging.error(f"Analysis failed: {str(e)}") |
|
return f"Critical analysis error: {str(e)}" |
|
except Exception as e: |
|
logging.exception("Unexpected error during analysis.") |
|
return f"Unexpected analysis error: {str(e)}" |
|
|
|
|
|
async def main(): |
|
logging.basicConfig( |
|
level=os.getenv("LOG_LEVEL", "INFO").upper(), |
|
format='%(asctime)s - %(levelname)s - %(message)s' |
|
) |
|
|
|
# Start Prometheus HTTP server for exposing metrics |
|
try: |
|
prometheus_port = int(os.getenv("PROMETHEUS_PORT", 8000)) |
|
start_http_server(prometheus_port) |
|
logging.info(f"Prometheus metrics server started on port { |
|
prometheus_port}") |
|
except Exception as e: |
|
logging.warning(f"Failed to start Prometheus metrics server: {e}") |
|
|
|
start_time = time.time() |
|
result = await analyze_content(headless=True) |
|
end_time = time.time() |
|
|
|
print(f"\nAnalysis completed in {end_time - start_time:.2f} seconds\n") |
|
print(result) |
|
|
|
if __name__ == "__main__": |
|
asyncio.run(main())
|
|
|