added crawler
ci/woodpecker/push/woodpecker Pipeline was successful
Details
ci/woodpecker/push/woodpecker Pipeline was successful
Details
This commit is contained in:
coleWesterveld
parent
9dea0bac65
commit
f06d01613f
|
|
@ -0,0 +1,2 @@
|
|||
.env
|
||||
marketline_cookies.json
|
||||
|
|
@ -0,0 +1,122 @@
|
|||
import google.generativeai as genai
|
||||
import json
|
||||
import os
|
||||
import time
|
||||
from dotenv import load_dotenv
|
||||
load_dotenv()
|
||||
|
||||
GOOGLE_API_KEY = os.environ.get("GOOGLE_API_KEY")
|
||||
|
||||
# json generated by the scraper (markeltine_crawler.py)
|
||||
INPUT_FILE = os.path.join("crawl_results", "successful_pages.json")
|
||||
|
||||
# output JSON any extracted deals from the scraped data
|
||||
OUTPUT_FILE = os.path.join("crawl_results", "extracted_arms_deals.json")
|
||||
|
||||
MODEL_NAME = "gemini-2.0-flash-lite"
|
||||
|
||||
# TODO: refine
|
||||
EXTRACTION_PROMPT = """
|
||||
From the document text provided below, extract key details about any military or arms exports.
|
||||
|
||||
Your task is to identify the following:
|
||||
- "company_name": The name of the company involved in manufacturing or selling.
|
||||
- "weapon_system": The specific type of weapon, vehicle, or military equipment.
|
||||
- "destination_country": The country receiving the goods.
|
||||
- "sale_value": The monetary value of the deal, including currency (e.g., "$15 Billion CAD").
|
||||
- "summary": A concise, one-sentence summary of the export deal or report.
|
||||
|
||||
If a specific piece of information cannot be found in the text, you MUST use the value "Not Found".
|
||||
|
||||
Provide your response as a single, clean JSON object. Do not add any explanatory text before or after the JSON.
|
||||
|
||||
---
|
||||
DOCUMENT TEXT:
|
||||
{text_content}
|
||||
"""
|
||||
|
||||
def load_scraped_data(filepath):
|
||||
"""Loads the scraped data from the JSON file."""
|
||||
try:
|
||||
with open(filepath, "r", encoding="utf-8") as f:
|
||||
return json.load(f)
|
||||
except FileNotFoundError:
|
||||
print(f"❌ Error: Input file not found at '{filepath}'.")
|
||||
print("Ensure you have run the scraper first.")
|
||||
return None
|
||||
|
||||
def save_extracted_data(filepath, data):
|
||||
"""Saves the final extracted data to a new JSON file."""
|
||||
with open(filepath, "w", encoding="utf-8") as f:
|
||||
json.dump(data, f, indent=4, ensure_ascii=False)
|
||||
print(f"\n✅ Success! Saved extracted info to '{filepath}'.")
|
||||
|
||||
|
||||
def process_content_with_gemini(text_content):
|
||||
"""
|
||||
Sends the text to the Gemini API with the extraction prompt and
|
||||
parses the JSON response.
|
||||
"""
|
||||
model = genai.GenerativeModel(MODEL_NAME)
|
||||
prompt = EXTRACTION_PROMPT.format(text_content=text_content)
|
||||
|
||||
try:
|
||||
response = model.generate_content(prompt)
|
||||
# Clean the response to ensure it's valid JSON. Gemini sometimes
|
||||
# wraps its JSON response in markdown backticks.
|
||||
clean_json = response.text.strip().replace("```json", "").replace("```", "")
|
||||
# print("GOT: ", clean_json)
|
||||
return json.loads(clean_json)
|
||||
except Exception as e:
|
||||
print(f" ❌ An error occurred while calling Gemini or parsing its response: {e}")
|
||||
return {"error": str(e)}
|
||||
|
||||
|
||||
def main():
|
||||
"""Main function to run the data extraction process."""
|
||||
if not GOOGLE_API_KEY:
|
||||
print("❌ Error: GOOGLE_API_KEY environment variable not set.")
|
||||
return
|
||||
|
||||
genai.configure(api_key=GOOGLE_API_KEY)
|
||||
|
||||
scraped_pages = load_scraped_data(INPUT_FILE)
|
||||
if not scraped_pages:
|
||||
print("❌ Error: No scraper results found. Run marketline_crawler.py to generate crawl_results/successful_pages.json")
|
||||
return
|
||||
|
||||
all_extracted_deals = []
|
||||
total_pages = len(scraped_pages)
|
||||
|
||||
print(f"🤖 Starting information extraction with Gemini for {total_pages} pages...")
|
||||
|
||||
for i, page in enumerate(scraped_pages):
|
||||
print(f"\nProcessing page {i+1}/{total_pages}: {page['url']}")
|
||||
|
||||
# Avoid processing pages with very little text
|
||||
if len(page.get('content', '')) < 150:
|
||||
print(" ⏩ Skipping page due to insufficient content.")
|
||||
continue
|
||||
|
||||
extracted_info = process_content_with_gemini(page['content'])
|
||||
|
||||
# Check if the extraction was successful and contains actual data
|
||||
if extracted_info and "error" not in extracted_info:
|
||||
if extracted_info.get("company_name") != "Not Found" or extracted_info.get("weapon_system") != "Not Found":
|
||||
print(f" ✔️ Found relevant info: {extracted_info.get('company_name', 'N/A')} | {extracted_info.get('weapon_system', 'N/A')}")
|
||||
# Add the source URL for reference
|
||||
extracted_info['source_url'] = page['url']
|
||||
all_extracted_deals.append(extracted_info)
|
||||
else:
|
||||
print(" ⚪ No relevant deals found on this page.")
|
||||
|
||||
# Add a small delay to respect API rate limits (1 second is safe)
|
||||
time.sleep(1)
|
||||
|
||||
if all_extracted_deals:
|
||||
save_extracted_data(OUTPUT_FILE, all_extracted_deals)
|
||||
else:
|
||||
print("\nNo relevant deals were extracted from any of the pages.")
|
||||
|
||||
if __name__ == "__main__":
|
||||
main()
|
||||
|
|
@ -0,0 +1,26 @@
|
|||
# a quick little side script to look into the results
|
||||
# NOT used for main workflow
|
||||
|
||||
import asyncio
|
||||
from playwright.async_api import async_playwright
|
||||
import json
|
||||
import os
|
||||
from crawl4ai import BrowserConfig, AsyncWebCrawler, CrawlerRunConfig
|
||||
from crawl4ai.deep_crawling import BFSDeepCrawlStrategy
|
||||
from crawl4ai.content_scraping_strategy import LXMLWebScrapingStrategy
|
||||
from crawl4ai.deep_crawling.scorers import KeywordRelevanceScorer
|
||||
|
||||
|
||||
# check how many pages are invalid password pages (it was not many -- like 7/100)
|
||||
with open("crawl_results/successful_pages.json", "r") as f:
|
||||
results = json.load(f)
|
||||
|
||||
counter = 0
|
||||
total = 0
|
||||
for result in results:
|
||||
total += 1
|
||||
if not "password is invalid" in result['content']:
|
||||
print("\n\n\n FOUND: \n", result['content'])
|
||||
counter+= 1
|
||||
|
||||
print(f"\n\n\n FINAL GOOD: {counter} OF {total} RESULTS")
|
||||
|
|
@ -0,0 +1 @@
|
|||
*.json
|
||||
|
|
@ -0,0 +1,183 @@
|
|||
import asyncio
|
||||
from playwright.async_api import async_playwright, Page
|
||||
import json
|
||||
import os
|
||||
from crawl4ai import BrowserConfig, AsyncWebCrawler, CrawlerRunConfig
|
||||
from crawl4ai.deep_crawling import BFSDeepCrawlStrategy
|
||||
from crawl4ai.content_scraping_strategy import ContentScrapingStrategy, ScrapingResult, LXMLWebScrapingStrategy
|
||||
from crawl4ai.processors.pdf import PDFContentScrapingStrategy
|
||||
from crawl4ai.deep_crawling.scorers import KeywordRelevanceScorer
|
||||
from crawl4ai.deep_crawling.filters import URLPatternFilter
|
||||
from datetime import datetime
|
||||
|
||||
# --- CONFIGURATION ---
|
||||
|
||||
# TODO: this will need to change for different organizations (ie univiersities)
|
||||
|
||||
# make this the link for university login when accessing marketline
|
||||
LOGIN_URL = "https://login.microsoftonline.com/be62a12b-2cad-49a1-a5fa-85f4f3156a7d/saml2?SAMLRequest=fZLBbtswEER%2FReBdokhJjk1YBtz4UANpasRODrkUK2plE6BIlUsl7d9Xtls0ufhIcPhmZ5ZLgt4Oaj3Gk3vCnyNSTH711pG6XNRsDE55IEPKQY%2Bkolb79bcHJbNcDcFHr71lyZoIQzTe3XtHY49hj%2BHNaHx%2BeqjZKcaBFOcwmWTWNNnojyPa4ZRp4NP5%2BE4D359M03iLk4TI87OH5Lvv%2BwNLNtNQxsEZ%2Fx9m%2FdG4rDc6ePJd9M4ah5n2PW9wJkHIJpUa2rRcgEih6iCdV13ZFaKawV3Lz%2BkkS7abmv1Y6Hne5lgVQhSFkPlMgABcdHPddvOmnE0yohG3jiK4WDOZyyrN71K5OEipylIV8pUlu79lfDGuNe54u7nmKiL19XDYpdeYLxjoEnESsNXyPKG6GIcPG7mNhX9rYKubpdOw5B%2F4V7NBPU7A7WbnrdG%2Fk7W1%2Fv0%2BIESsmWB8dX3y%2Ba%2Bs%2FgA%3D&RelayState=https%3A%2F%2Fauth.lib.uoguelph.ca%2Fopenathens%2Fsaml%2F%3Fuuid%3Db3nuk1o5lh78w6j657yd773oxfeqzc0v%26csrfmiddlewaretoken%3D4EzWMhPgP6L5YXtK3FGIgKKQ5KguVDwOuod2abzLQRV6kagUu0BBVWsJVI8N78tT%26opshib%3DLogin%2Bwith%2Byour%2BGryphmail%2BPassword%26staff_mode%3DTrue&sso_reload=true"
|
||||
|
||||
# shouldnt need to change. this is what we will wait for to load after logging in to trigger saving cookies.
|
||||
HOMEPAGE_URL = "https://advantage.marketline.com/HomePage/Home"
|
||||
|
||||
# the root page to seed crawling
|
||||
CRAWLPAGE_URL = "https://advantage.marketline.com/Search?industry=2800001"
|
||||
|
||||
# name of file where cookies are saved
|
||||
COOKIES_FILE = "marketline_cookies.json"
|
||||
|
||||
# --- CRAWLER SETTINGS ---
|
||||
DEPTH = 2
|
||||
COUNT = 10 # Increased for better testing
|
||||
|
||||
# TODO: maybe make this list more comprehensive?
|
||||
SCRAPER_KEYWORDS = [
|
||||
# Core Terms
|
||||
"arms export", "arms sale", "arms trade", "weapons export", "weapons deal",
|
||||
"military export", "defence contract", "defense contract",
|
||||
|
||||
# Canadian Context
|
||||
"canadian armed forces", "global affairs canada", "canadian defence",
|
||||
"canadian military", "royal canadian navy", "royal canadian air force",
|
||||
|
||||
# Equipment & Technology
|
||||
"armoured vehicle", "light armoured vehicle", "lav", "naval ship", "warship",
|
||||
"frigate", "fighter jet", "military aircraft", "surveillance", "radar",
|
||||
"artillery", "munitions", "firearms", "aerospace",
|
||||
|
||||
# Action & Policy Terms
|
||||
"procurement", "acquisition", "military aid", "export permit", "itar"
|
||||
]
|
||||
|
||||
# runs login process and saves cookies so that we can run the scraping with authentication
|
||||
async def login_and_save_cookies():
|
||||
|
||||
async with async_playwright() as p:
|
||||
browser = await p.chromium.launch(headless=False)
|
||||
context = await browser.new_context()
|
||||
page = await context.new_page()
|
||||
|
||||
try:
|
||||
await page.goto(LOGIN_URL)
|
||||
await page.wait_for_url(HOMEPAGE_URL, timeout=300000)
|
||||
|
||||
print("Login detected. Saving session cookies...")
|
||||
cookies = await context.cookies()
|
||||
with open(COOKIES_FILE, "w") as f:
|
||||
json.dump(cookies, f)
|
||||
|
||||
print("Cookies saved successfully!")
|
||||
await crawl_with_saved_cookies()
|
||||
|
||||
except Exception as e:
|
||||
print(f"Login failed: {e}")
|
||||
print("Error details:")
|
||||
print(await page.content())
|
||||
|
||||
finally:
|
||||
await context.close()
|
||||
await browser.close()
|
||||
|
||||
def save_results_to_json(successful_data, failed_pages):
|
||||
"""
|
||||
Saves the successful and failed crawl results into separate JSON files
|
||||
in a dedicated directory.
|
||||
"""
|
||||
output_dir = "crawl_results"
|
||||
os.makedirs(output_dir, exist_ok=True)
|
||||
print(f"\n💾 Saving results to '{output_dir}' directory...")
|
||||
|
||||
# Define file paths
|
||||
successful_file = os.path.join(output_dir, "successful_pages.json")
|
||||
failed_file = os.path.join(output_dir, "failed_pages.json")
|
||||
|
||||
# Save successfully scraped data
|
||||
with open(successful_file, "w", encoding="utf-8") as f:
|
||||
json.dump(successful_data, f, indent=4, ensure_ascii=False)
|
||||
print(f" Saved data for {len(successful_data)} successful pages to '{successful_file}'")
|
||||
|
||||
# Save failed pages if any
|
||||
if failed_pages:
|
||||
with open(failed_file, "w", encoding="utf-8") as f:
|
||||
json.dump(failed_pages, f, indent=4, ensure_ascii=False)
|
||||
print(f" Saved info for {len(failed_pages)} failed pages to '{failed_file}'")
|
||||
|
||||
|
||||
# runs the crawler with the cookies collected during login
|
||||
async def crawl_with_saved_cookies():
|
||||
|
||||
if not os.path.exists(COOKIES_FILE):
|
||||
print("No cookies found. Please run login first.")
|
||||
return
|
||||
|
||||
with open(COOKIES_FILE, "r") as f:
|
||||
cookies = json.load(f)
|
||||
|
||||
browser_config = BrowserConfig(cookies=cookies)
|
||||
|
||||
config = CrawlerRunConfig(
|
||||
deep_crawl_strategy=BFSDeepCrawlStrategy(
|
||||
max_depth=DEPTH,
|
||||
max_pages=COUNT,
|
||||
url_scorer=KeywordRelevanceScorer(keywords=SCRAPER_KEYWORDS,),
|
||||
),
|
||||
scraping_strategy=LXMLWebScrapingStrategy(),
|
||||
# TODO: scrape the PDFs better
|
||||
# scraping_strategy=PDFCrawlerStrategy(),
|
||||
verbose=True,
|
||||
stream=True,
|
||||
page_timeout=30000
|
||||
)
|
||||
|
||||
successful_data = []
|
||||
failed_pages = []
|
||||
|
||||
async with AsyncWebCrawler(config=browser_config) as crawler:
|
||||
async for result in await crawler.arun(CRAWLPAGE_URL, config=config):
|
||||
if result.success:
|
||||
depth = result.metadata.get("depth", 0)
|
||||
score = result.metadata.get("score", 0)
|
||||
|
||||
# here we could look at a few things, the HTML, markdown, raw text, etc.
|
||||
scraped_content = result.markdown
|
||||
|
||||
print(f"✅ Depth {depth} | Score: {score:.2f} | {result.url}")
|
||||
# NEW: Print a preview of the content to confirm it's being scraped
|
||||
print(f" 📄 Content length: {len(scraped_content)}. Preview: {scraped_content[:120]}...")
|
||||
|
||||
successful_data.append({
|
||||
"url": result.url,
|
||||
"content": scraped_content,
|
||||
"depth": depth,
|
||||
"score": round(score, 2)
|
||||
})
|
||||
else:
|
||||
failed_pages.append({
|
||||
'url': result.url,
|
||||
'error': result.error_message,
|
||||
'depth': result.metadata.get("depth", 0)
|
||||
})
|
||||
print(f"❌ Failed: {result.url} - {result.error_message}")
|
||||
|
||||
print(f"📊 Results: {len(successful_data)} successful, {len(failed_pages)} failed")
|
||||
|
||||
save_results_to_json(successful_data, failed_pages)
|
||||
|
||||
# Analyze failures by depth
|
||||
if failed_pages:
|
||||
failure_by_depth = {}
|
||||
for failure in failed_pages:
|
||||
depth = failure['depth']
|
||||
failure_by_depth[depth] = failure_by_depth.get(depth, 0) + 1
|
||||
|
||||
print("❌ Failures by depth:")
|
||||
for depth, count in sorted(failure_by_depth.items()):
|
||||
print(f" Depth {depth}: {count} failures")
|
||||
|
||||
if __name__ == "__main__":
|
||||
# Choose which function to run
|
||||
# 1. First, run the login function once to get your cookies
|
||||
asyncio.run(login_and_save_cookies())
|
||||
|
||||
# 2. Then, comment out the login line and run the crawl
|
||||
# asyncio.run(crawl_with_saved_cookies())
|
||||
|
|
@ -0,0 +1,29 @@
|
|||
beautifulsoup4==4.13.4
|
||||
cachetools==5.5.2
|
||||
certifi==2025.7.14
|
||||
charset-normalizer==3.4.2
|
||||
dotenv==0.9.9
|
||||
google==3.0.0
|
||||
google-ai-generativelanguage==0.1.0
|
||||
google-api-core==2.25.1
|
||||
google-api-python-client==2.177.0
|
||||
google-auth==2.40.3
|
||||
google-auth-httplib2==0.2.0
|
||||
google-generativeai==0.1.0rc1
|
||||
googleapis-common-protos==1.70.0
|
||||
grpcio==1.70.0
|
||||
grpcio-status==1.62.3
|
||||
httplib2==0.22.0
|
||||
idna==3.10
|
||||
proto-plus==1.26.1
|
||||
protobuf==4.25.8
|
||||
pyasn1==0.6.1
|
||||
pyasn1-modules==0.4.2
|
||||
pyparsing==3.1.4
|
||||
python-dotenv==1.0.1
|
||||
requests==2.32.4
|
||||
rsa==4.9.1
|
||||
soupsieve==2.7
|
||||
typing-extensions==4.13.2
|
||||
uritemplate==4.1.1
|
||||
urllib3==2.2.3
|
||||
Loading…
Reference in New Issue