Chariah
446 lines
18 KiB
Python
446 lines
18 KiB
Python
import os, json, asyncio, re, shutil, sqlite3, urllib.parse, random, logging
|
|
from datetime import datetime, timedelta
|
|
from fastapi import FastAPI, Query
|
|
from fastapi.middleware.cors import CORSMiddleware
|
|
import uvicorn
|
|
from playwright.async_api import async_playwright
|
|
|
|
logging.basicConfig(level=logging.INFO, format='%(asctime)s [%(levelname)s] %(message)s')
|
|
logger = logging.getLogger(__name__)
|
|
|
|
app = FastAPI()
|
|
app.add_middleware(
|
|
CORSMiddleware,
|
|
allow_origins=["http://localhost:3006", "http://127.0.0.1:3006"],
|
|
allow_methods=["POST"],
|
|
allow_headers=["*"],
|
|
)
|
|
PORT = int(os.getenv("PORT", "3008"))
|
|
|
|
OLLAMA_URL = os.getenv("OLLAMA_URL", "http://localhost:11434")
|
|
CLASSIFY_MODEL = os.getenv("CLASSIFY_MODEL", "dolphin-llama3:8b")
|
|
|
|
FX_PROFILE = os.getenv('FX_PROFILE', '')
|
|
|
|
BROAD_KEYWORDS = [
|
|
"website", "web design", "web develop", "web dev",
|
|
"build my website", "build a website", "create a website",
|
|
"need web", "looking for web", "new website",
|
|
"landing page", "wordpress",
|
|
"need a website", "my website", "business website",
|
|
"need a designer", "help with my website",
|
|
"redesign", "update my website",
|
|
"looking for", "need a", "need an", "looking to",
|
|
"help me build", "create my", "build me",
|
|
"design my", "make me a", "would like",
|
|
"need someone", "hire a", "looking to hire",
|
|
"want someone", "need help with",
|
|
]
|
|
|
|
def kw_match(text: str) -> bool:
|
|
t = text.lower()
|
|
return any(kw in t for kw in BROAD_KEYWORDS)
|
|
|
|
FB_SEARCHES = [
|
|
"looking for web developer",
|
|
"need a website designed",
|
|
"need website built South Africa",
|
|
"need someone to build my website",
|
|
"need web designer",
|
|
"looking for someone to create website",
|
|
"need ecommerce website built",
|
|
"want to hire web developer",
|
|
"website quote please",
|
|
"need wordpress website",
|
|
"I need a website for my business",
|
|
"need a site for my business",
|
|
]
|
|
|
|
VIEWPORTS = [
|
|
{'width': 1280, 'height': 800},
|
|
{'width': 1366, 'height': 768},
|
|
{'width': 1440, 'height': 900},
|
|
{'width': 1536, 'height': 864},
|
|
{'width': 1920, 'height': 1080},
|
|
]
|
|
|
|
async def get_fb_cookies(profile_path: str | None = None):
|
|
cookie_db_path = profile_path or FX_PROFILE
|
|
if not cookie_db_path:
|
|
logger.warning("No profile path provided")
|
|
return []
|
|
cookie_db = os.path.join(cookie_db_path, 'cookies.sqlite')
|
|
if not os.path.exists(cookie_db):
|
|
logger.warning("Cookie DB not found at %s", cookie_db)
|
|
return []
|
|
tmp = os.path.join(os.path.dirname(cookie_db), f'cookies_tmp_{os.getpid()}.sqlite')
|
|
for attempt in range(3):
|
|
try:
|
|
shutil.copy2(cookie_db, tmp)
|
|
break
|
|
except PermissionError:
|
|
if attempt < 2:
|
|
await asyncio.sleep(0.5)
|
|
else:
|
|
logger.error("Failed to copy cookie DB after 3 attempts")
|
|
return []
|
|
try:
|
|
conn = sqlite3.connect(tmp)
|
|
c = conn.cursor()
|
|
c.execute("SELECT name, value, host, path FROM moz_cookies WHERE host LIKE '%facebook.com'")
|
|
rows = c.fetchall()
|
|
conn.close()
|
|
try:
|
|
os.remove(tmp)
|
|
except Exception:
|
|
pass
|
|
return [{
|
|
"name": name, "value": value,
|
|
"domain": host if host.startswith('.') else f'.{host}',
|
|
"path": path,
|
|
"httpOnly": True, "secure": True, "sameSite": "Lax",
|
|
} for name, value, host, path in rows]
|
|
except Exception as e:
|
|
logger.error("Cookie DB read error: %s", e)
|
|
try:
|
|
os.remove(tmp)
|
|
except Exception:
|
|
pass
|
|
return []
|
|
|
|
WEEKDAY_ORDER = ['monday', 'tuesday', 'wednesday', 'thursday', 'friday', 'saturday', 'sunday']
|
|
|
|
def _parse_fb_date(block: list[str]) -> str:
|
|
for l in block:
|
|
l = l.strip()
|
|
m = re.match(r'(\d+)\s*(h|hr|hrs|hour|hours)\s*ago', l)
|
|
if m:
|
|
hours = int(m.group(1))
|
|
d = datetime.now() - timedelta(hours=hours)
|
|
return d.strftime('%Y-%m-%d')
|
|
if l.lower() == 'yesterday':
|
|
d = datetime.now() - timedelta(days=1)
|
|
return d.strftime('%Y-%m-%d')
|
|
low = l.lower()
|
|
if low in WEEKDAY_ORDER:
|
|
day_idx = WEEKDAY_ORDER.index(low)
|
|
today_idx = datetime.now().weekday()
|
|
days_ago = (today_idx - day_idx) % 7
|
|
d = datetime.now() - timedelta(days=days_ago)
|
|
return d.strftime('%Y-%m-%d')
|
|
for fmt in ['%Y-%m-%d', '%d %B %Y', '%B %d %Y', '%d %b %Y', '%b %d %Y', '%d %b', '%b %d']:
|
|
try:
|
|
dt = datetime.strptime(l, fmt)
|
|
return dt.strftime('%Y-%m-%d')
|
|
except ValueError:
|
|
pass
|
|
return datetime.now().strftime('%Y-%m-%d')
|
|
|
|
def _extract_posts_from_text(raw: str, url: str) -> list[dict]:
|
|
lines = [l.strip() for l in raw.split('\n')]
|
|
blocks = []
|
|
cur = []
|
|
fb_run = 0
|
|
for l in lines:
|
|
if 'facebook' in l.lower():
|
|
fb_run += 1
|
|
if cur and fb_run >= 2:
|
|
blocks.append(cur)
|
|
cur = []
|
|
else:
|
|
fb_run = 0
|
|
if l and len(l) >= 15:
|
|
words = re.findall(r'[A-Za-z]{2,}', l)
|
|
if len(words) >= 2:
|
|
cur.append(l)
|
|
if cur:
|
|
blocks.append(cur)
|
|
|
|
posts = []
|
|
seen_texts = set()
|
|
for block in blocks:
|
|
if not block:
|
|
continue
|
|
kw_indices = [i for i, l in enumerate(block) if kw_match(l)]
|
|
if not kw_indices:
|
|
continue
|
|
i = kw_indices[0]
|
|
start = max(0, i - 2)
|
|
end = min(len(block), i + 5)
|
|
snippet = ' '.join(block[start:end])
|
|
if len(snippet) < 40:
|
|
continue
|
|
dekey = snippet[:80]
|
|
if dekey in seen_texts:
|
|
continue
|
|
seen_texts.add(dekey)
|
|
posts.append({
|
|
"title": snippet[:300],
|
|
"content": snippet[:1000],
|
|
"author": block[start] if start < i else '',
|
|
"url": url,
|
|
"source": "facebook",
|
|
"date": _parse_fb_date(block),
|
|
})
|
|
return posts
|
|
|
|
async def human_scroll(page, steps: int = None, total_delay: float = None):
|
|
steps = steps or random.randint(2, 5)
|
|
total_delay = total_delay or random.uniform(6, 18)
|
|
step_delay = total_delay / steps
|
|
for i in range(steps):
|
|
scroll_dist = random.randint(200, 600)
|
|
await page.evaluate(f"window.scrollBy(0, {scroll_dist})")
|
|
await page.wait_for_timeout(int(step_delay * 1000))
|
|
if random.random() < 0.3:
|
|
await page.evaluate(f"window.scrollBy(0, -{random.randint(100, 300)})")
|
|
await page.wait_for_timeout(random.randint(1000, 3000))
|
|
|
|
async def random_idle(page):
|
|
action = random.random()
|
|
if action < 0.15:
|
|
try:
|
|
elems = await page.query_selector_all('a, span, div[role="button"]')
|
|
if elems:
|
|
el = random.choice(elems[:20])
|
|
box = await el.bounding_box()
|
|
if box:
|
|
x = box['x'] + box['width'] / 2 + random.randint(-20, 20)
|
|
y = box['y'] + box['height'] / 2 + random.randint(-10, 10)
|
|
await page.mouse.move(x, y, steps=random.randint(20, 40))
|
|
await page.wait_for_timeout(random.randint(500, 2000))
|
|
except Exception:
|
|
pass
|
|
|
|
async def search_facebook(page, query: str) -> list[dict]:
|
|
url = f'https://www.facebook.com/search/posts/?q={urllib.parse.quote(query)}'
|
|
try:
|
|
await page.goto(url, wait_until='domcontentloaded', timeout=30000)
|
|
await page.wait_for_timeout(random.randint(3000, 8000))
|
|
|
|
await human_scroll(page, steps=random.randint(2, 4), total_delay=random.uniform(6, 15))
|
|
|
|
# Accidental overscroll: 10% chance to jump to bottom and back
|
|
if random.random() < 0.1:
|
|
await page.evaluate("window.scrollTo(0, document.body.scrollHeight)")
|
|
await page.wait_for_timeout(random.randint(1000, 3000))
|
|
await page.evaluate("window.scrollBy(0, -300)")
|
|
await page.wait_for_timeout(random.randint(1000, 3000))
|
|
|
|
if random.random() < 0.2:
|
|
await page.evaluate("window.scrollTo(0, 0)")
|
|
await page.wait_for_timeout(random.randint(2000, 5000))
|
|
|
|
if random.random() < 0.3:
|
|
await random_idle(page)
|
|
|
|
raw = await page.evaluate('document.body.innerText')
|
|
except Exception as e:
|
|
logger.warning("Facebook search failed: %s", e)
|
|
return []
|
|
return _extract_posts_from_text(raw, url)
|
|
|
|
def cleanup_chrome():
|
|
import subprocess, signal
|
|
try:
|
|
subprocess.run(["taskkill", "/F", "/IM", "chrome-headless-shell.exe"], capture_output=True, timeout=5)
|
|
except Exception:
|
|
pass
|
|
|
|
async def scrape_facebook(profile_path: str | None = None, force: bool = False) -> dict:
|
|
cleanup_chrome()
|
|
fb_cookies = await get_fb_cookies(profile_path)
|
|
if not fb_cookies:
|
|
logger.warning("No Facebook cookies available")
|
|
return {"success": False, "leads": [], "flagged": False, "flag_reason": None, "error": "No cookies available"}
|
|
try:
|
|
async with async_playwright() as pw:
|
|
browser = await pw.chromium.launch(
|
|
headless=True,
|
|
args=['--disable-blink-features=AutomationControlled']
|
|
)
|
|
viewport = random.choice(VIEWPORTS)
|
|
context = await browser.new_context(
|
|
user_agent='Mozilla/5.0 (Windows NT 10.0; Win64; x64; rv:150.0) Gecko/20100101 Firefox/150.0',
|
|
viewport=viewport,
|
|
)
|
|
|
|
await context.add_init_script("""
|
|
Object.defineProperty(navigator, 'webdriver', { get: () => false });
|
|
Object.defineProperty(navigator, 'languages', { get: () => ['en-US', 'en'] });
|
|
Object.defineProperty(navigator, 'platform', { get: () => 'Win32' });
|
|
window.chrome = { runtime: {} };
|
|
Object.defineProperty(navigator, 'hardwareConcurrency', { get: () => 8 });
|
|
Object.defineProperty(navigator, 'deviceMemory', { get: () => 8 });
|
|
""")
|
|
for c in fb_cookies:
|
|
try:
|
|
await context.add_cookies([c])
|
|
except Exception as e:
|
|
logger.warning("Failed to inject cookie %s: %s", c.get('name'), e)
|
|
|
|
page = await context.new_page()
|
|
# Navigate through google first for a legitimate Referer header
|
|
await page.goto('https://www.google.com/', wait_until='domcontentloaded', timeout=15000)
|
|
await page.wait_for_timeout(random.randint(1000, 3000))
|
|
await page.goto('https://www.facebook.com/', wait_until='domcontentloaded', timeout=30000)
|
|
await page.wait_for_timeout(random.randint(3000, 8000))
|
|
|
|
url = page.url
|
|
if '/login' in url.lower():
|
|
logger.warning("Facebook login page detected — account flagged")
|
|
await browser.close()
|
|
return {"success": False, "leads": [], "flagged": True, "flag_reason": "login_page", "error": "Facebook login page detected"}
|
|
|
|
# Browse feed like a human before searching
|
|
await human_scroll(page, steps=random.randint(2, 4), total_delay=random.uniform(8, 20))
|
|
if random.random() < 0.25:
|
|
await page.evaluate("window.scrollTo(0, 0)")
|
|
await page.wait_for_timeout(random.randint(2000, 5000))
|
|
await human_scroll(page, steps=random.randint(1, 2))
|
|
|
|
# False start: 30% chance to idle and leave without scraping (skipped when force=true)
|
|
if not force and random.random() < 0.3:
|
|
await page.wait_for_timeout(random.randint(8000, 20000))
|
|
await browser.close()
|
|
return {"success": True, "leads": [], "flagged": False, "flag_reason": None, "error": None}
|
|
|
|
all_posts = []
|
|
searches = random.sample(FB_SEARCHES, k=random.randint(3, 6))
|
|
for i, query in enumerate(searches):
|
|
try:
|
|
posts = await search_facebook(page, query)
|
|
all_posts.extend(posts)
|
|
# Between searches: random break with possible small scroll
|
|
if random.random() < 0.4:
|
|
await page.evaluate(f"window.scrollBy(0, {random.randint(-300, 300)})")
|
|
delay = random.uniform(8, 25)
|
|
await page.wait_for_timeout(int(delay * 1000))
|
|
# Tab switch: 15% chance after 2nd-3rd search
|
|
if i == random.randint(1, 2) and random.random() < 0.15:
|
|
new_page = await context.new_page()
|
|
await new_page.goto('https://www.messenger.com/', wait_until='domcontentloaded', timeout=15000)
|
|
await new_page.wait_for_timeout(random.randint(3000, 8000))
|
|
await new_page.close()
|
|
await page.wait_for_timeout(random.randint(1000, 3000))
|
|
except Exception as e:
|
|
logger.warning("Facebook search '%s' failed: %s", query, e)
|
|
|
|
# Idle before closing — human would pause
|
|
if random.random() < 0.5:
|
|
await page.wait_for_timeout(random.randint(3000, 10000))
|
|
|
|
await browser.close()
|
|
|
|
seen = set()
|
|
deduped = []
|
|
for p in all_posts:
|
|
key = p.get('content', '')[:100]
|
|
if key not in seen:
|
|
seen.add(key)
|
|
deduped.append(p)
|
|
|
|
leads = deduped[:20]
|
|
if leads:
|
|
leads = await classify_leads(leads)
|
|
|
|
return {"success": True, "leads": leads[:15], "flagged": False, "flag_reason": None, "error": None}
|
|
except Exception as e:
|
|
logger.error("Facebook scrape failed: %s", e)
|
|
return {"success": False, "leads": [], "flagged": False, "flag_reason": None, "error": str(e)}
|
|
|
|
async def ask_ollama(prompt: str) -> str:
|
|
import httpx
|
|
async with httpx.AsyncClient(timeout=120) as c:
|
|
r = await c.post(f"{OLLAMA_URL}/api/chat", json={
|
|
"model": CLASSIFY_MODEL,
|
|
"messages": [
|
|
{"role": "system", "content": "You classify forum posts. Return only valid JSON."},
|
|
{"role": "user", "content": prompt}
|
|
],
|
|
"stream": False,
|
|
"options": {"temperature": 0.05, "num_predict": 1024},
|
|
})
|
|
r.raise_for_status()
|
|
data = r.json()
|
|
return data["message"]["content"]
|
|
|
|
async def classify_leads(results: list[dict]) -> list[dict]:
|
|
if not results:
|
|
return []
|
|
briefs = [r["title"][:200] for r in results]
|
|
prompt = f"""Classify each post as LEAD or NOT.
|
|
LEAD = someone REQUESTING/POSTING/WANTING a website built, designed, or created for them.
|
|
LEAD examples: "Need a website for my business", "Looking for web developer to build my site", "I need someone to create my website", "Want a new website for my company", "Looking for someone to design my WordPress site"
|
|
|
|
NOT LEAD:
|
|
- Offering web design services: "I build websites", "I offer web design", "Affordable web design packages"
|
|
- Already have a website and need marketing, SEO, content, video, link building, email marketing, affiliates
|
|
- Recruiting employees, hiring staff, looking for business partners
|
|
- Selling products, promoting services, affiliate offers
|
|
- "Need web hosting", "Looking for a partner", "Looking for content writer", "Video spokesperson"
|
|
|
|
For each numbered post, answer ONLY "yes" (LEAD) or "no" (NOT LEAD):
|
|
{chr(10).join(f'{i+1}. {t}' for i, t in enumerate(briefs))}
|
|
Return a JSON array like ["yes","no","yes"] matching the order above."""
|
|
ai_succeeded = False
|
|
try:
|
|
raw = await ask_ollama(prompt)
|
|
raw = raw.strip()
|
|
if raw.startswith("```"):
|
|
first_nl = raw.find('\n')
|
|
if first_nl != -1:
|
|
raw = raw[first_nl + 1:]
|
|
if raw.endswith("```"):
|
|
raw = raw[:-3]
|
|
raw = raw.strip()
|
|
answers = json.loads(raw)
|
|
if isinstance(answers, list) and len(answers) == len(results):
|
|
ai_succeeded = True
|
|
filtered = []
|
|
for i, ans in enumerate(answers):
|
|
if isinstance(ans, str) and ans.lower() == 'yes':
|
|
filtered.append(results[i])
|
|
if filtered:
|
|
return filtered[:10]
|
|
logger.info("AI classified all %d items as NOT LEAD — returning empty", len(results))
|
|
return []
|
|
except Exception as e:
|
|
logger.warning("AI classification failed, falling back to keyword filter: %s", e)
|
|
|
|
if not ai_succeeded:
|
|
strict_keywords = [
|
|
"website", "web design", "web develop", "web dev",
|
|
"build my website", "build a website", "create a website",
|
|
"need web", "looking for web", "new website",
|
|
"landing page", "wordpress",
|
|
"need a website", "my website", "business website",
|
|
"need a designer", "help with my website",
|
|
"redesign", "update my website",
|
|
]
|
|
filtered = []
|
|
for r in results:
|
|
t = r['title'].lower()
|
|
if not any(kw in t for kw in strict_keywords):
|
|
continue
|
|
if any(kw in t for kw in ['i build', 'i offer', 'i create', 'i am a', 'web developer available',
|
|
'affordable web', 'web hosting', 'free website',
|
|
'limited time', 'special offer', 'sign up now',
|
|
'offer']):
|
|
continue
|
|
filtered.append(r)
|
|
return filtered[:10]
|
|
return []
|
|
|
|
@app.get("/health")
|
|
async def health():
|
|
return {"status": "ok"}
|
|
|
|
@app.post("/scrape/facebook")
|
|
async def scrape_facebook_endpoint(profile_path: str | None = Query(None), force: bool = Query(False)):
|
|
result = await scrape_facebook(profile_path, force)
|
|
return result
|
|
|
|
if __name__ == "__main__":
|
|
uvicorn.run(app, host="0.0.0.0", port=PORT)
|