Restore file structure to last legit commit b245b35, remove remote CRM_ENVR-main files
This commit is contained in:
+434
-231
@@ -1,10 +1,5 @@
|
||||
import os, json, asyncio, re, shutil, sqlite3, urllib.parse, random, logging, httpx
|
||||
import os, json, asyncio, re, shutil, sqlite3, urllib.parse, random, logging
|
||||
from datetime import datetime, timedelta
|
||||
from dotenv import load_dotenv
|
||||
|
||||
dotenv_path = os.path.join(os.path.dirname(__file__), '..', '.env.local')
|
||||
if os.path.exists(dotenv_path):
|
||||
load_dotenv(dotenv_path)
|
||||
from fastapi import FastAPI, Query
|
||||
from fastapi.middleware.cors import CORSMiddleware
|
||||
import uvicorn
|
||||
@@ -24,10 +19,8 @@ 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', '')
|
||||
BU_API_KEY = os.getenv('BU_API_KEY', '')
|
||||
|
||||
BU_API_BASE = "https://api.browser-use.com/api/v3"
|
||||
FX_PROFILE = os.getenv('FX_PROFILE', '')
|
||||
|
||||
BROAD_KEYWORDS = [
|
||||
"website", "web design", "web develop", "web dev",
|
||||
@@ -45,164 +38,107 @@ BROAD_KEYWORDS = [
|
||||
]
|
||||
|
||||
OFFER_PATTERNS = [
|
||||
r"\bfree\s+domain\b", r"\bweb\s+hosting\b", r"\bfree\s+website\b",
|
||||
r"\bprofessional\s+website", r"\baffordable\s+web\b", r"\bget\s+a\s+free\b",
|
||||
r"\bsee\s+more\b", r"\bbook\s+your\b", r"\bhire\s+a\s+freelancer\b",
|
||||
r"\bcontent\s+strategy\b", r"\bdigital\s+marketing\b", r"\bseo\s+service",
|
||||
r"\bsocial\s+media\b", r"\blimited\s+time\s+offer\b", r"\bspecial\s+offer\b",
|
||||
r"\bsign\s+up\s+now\b", r"\br\d[\d,.]", r"\bstarting\s+at\b", r"\bper\s+month\b",
|
||||
r"\bmonthly\b", r"\bfreelancer\s+marketplace\b", r"\bfirst\s+order\b",
|
||||
r"\bdomain\s+name\b", r"\bregister\s+your\s+domain\b", r"\bsponsored\b",
|
||||
r"\bpromoted\b", r"\badvertisement\b",
|
||||
r"\bfree\s+domain\b",
|
||||
r"\bweb\s+hosting\b",
|
||||
r"\bfree\s+website\b",
|
||||
r"\bprofessional\s+website",
|
||||
r"\baffordable\s+web\b",
|
||||
r"\bget\s+a\s+free\b",
|
||||
r"\bsee\s+more\b",
|
||||
r"\bbook\s+your\b",
|
||||
r"\bhire\s+a\s+freelancer\b",
|
||||
r"\bcontent\s+strategy\b",
|
||||
r"\bdigital\s+marketing\b",
|
||||
r"\bseo\s+service",
|
||||
r"\bsocial\s+media\b",
|
||||
r"\blimited\s+time\s+offer\b",
|
||||
r"\bspecial\s+offer\b",
|
||||
r"\bsign\s+up\s+now\b",
|
||||
r"\br\d[\d,.]",
|
||||
r"\bstarting\s+at\b",
|
||||
r"\bper\s+month\b",
|
||||
r"\bmonthly\b",
|
||||
r"\bfreelancer\s+marketplace\b",
|
||||
r"\bfirst\s+order\b",
|
||||
r"\bdomain\s+name\b",
|
||||
r"\bregister\s+your\s+domain\b",
|
||||
r"\bsponsored\b",
|
||||
r"\bpromoted\b",
|
||||
r"\badvertisement\b",
|
||||
]
|
||||
|
||||
REQUEST_PATTERNS = [
|
||||
r"\bi\s+need\b", r"\bi\s+need\s+someone\b", r"\bi\s+need\s+a\b", r"\bi\s+need\s+an\b",
|
||||
r"\bi\s+want\b", r"\bi\s+would\s+like\b", r"\bi'm\s+looking\b", r"\bi\s+am\s+looking\b",
|
||||
r"\blooking\s+for\b", r"\blooking\s+to\b", r"\bwho\s+can\b", r"\bcan\s+someone\b",
|
||||
r"\bhelp\s+me\b", r"\bneed\s+help\b", r"\bwould\s+like\s+to\b", r"\bwant\s+to\b",
|
||||
r"\banyone\s+know\b", r"\bdoes\s+anyone\b", r"\brecommend\b", r"\bsuggestion\b",
|
||||
r"\bi\s+need\b",
|
||||
r"\bi\s+need\s+someone\b",
|
||||
r"\bi\s+need\s+a\b",
|
||||
r"\bi\s+need\s+an\b",
|
||||
r"\bi\s+want\b",
|
||||
r"\bi\s+would\s+like\b",
|
||||
r"\bi'm\s+looking\b",
|
||||
r"\bi\s+am\s+looking\b",
|
||||
r"\blooking\s+for\b",
|
||||
r"\blooking\s+to\b",
|
||||
r"\bwho\s+can\b",
|
||||
r"\bcan\s+someone\b",
|
||||
r"\bhelp\s+me\b",
|
||||
r"\bneed\s+help\b",
|
||||
r"\bwould\s+like\s+to\b",
|
||||
r"\bwant\s+to\b",
|
||||
r"\banyone\s+know\b",
|
||||
r"\bdoes\s+anyone\b",
|
||||
r"\brecommend\b",
|
||||
r"\bsuggestion\b",
|
||||
r"\bquote\s+(please|for|to)\b",
|
||||
]
|
||||
|
||||
def kw_match(text: str) -> bool:
|
||||
t = text.lower()
|
||||
return any(kw in t for kw in BROAD_KEYWORDS)
|
||||
|
||||
def is_request(text: str) -> bool:
|
||||
t = text.lower()
|
||||
return any(re.search(p, t) for p in REQUEST_PATTERNS)
|
||||
|
||||
def is_offer(text: str) -> bool:
|
||||
t = text.lower()
|
||||
return any(re.search(p, t) for p in OFFER_PATTERNS)
|
||||
|
||||
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",
|
||||
"need wordpress website", "I need a website for my business",
|
||||
"need a site for my business", "looking for website designer South Africa",
|
||||
"need website for my small business", "who can build me a website",
|
||||
"want to create a website for my business", "looking for affordable web design",
|
||||
"need a web developer South Africa", "help me build a website",
|
||||
"need someone to design my website", "looking for web designer near me",
|
||||
"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",
|
||||
"need wordpress website",
|
||||
"I need a website for my business",
|
||||
"need a site for my business",
|
||||
"looking for website designer South Africa",
|
||||
"need website for my small business",
|
||||
"who can build me a website",
|
||||
"want to create a website for my business",
|
||||
"looking for affordable web design",
|
||||
"need a web developer South Africa",
|
||||
"help me build a website",
|
||||
"need someone to design my website",
|
||||
"looking for web designer near me",
|
||||
"need a website for my startup",
|
||||
]
|
||||
|
||||
def is_offer(text: str) -> bool:
|
||||
return any(re.search(p, text.lower()) for p in OFFER_PATTERNS)
|
||||
|
||||
def is_request(text: str) -> bool:
|
||||
return any(re.search(p, text.lower()) for p in REQUEST_PATTERNS)
|
||||
|
||||
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:
|
||||
d = datetime.now() - timedelta(hours=int(m.group(1)))
|
||||
return d.strftime('%Y-%m-%d')
|
||||
if l.lower() == 'yesterday':
|
||||
return (datetime.now() - timedelta(days=1)).strftime('%Y-%m-%d')
|
||||
low = l.lower()
|
||||
if low in WEEKDAY_ORDER:
|
||||
day_idx = WEEKDAY_ORDER.index(low)
|
||||
today_idx = datetime.now().weekday()
|
||||
d = datetime.now() - timedelta(days=(today_idx - day_idx) % 7)
|
||||
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:
|
||||
return datetime.strptime(l, fmt).strftime('%Y-%m-%d')
|
||||
except ValueError:
|
||||
pass
|
||||
return datetime.now().strftime('%Y-%m-%d')
|
||||
|
||||
def _clean_fb_text(text: str) -> str:
|
||||
cleaned = []
|
||||
for l in text.split('\n'):
|
||||
s = l.strip()
|
||||
if len(s) < 3 or all(not c.isalpha() for c in s):
|
||||
continue
|
||||
cleaned.append(s)
|
||||
return '\n'.join(cleaned)
|
||||
|
||||
def _extract_posts_from_elements(elements: list[dict], base_url: str) -> list[dict]:
|
||||
posts = []
|
||||
seen = set()
|
||||
for el in elements:
|
||||
raw = el.get('text', '')
|
||||
if len(raw) < 40:
|
||||
continue
|
||||
text = _clean_fb_text(raw)
|
||||
if len(text) < 40 or is_offer(text):
|
||||
continue
|
||||
dekey = text[:80]
|
||||
if dekey in seen:
|
||||
continue
|
||||
seen.add(dekey)
|
||||
score = 2 if is_request(text) else 0
|
||||
posts.append({
|
||||
"title": text[:300], "content": text[:1000],
|
||||
"author": el.get('author', ''), "url": el.get('url') or base_url,
|
||||
"source": "facebook", "date": _parse_fb_date(text.split('\n')),
|
||||
"_score": score,
|
||||
})
|
||||
posts.sort(key=lambda p: p['_score'], reverse=True)
|
||||
for p in posts:
|
||||
p.pop('_score', None)
|
||||
return posts
|
||||
|
||||
def _extract_posts_from_text(raw: str, url: str) -> list[dict]:
|
||||
lines = [l.strip() for l in raw.split('\n') if l.strip() and len(l.strip()) >= 15]
|
||||
posts, seen = [], set()
|
||||
cur = []
|
||||
for l in lines:
|
||||
words = re.findall(r'[A-Za-z]{2,}', l)
|
||||
if len(words) < 2:
|
||||
continue
|
||||
lower = l.lower()
|
||||
if any(kw in lower for kw in BROAD_KEYWORDS) and not is_offer(l):
|
||||
snippet = ' '.join(cur[-3:] + [l]) if cur else l
|
||||
if len(snippet) >= 40 and not is_offer(snippet) and snippet[:80] not in seen:
|
||||
seen.add(snippet[:80])
|
||||
posts.append({"title": snippet[:300], "content": snippet[:1000],
|
||||
"author": '', "url": url, "source": "facebook",
|
||||
"date": _parse_fb_date(cur + [l] if cur else [l])})
|
||||
cur = []
|
||||
cur.append(l)
|
||||
if len(cur) > 10:
|
||||
cur.pop(0)
|
||||
return posts
|
||||
|
||||
async def _get_article_elements(page) -> list[dict]:
|
||||
return await page.evaluate('''() => {
|
||||
const results = [];
|
||||
const seenTexts = new Set();
|
||||
const selectors = ['div[role="article"]', 'div[role="feed"] > div', 'div.x1yztbdb', 'div[data-pagelet]'];
|
||||
for (const sel of selectors) {
|
||||
const els = document.querySelectorAll(sel);
|
||||
for (const el of els) {
|
||||
const text = (el.innerText || '').trim();
|
||||
if (text.length < 40) continue;
|
||||
const key = text.substring(0, 80);
|
||||
if (seenTexts.has(key)) continue;
|
||||
seenTexts.add(key);
|
||||
const links = el.querySelectorAll('a');
|
||||
let author = '', postUrl = '';
|
||||
for (const a of links) {
|
||||
const t = (a.innerText || '').trim();
|
||||
if (t && t.length > 2 && t.length < 80 && /^[a-zA-ZÀ-ÿ][a-zA-ZÀ-ÿ .'-]+$/.test(t)) {
|
||||
author = t; break;
|
||||
}
|
||||
}
|
||||
for (const a of links) {
|
||||
const h = (a.getAttribute('href') || '');
|
||||
if (h.includes('/posts/') || h.includes('/photo/') || h.includes('/video/') || h.includes('/groups/')) {
|
||||
postUrl = h.startsWith('http') ? h : 'https://www.facebook.com' + h; break;
|
||||
}
|
||||
}
|
||||
results.push({ text, author, url: postUrl });
|
||||
}
|
||||
if (results.length > 0) break;
|
||||
}
|
||||
return results;
|
||||
}''')
|
||||
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):
|
||||
@@ -217,109 +153,367 @@ async def get_fb_cookies(profile_path: str | None = None):
|
||||
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)
|
||||
rows = conn.cursor().execute(
|
||||
"SELECT name, value, host, path FROM moz_cookies WHERE host LIKE '%facebook.com'"
|
||||
).fetchall()
|
||||
c = conn.cursor()
|
||||
c.execute("SELECT name, value, host, path FROM moz_cookies WHERE host LIKE '%facebook.com'")
|
||||
rows = c.fetchall()
|
||||
conn.close()
|
||||
os.remove(tmp)
|
||||
return [{"name": n, "value": v, "domain": h if h.startswith('.') else f'.{h}',
|
||||
"path": p, "httpOnly": True, "secure": True, "sameSite": "Lax"}
|
||||
for n, v, h, p in rows]
|
||||
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 error: %s", e)
|
||||
try: os.remove(tmp)
|
||||
except: pass
|
||||
logger.error("Cookie DB read error: %s", e)
|
||||
try:
|
||||
os.remove(tmp)
|
||||
except Exception:
|
||||
pass
|
||||
return []
|
||||
|
||||
async def create_browser_session():
|
||||
if not BU_API_KEY:
|
||||
raise RuntimeError("BU_API_KEY not set")
|
||||
async with httpx.AsyncClient(timeout=30) as client:
|
||||
r = await client.post(
|
||||
f"{BU_API_BASE}/browsers",
|
||||
headers={"X-Browser-Use-API-Key": BU_API_KEY},
|
||||
json={"timeout": 60, "proxyCountryCode": "za"}
|
||||
)
|
||||
r.raise_for_status()
|
||||
data = r.json()
|
||||
return data["id"], data["cdpUrl"]
|
||||
WEEKDAY_ORDER = ['monday', 'tuesday', 'wednesday', 'thursday', 'friday', 'saturday', 'sunday']
|
||||
|
||||
async def stop_browser_session(session_id: str):
|
||||
if not BU_API_KEY:
|
||||
return
|
||||
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 _clean_fb_text(text: str) -> str:
|
||||
cleaned_lines = []
|
||||
for l in text.split('\n'):
|
||||
stripped = l.strip()
|
||||
if not stripped:
|
||||
continue
|
||||
if len(stripped) < 3:
|
||||
continue
|
||||
if all(not c.isalpha() for c in stripped):
|
||||
continue
|
||||
cleaned_lines.append(stripped)
|
||||
return '\n'.join(cleaned_lines)
|
||||
|
||||
def _extract_posts_from_elements(elements: list[dict], base_url: str) -> list[dict]:
|
||||
posts = []
|
||||
seen_texts = set()
|
||||
for el in elements:
|
||||
raw_text = el.get('text', '')
|
||||
if len(raw_text) < 40:
|
||||
continue
|
||||
text = _clean_fb_text(raw_text)
|
||||
if len(text) < 40:
|
||||
continue
|
||||
if is_offer(text):
|
||||
continue
|
||||
lines = text.split('\n')
|
||||
dekey = text[:80]
|
||||
if dekey in seen_texts:
|
||||
continue
|
||||
seen_texts.add(dekey)
|
||||
request_score = 2 if is_request(text) else 0
|
||||
post_url = el.get('url') or base_url
|
||||
posts.append({
|
||||
"title": text[:300],
|
||||
"content": text[:1000],
|
||||
"author": el.get('author', ''),
|
||||
"url": post_url,
|
||||
"source": "facebook",
|
||||
"date": _parse_fb_date(lines),
|
||||
"_score": request_score,
|
||||
})
|
||||
posts.sort(key=lambda p: p.get('_score', 0), reverse=True)
|
||||
for p in posts:
|
||||
p.pop('_score', None)
|
||||
return posts
|
||||
|
||||
def _extract_posts_from_text(raw: str, url: str) -> list[dict]:
|
||||
lines = [l.strip() for l in raw.split('\n') if l.strip() and len(l.strip()) >= 15]
|
||||
posts = []
|
||||
seen_texts = set()
|
||||
cur = []
|
||||
for l in lines:
|
||||
words = re.findall(r'[A-Za-z]{2,}', l)
|
||||
if len(words) < 2:
|
||||
continue
|
||||
lower = l.lower()
|
||||
if any(kw in lower for kw in BROAD_KEYWORDS) and not is_offer(l):
|
||||
if cur:
|
||||
snippet = ' '.join(cur[-3:] + [l])
|
||||
if len(snippet) >= 40 and not is_offer(snippet):
|
||||
dekey = snippet[:80]
|
||||
if dekey not in seen_texts:
|
||||
seen_texts.add(dekey)
|
||||
posts.append({
|
||||
"title": snippet[:300],
|
||||
"content": snippet[:1000],
|
||||
"author": '',
|
||||
"url": url,
|
||||
"source": "facebook",
|
||||
"date": _parse_fb_date(cur + [l]),
|
||||
})
|
||||
cur = []
|
||||
else:
|
||||
if not is_offer(l):
|
||||
snippet = l
|
||||
if len(snippet) >= 40:
|
||||
dekey = snippet[:80]
|
||||
if dekey not in seen_texts:
|
||||
seen_texts.add(dekey)
|
||||
posts.append({
|
||||
"title": snippet[:300],
|
||||
"content": snippet[:1000],
|
||||
"author": '',
|
||||
"url": url,
|
||||
"source": "facebook",
|
||||
"date": _parse_fb_date([l]),
|
||||
})
|
||||
cur.append(l)
|
||||
if len(cur) > 10:
|
||||
cur.pop(0)
|
||||
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 _get_article_elements(page) -> list[dict]:
|
||||
return await page.evaluate('''() => {
|
||||
const results = [];
|
||||
const seenTexts = new Set();
|
||||
const selectors = [
|
||||
'div[role="article"]',
|
||||
'div[role="feed"] > div',
|
||||
'div.x1yztbdb',
|
||||
'div[data-pagelet]',
|
||||
];
|
||||
for (const sel of selectors) {
|
||||
const els = document.querySelectorAll(sel);
|
||||
for (const el of els) {
|
||||
const text = (el.innerText || '').trim();
|
||||
if (text.length < 40) continue;
|
||||
const key = text.substring(0, 80);
|
||||
if (seenTexts.has(key)) continue;
|
||||
seenTexts.add(key);
|
||||
const links = el.querySelectorAll('a');
|
||||
let author = '';
|
||||
for (const a of links) {
|
||||
const t = (a.innerText || '').trim();
|
||||
if (t && t.length > 2 && t.length < 80 && /^[a-zA-ZÀ-ÿ][a-zA-ZÀ-ÿ .'-]+$/.test(t)) {
|
||||
author = t;
|
||||
break;
|
||||
}
|
||||
}
|
||||
let postUrl = '';
|
||||
for (const a of links) {
|
||||
const h = (a.getAttribute('href') || '');
|
||||
if (h.includes('/posts/') || h.includes('/photo/') || h.includes('/video/') || h.includes('/groups/')) {
|
||||
postUrl = h.startsWith('http') ? h : 'https://www.facebook.com' + h;
|
||||
break;
|
||||
}
|
||||
}
|
||||
results.push({ text, author, url: postUrl });
|
||||
}
|
||||
if (results.length > 0) break;
|
||||
}
|
||||
return results;
|
||||
}''')
|
||||
|
||||
async def _ensure_page(page, context):
|
||||
try:
|
||||
async with httpx.AsyncClient(timeout=10) as client:
|
||||
await client.patch(
|
||||
f"{BU_API_BASE}/browsers/{session_id}",
|
||||
headers={"X-Browser-Use-API-Key": BU_API_KEY, "Content-Type": "application/json"},
|
||||
json={"action": "stop"}
|
||||
)
|
||||
except Exception as e:
|
||||
logger.warning("Failed to stop BU session: %s", e)
|
||||
await page.evaluate('1')
|
||||
return page
|
||||
except Exception:
|
||||
logger.warning("Page was closed mid-scrape, creating a fresh page")
|
||||
page = await context.new_page()
|
||||
try:
|
||||
await page.goto('https://www.google.com/', wait_until='domcontentloaded', timeout=15000)
|
||||
await page.wait_for_timeout(random.randint(1000, 3000))
|
||||
except Exception:
|
||||
logger.warning("Google navigation failed during page recreation")
|
||||
await page.goto('https://www.facebook.com/', wait_until='domcontentloaded', timeout=30000)
|
||||
await page.wait_for_timeout(random.randint(3000, 8000))
|
||||
return page
|
||||
|
||||
async def search_facebook(page, query: str) -> list[dict]:
|
||||
async def search_facebook(page, context, query: str):
|
||||
page = await _ensure_page(page, context)
|
||||
url = f'https://www.facebook.com/search/posts/?q={urllib.parse.quote(query)}'
|
||||
try:
|
||||
await page.goto(url, wait_until='domcontentloaded', timeout=30000)
|
||||
try:
|
||||
await page.wait_for_selector('div[role="article"], div[role="feed"]', timeout=10000)
|
||||
await page.wait_for_selector('div[role="article"], div[role="feed"]', timeout=15000)
|
||||
except Exception:
|
||||
pass
|
||||
await page.wait_for_timeout(2000)
|
||||
await page.evaluate("window.scrollBy(0, 600)")
|
||||
await page.wait_for_timeout(1500)
|
||||
await page.wait_for_timeout(random.randint(3000, 8000))
|
||||
|
||||
await human_scroll(page, steps=random.randint(2, 4), total_delay=random.uniform(6, 15))
|
||||
|
||||
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_articles = await _get_article_elements(page)
|
||||
posts = _extract_posts_from_elements(raw_articles, url) if raw_articles else []
|
||||
if not posts:
|
||||
raw = await page.evaluate('document.body.innerText')
|
||||
posts = _extract_posts_from_text(raw, url)
|
||||
return posts
|
||||
except Exception as e:
|
||||
logger.warning("Facebook search '%s' failed: %s", query, e)
|
||||
return []
|
||||
return page, []
|
||||
return page, posts
|
||||
|
||||
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"}
|
||||
|
||||
session_id = None
|
||||
try:
|
||||
session_id, cdp_url = await create_browser_session()
|
||||
logger.info("BU session %s created, connecting via CDP", session_id)
|
||||
except Exception as e:
|
||||
logger.error("Failed to create BU session: %s", e)
|
||||
return {"success": False, "leads": [], "flagged": False, "flag_reason": None, "error": f"BU session failed: {e}"}
|
||||
|
||||
try:
|
||||
async with async_playwright() as pw:
|
||||
browser = await pw.chromium.connect_over_cdp(cdp_url)
|
||||
ctx = browser.contexts[0] if browser.contexts else await browser.new_context()
|
||||
if not browser.contexts:
|
||||
for c in fb_cookies:
|
||||
try: await ctx.add_cookies([c])
|
||||
except Exception: pass
|
||||
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) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/125.0.0.0 Safari/537.36',
|
||||
viewport=viewport,
|
||||
)
|
||||
|
||||
page = ctx.pages[0] if ctx.pages else await ctx.new_page()
|
||||
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
|
||||
try:
|
||||
await page.goto('https://www.google.com/', wait_until='domcontentloaded', timeout=15000)
|
||||
await page.wait_for_timeout(random.randint(1000, 3000))
|
||||
except Exception:
|
||||
logger.warning("Google navigation failed (IP may be blocked), trying Facebook directly")
|
||||
await page.goto('https://www.facebook.com/', wait_until='domcontentloaded', timeout=30000)
|
||||
await page.wait_for_timeout(2000)
|
||||
await page.wait_for_timeout(random.randint(3000, 8000))
|
||||
|
||||
url = page.url
|
||||
if '/login' in url.lower():
|
||||
logger.warning("Facebook login page detected")
|
||||
logger.warning("Facebook login page detected — account flagged")
|
||||
await browser.close()
|
||||
await stop_browser_session(session_id)
|
||||
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(5, 8))
|
||||
for query in searches:
|
||||
posts = await search_facebook(page, query)
|
||||
for i, query in enumerate(searches):
|
||||
page, posts = await search_facebook(page, context, query)
|
||||
all_posts.extend(posts)
|
||||
await page.wait_for_timeout(1000)
|
||||
if not posts:
|
||||
continue
|
||||
# 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()
|
||||
page = await _ensure_page(page, context)
|
||||
|
||||
# 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 = []
|
||||
@@ -333,17 +527,13 @@ async def scrape_facebook(profile_path: str | None = None, force: bool = False)
|
||||
if leads:
|
||||
leads = await classify_leads(leads)
|
||||
|
||||
await browser.close()
|
||||
await stop_browser_session(session_id)
|
||||
|
||||
return {"success": True, "leads": leads[:15], "flagged": False, "flag_reason": None, "error": None}
|
||||
|
||||
except Exception as e:
|
||||
logger.error("Facebook scrape failed: %s", e)
|
||||
await stop_browser_session(session_id)
|
||||
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,
|
||||
@@ -355,7 +545,8 @@ async def ask_ollama(prompt: str) -> str:
|
||||
"options": {"temperature": 0.05, "num_predict": 1024},
|
||||
})
|
||||
r.raise_for_status()
|
||||
return r.json()["message"]["content"]
|
||||
data = r.json()
|
||||
return data["message"]["content"]
|
||||
|
||||
async def classify_leads(results: list[dict]) -> list[dict]:
|
||||
if not results:
|
||||
@@ -377,7 +568,8 @@ For each numbered post, answer ONLY "yes" (LEAD) or "no" (NOT LEAD):
|
||||
Return a JSON array like ["yes","no","yes"] matching the order above."""
|
||||
ai_succeeded = False
|
||||
try:
|
||||
raw = (await ask_ollama(prompt)).strip()
|
||||
raw = await ask_ollama(prompt)
|
||||
raw = raw.strip()
|
||||
if raw.startswith("```"):
|
||||
first_nl = raw.find('\n')
|
||||
if first_nl != -1:
|
||||
@@ -388,24 +580,35 @@ Return a JSON array like ["yes","no","yes"] matching the order above."""
|
||||
answers = json.loads(raw)
|
||||
if isinstance(answers, list) and len(answers) == len(results):
|
||||
ai_succeeded = True
|
||||
filtered = [results[i] for i, ans in enumerate(answers) if isinstance(ans, str) and ans.lower() == 'yes']
|
||||
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 = ["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", "need someone", "hire a",
|
||||
"want someone", "need help with", "would like", "build me", "design my",
|
||||
"make me a", "create my"]
|
||||
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",
|
||||
"looking for", "need a", "need an", "looking to",
|
||||
"need someone", "hire a", "want someone",
|
||||
"need help with", "would like", "build me",
|
||||
"design my", "make me a", "create my",
|
||||
]
|
||||
filtered = []
|
||||
for r in results:
|
||||
t = r['title'].lower()
|
||||
if not any(kw in t for kw in strict):
|
||||
if not any(kw in t for kw in strict_keywords):
|
||||
continue
|
||||
if any(kw in t for kw in ['i build website', 'i offer web', 'i am a web developer',
|
||||
'affordable web design package', 'web hosting']):
|
||||
|
||||
Reference in New Issue
Block a user