drjones/android-fleet-godhead
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

proxyfly engine v2: 3-source harvest, stratified testing, 2-strike health, transparent incl, latency cap

Browse Source
This commit is contained in:
drjones
2026-08-13 17:15:43 -07:00
parent 67c8579db9
commit 38087a74c1
2 changed files with 106 additions and 41 deletions
Download Patch File Download Diff File Expand all files Collapse all files

118
proxy_harvest.py
View File

@@ -15,13 +15,16 @@ from concurrent.futures import ThreadPoolExecutor
DB_PATH = "/opt/android-fleet/fleet.db"
SOURCES = [
"https://cdn.jsdelivr.net/gh/proxifly/free-proxy-list@main/proxies/all/data.json",
("proxifly", "https://cdn.jsdelivr.net/gh/proxifly/free-proxy-list@main/proxies/all/data.json"),
("geonode", "https://proxylist.geonode.com/api/proxy-list?limit=500&page=1&sort_by=lastChecked&sort_type=desc"),
("proxyscrape", "https://api.proxyscrape.com/v4/free-proxy-list/get?request=display_proxies&proxy_format=protocolipport&format=json"),
]
MAX_FETCH = 200 # candidates per cycle
MAX_TEST = 120 # live-test at most this many per cycle
MAX_POOL = 40 # cap the proxies table (perf)
MAX_POOL = 60 # cap the proxies table (perf)
TEST_TIMEOUT = 7
HARVEST_INTERVAL = 1500 # 25 min
HARVEST_INTERVAL = 480 # 8 min — free proxies churn fast, harvest often
MAX_LATENCY_MS = 4000 # drop anything slower than this (routeable only)
def log(msg):
print(f"[harvest] {msg}", flush=True)
@@ -32,48 +35,87 @@ def dbcon():
return con
def fetch_candidates():
cands = []
for src in SOURCES:
cands = {"socks5": [], "socks4": [], "http": []}
for src_name, src in SOURCES:
try:
with urllib.request.urlopen(src, timeout=45) as r:
req = urllib.request.Request(src, headers={
"User-Agent": "Mozilla/5.0 (Windows NT 10.0; Win64; x64) AppleWebKit/537.36 Chrome/120 Safari/537.36",
"Accept": "application/json"})
with urllib.request.urlopen(req, timeout=45) as r:
data = json.loads(r.read())
for p in data:
proto = p.get("protocol", "")
if proto not in ("socks5", "https", "http", "socks4"):
continue
if p.get("anonymity") == "transparent":
continue # leaks real IP — useless for the fleet
cands.append({
"host": p["ip"], "port": p["port"], "proto": proto,
"country": (p.get("geolocation") or {}).get("country", "?"),
})
if src_name == "proxifly":
for p in data:
proto = p.get("protocol", "")
if proto not in ("socks5", "https", "http", "socks4"):
continue
# transparent = leaks real IP only on plain-HTTP; our traffic is
# HTTPS, so include them (mark them in the name later).
cands.setdefault(proto, []).append({
"host": p["ip"], "port": p["port"], "proto": proto,
"country": (p.get("geolocation") or {}).get("country", "?"),
"anon": p.get("anonymity", ""),
})
elif src_name == "geonode":
for p in data.get("data", []):
protos = [x for x in (p.get("protocols") or []) if x in ("socks5", "socks4", "http", "https")]
if not protos:
continue
for proto in protos[:1]:
cands.setdefault(proto, []).append({
"host": p["ip"], "port": int(p["port"]), "proto": proto,
"country": p.get("country", "?"),
"anon": p.get("anonymityLevel", ""),
})
elif src_name == "proxyscrape":
for p in data.get("proxies", []):
proto = p.get("protocol", "")
if proto not in ("socks5", "socks4", "http"):
continue
cands.setdefault(proto, []).append({
"host": p["ip"], "port": int(p["port"]), "proto": proto,
"country": p.get("country", "?"),
"anon": "elite",
})
except Exception as e:
log(f"source error {src}: {e}")
log(f"source error {src_name}: {e}")
# STRATIFY + SHUFFLE: feeds are protocol-grouped; sampling the head would
# test only http proxies. Take an even mix across protocols.
import random as _random
out = []
for pool in cands.values():
_random.shuffle(pool)
while len(out) < MAX_FETCH and any(cands.values()):
for proto in ("socks5", "socks4", "http"):
if cands[proto]:
out.append(cands[proto].pop())
# dedupe by host:port
seen = set()
out = []
for c in cands:
final = []
for c in out:
k = f"{c['host']}:{c['port']}"
if k not in seen:
seen.add(k)
out.append(c)
return out[:MAX_FETCH]
final.append(c)
return final[:MAX_FETCH]
def test_proxy(c):
"""Returns (c, egress_ip, latency_ms) or (c, None, None)."""
"""Returns (c, egress_ip, latency_ms) or (c, None, None). HTTPS first, HTTP fallback."""
proto_map = {"socks5": "socks5h://", "https": "https://", "http": "http://",
"socks4": "socks4://"}
url = proto_map[c["proto"]] + f"{c['host']}:{c['port']}"
import urllib.request as ur
opener = ur.build_opener(ur.ProxyHandler({"http": url, "https": url}))
t0 = time.time()
try:
with opener.open("https://api.ipify.org", timeout=TEST_TIMEOUT) as r:
ip = r.read().decode().strip()
lat = int((time.time() - t0) * 1000)
return (c, ip, lat)
except Exception:
return (c, None, None)
opener.addheaders = [("User-Agent", "Mozilla/5.0 (Linux; Android 9) AppleWebKit/537.36")]
for target in ("https://api.ipify.org", "http://api.ipify.org"):
t0 = time.time()
try:
with opener.open(target, timeout=TEST_TIMEOUT) as r:
ip = r.read().decode().strip()
lat = int((time.time() - t0) * 1000)
return (c, ip, lat)
except Exception:
continue
return (c, None, None)
def check_clean(ips):
"""ip-api batch: flags for each ip. Returns {ip: dict}."""
@@ -109,6 +151,8 @@ def dnsbl_listed(ip):
def prune_pool():
con = dbcon()
# drop dead proxifly entries every cycle (they die fast — keep the list honest)
con.execute("DELETE FROM proxies WHERE source='proxifly' AND health='dead' AND assigned_mac IS NULL")
rows = con.execute("SELECT id, egress_ip, source FROM proxies ORDER BY id").fetchall()
# remove duplicates by egress ip (keep lowest id)
seen = set()
@@ -151,12 +195,11 @@ def harvest():
dns = dns_flags.get(c["egress_ip"], False)
clean = 1 if (not f.get("hosting") and not f.get("proxy") and not dns) else 0
mobile = 1 if f.get("mobile") else 0
# keep only clean ones + a few fallbacks, skip dirty datacenter spam
if clean == 0:
why = "hosting" if f.get("hosting") else ("proxy" if f.get("proxy") else ("dnsbl" if dns else "unknown"))
log(f"rejected {c['egress_ip']} ({why})")
continue
name = f"proxifly-{c['country']}-{c['egress_ip'].split('.')[-2]}"
if c["latency_ms"] and c["latency_ms"] > MAX_LATENCY_MS:
continue # too slow to route through
# BANK EVERYTHING ALIVE: grade it, route clean-first. The list stays populated.
suffix = "-T" if c.get("anon") == "transparent" else ""
name = f"proxifly-{c['country']}-{c['egress_ip'].split('.')[-2]}{suffix}"
con.execute("""INSERT INTO proxies(name, host, port, proto, health, latency_ms,
assigned_mac, last_check, clean, country, egress_ip, source)
VALUES(?,?,?,?,'healthy',?,NULL,?,?,?,?,?)
@@ -182,13 +225,14 @@ def assign_best():
have = con.execute("SELECT * FROM proxies WHERE assigned_mac=?", (d["mac"],)).fetchone()
if have:
continue # already assigned — keep stable
# assignment preference: verified-clean first, then any live proxy
best = con.execute("""SELECT * FROM proxies
WHERE clean=1 AND health='healthy' AND assigned_mac IS NULL
ORDER BY latency_ms ASC LIMIT 1""").fetchone()
if not best:
best = con.execute("""SELECT * FROM proxies
WHERE health='healthy' AND assigned_mac IS NULL
ORDER BY latency_ms ASC LIMIT 1""").fetchone()
ORDER BY clean DESC, latency_ms ASC LIMIT 1""").fetchone()
if not best:
continue
con.execute("UPDATE proxies SET assigned_mac=? WHERE id=?", (d["mac"], best["id"]))