File size: 9,677 Bytes
a5cd74c |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 |
from __future__ import annotations
import os, re, json, uuid, random, string, logging, asyncio
from datetime import datetime, timedelta
from typing import List, Callable, Any, Optional
import httpx
from fastapi import FastAPI, HTTPException
from fastapi.responses import StreamingResponse, JSONResponse
from pydantic import BaseModel, Field
# ββββββββββββββββββββββββββ logging ββββββββββββββββββββββββββββββββββββββ
logging.basicConfig(
level=os.getenv("LOG_LEVEL", "INFO"),
format="%(asctime)s | %(levelname)-8s | %(name)s | %(message)s",
)
log = logging.getLogger("snapzion-service")
log.info("snapzion service starting β¦")
# ββββββββββββββββββββββββββ ENV & constants βββββββββββββββββββββββββββββ
SYSTEM_PROMPT = os.getenv(
"SYSTEM_PROMPT",
"You are a prompt-safety model. Decide if the prompt is safe. "
"Respond with 'safe' or 'not safe'.",
)
SAFETY_API_KEY = os.getenv("SAFETY_API_KEY", "sk-F8l9ALDrJSpVCWJ3G1XbqP09oE3UD09Jf0t4WSlnrSJFdTtX")
SAFETY_MODEL_URL = os.getenv(
"SAFETY_MODEL_URL",
"https://api.typegpt.net/v1/chat/completions",
)
MAX_RETRIES = int(os.getenv("MAX_RETRIES", "5"))
INITIAL_DELAY = float(os.getenv("INITIAL_DELAY", "0.5"))
MAX_DELAY = float(os.getenv("MAX_DELAY", "2.5"))
# ββββββββββββββββββββββββββ FastAPI / HTTPX ββββββββββββββββββββββββββββ
app = FastAPI(title="Snapzion Image-Gen API | NAI", version="2.4.1")
_http: Optional[httpx.AsyncClient] = None
@app.on_event("startup")
async def _startup():
global _http
_http = httpx.AsyncClient(
timeout=30,
limits=httpx.Limits(max_connections=100, max_keepalive_connections=40),
)
log.info("HTTPX pool ready β")
# ββββββββββββββββββββββββββ Pydantic models ββββββββββββββββββββββββββββ
class ChatMessage(BaseModel):
role: str
content: str
class ChatRequest(BaseModel):
model: str
messages: List[ChatMessage]
stream: bool = Field(default=False)
# ββββββββββββββββββββββββββ Helpers ββββββββββββββββββββββββββββββββββββ
def _fake_user() -> tuple[str, str, str]:
first = random.choice("Alice Bob Carol David Evelyn Frank Grace Hector Ivy Jackie".split())
last = random.choice("Smith Johnson Davis Miller Thompson Garcia Brown Wilson Martin Clark".split())
email = ''.join(random.choices(string.ascii_lowercase + string.digits, k=8)) + "@example.com"
cust = "cus_" + ''.join(random.choices(string.ascii_letters + string.digits, k=14))
return f"{first} {last}", email, cust
async def _retry(fn: Callable, *a, **kw) -> Any:
max_tries = kw.pop("max_retries", MAX_RETRIES)
delay = INITIAL_DELAY
for n in range(1, max_tries + 1):
try:
return await fn(*a, **kw)
except httpx.HTTPStatusError as exc:
if exc.response.status_code == 400:
log.warning("%s try %d/%d: HTTP 400 error: %s", fn.__name__, n, max_tries, exc)
if n == max_tries:
log.error("%s failed after %d tries: HTTP 400 error: %s", fn.__name__, n, exc)
raise
else:
log.error("%s failed with status %d: %s", fn.__name__, exc.response.status_code, exc)
raise
except Exception as exc:
if n == max_tries:
log.error("%s failed after %d tries: %s", fn.__name__, n, exc)
raise
log.warning("%s try %d/%d: %s", fn.__name__, n, max_tries, exc)
await asyncio.sleep(delay + random.uniform(0, 0.4))
delay = min(delay * 2, MAX_DELAY)
# ββββββββββββββββββββββββββ Safety check βββββββββββββββββββββββββββββββ
async def _raw_safety(prompt: str) -> bool:
assert _http
hdrs = {"Authorization": f"Bearer {SAFETY_API_KEY}", "Content-Type": "application/json"}
payload = {
"model": "meta-llama/Meta-Llama-3-8B-Instruct-Lite",
"messages": [
{"role": "system", "content": SYSTEM_PROMPT},
{"role": "user", "content": prompt},
],
}
r = await _http.post(SAFETY_MODEL_URL, json=payload, headers=hdrs)
r.raise_for_status()
raw = r.json()["choices"][0]["message"]["content"].strip().lower()
log.debug("Safety raw reply: %r", raw)
if re.search(r"\b(not\s+safe|unsafe)\b", raw):
log.warning("Prompt-safety verdict: NOT SAFE")
return False
if re.search(r"\bsafe\b", raw):
log.info("Prompt-safety verdict: SAFE")
return True
log.warning("Prompt-safety unknown reply %r β NOT SAFE", raw)
return False
async def is_safe(prompt: str) -> bool:
return await _retry(_raw_safety, prompt)
# ββββββββββββββββββββββββββ Blackbox Image API βββββββββββββββββββββββββ
async def _raw_blackbox(prompt: str) -> str:
assert _http
name, email, _ = _fake_user()
user_id = ''.join(random.choices(string.digits, k=21))
expiry = (datetime.utcnow().replace(microsecond=0) + timedelta(days=30)).isoformat() + "Z"
payload = {
"query": prompt,
"session": {
"user": {
"name": name,
"email": email,
"image": "https://lh3.googleusercontent.com/a/ACg8ocI-ze5Qe42S-j8xaCL6X7KSVwfiOae4fONqpTxzt0d2_a2FIld1=s96-c",
"id": user_id
},
"expires": expiry
}
}
headers = {
"accept": "*/*",
"accept-language": "en-US,en;q=0.9,ru;q=0.8",
"content-type": "text/plain;charset=UTF-8",
"origin": "https://www.blackbox.ai",
"priority": "u=1, i",
"referer": "https://www.blackbox.ai/",
"sec-ch-ua": '"Google Chrome";v="135", "Not-A.Brand";v="8", "Chromium";v="135"',
"sec-ch-ua-mobile": "?0",
"sec-ch-ua-platform": '"Windows"',
"sec-fetch-dest": "empty",
"sec-fetch-mode": "cors",
"sec-fetch-site": "same-origin",
"user-agent": (
"Mozilla/5.0 (Windows NT 10.0; Win64; x64) "
"AppleWebKit/537.36 (KHTML, like Gecko) "
"Chrome/135.0.0.0 Safari/537.36"
),
}
resp = await _http.post("https://www.blackbox.ai/api/image-generator", json=payload, headers=headers)
resp.raise_for_status()
try:
return resp.json().get("markdown", "").strip()
except json.JSONDecodeError:
return resp.text.strip()
async def blackbox(prompt: str) -> str:
return await _retry(_raw_blackbox, prompt)
# ββββββββββββββββββββββββββ Main route βββββββββββββββββββββββββββββββββ
@app.post("/v1/chat/completions")
async def chat(req: ChatRequest):
if _http is None:
raise HTTPException(503, "HTTP client not ready")
user_prompt = next((m.content for m in reversed(req.messages) if m.role == "user"), "")
if not user_prompt:
raise HTTPException(400, "User prompt missing")
try:
if not await is_safe(user_prompt):
return JSONResponse({"error": "Your prompt is considered unsafe."}, status_code=400)
except httpx.HTTPStatusError as exc:
return JSONResponse({"error": f"Safety check failed: HTTP {exc.response.status_code}", "reason": str(exc)}, status_code=503)
try:
md = await blackbox(user_prompt)
except httpx.HTTPStatusError as exc:
return JSONResponse({"error": f"Image generation failed: HTTP {exc.response.status_code}", "reason": str(exc)}, status_code=503)
except Exception as exc:
return JSONResponse({"error": "Image generation failed after retries.", "reason": str(exc)}, status_code=503)
md = re.sub(r"!\[[^\]]*\]\(https://storage\.googleapis\.com([^\)]*)\)",
f"", md)
uid, ts = str(uuid.uuid4()), int(datetime.now().timestamp())
if not req.stream:
return {
"id": uid,
"object": "chat.completion",
"created": ts,
"model": "Image-Generator",
"choices": [{
"index": 0,
"message": {"role": "assistant", "content": md},
"finish_reason": "stop",
}],
"usage": None,
}
async def sse():
chunk1 = {"id": uid, "object": "chat.completion.chunk", "created": ts, "model": "Image-Generator",
"choices": [{"index": 0, "delta": {"role": "assistant", "content": md}, "finish_reason": None}], "usage": None}
yield f"data: {json.dumps(chunk1)}\n\n"
chunk2 = {"id": uid, "object": "chat.completion.chunk", "created": ts, "model": "Image-Generator",
"choices": [{"index": 0, "delta": {"role": "assistant", "content": ""}, "finish_reason": "stop"}], "usage": None}
yield f"data: {json.dumps(chunk2)}\n\n"
yield "data: [DONE]\n\n"
return StreamingResponse(sse(), media_type="text/event-stream")
|