File size: 9,331 Bytes
5f920e0 4f95612 5f920e0 4f95612 | 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 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 | import collections
import logging
import os
import sys
import threading
import time
from contextlib import asynccontextmanager
from pathlib import Path
from dotenv import load_dotenv
from fastapi import Depends, FastAPI, Header, HTTPException
from fastapi.middleware.cors import CORSMiddleware
from fastapi.responses import FileResponse, JSONResponse
load_dotenv()
logging.basicConfig(
level=logging.INFO,
format="%(asctime)s | %(levelname)-8s | %(message)s",
datefmt="%H:%M:%S",
handlers=[logging.StreamHandler(sys.stdout)],
)
log = logging.getLogger(__name__)
# ---------------------------------------------------------------------------
# Configuration
# ---------------------------------------------------------------------------
PORT = int(os.environ.get("PORT", 8000))
API_KEY = os.environ.get("API_KEY", "")
MODEL_VERSION = os.environ.get("MODEL_VERSION", "v1.0.0")
HF_TOKEN = os.environ.get("HF_TOKEN", "")
DEVICE = os.environ.get("DEVICE", "cpu")
# Rate limiting — spec Section 5 / error code 429
RATE_LIMIT_REQUESTS = int(os.environ.get("RATE_LIMIT_REQUESTS", 60))
RATE_LIMIT_WINDOW_S = int(os.environ.get("RATE_LIMIT_WINDOW_S", 60))
_rate_buckets: dict = {}
_rate_lock: threading.Lock = threading.Lock()
ONNX_PATH = Path("models/onnx/immuniwatch_classifier.onnx")
THRESHOLDS_PATH = Path("models/onnx/thresholds.json")
CONFIG_PATH = Path("models/onnx/model_config.json")
LORA_REPO = "AHFIDAILabs/immuniwatch-lora-classifier"
# Uptime tracking — used by /health
_start_time = time.time()
# ---------------------------------------------------------------------------
# Model file download — runs at startup on HuggingFace Spaces
# Downloads ONNX files from the model repo if not present locally.
# On local dev they already exist in models/onnx/ (gitignored).
# ---------------------------------------------------------------------------
def _download_model_files() -> None:
files = [
("immuniwatch_classifier.onnx", ONNX_PATH),
("immuniwatch_classifier.onnx.data", ONNX_PATH.parent / "immuniwatch_classifier.onnx.data"),
("thresholds.json", THRESHOLDS_PATH),
("model_config.json", CONFIG_PATH),
]
missing = [(fname, path) for fname, path in files if not path.exists()]
if not missing:
return
log.info("Downloading %d model file(s) from %s ...", len(missing), LORA_REPO)
try:
from huggingface_hub import hf_hub_download
for fname, path in missing:
log.info(" -> %s", fname)
path.parent.mkdir(parents=True, exist_ok=True)
hf_hub_download(
repo_id= LORA_REPO,
filename= fname,
local_dir= str(path.parent),
token= HF_TOKEN or None,
)
log.info("Model files ready.")
except Exception as exc:
log.error("Model download failed: %s", exc)
log.error("Upload ONNX files to %s on HuggingFace Hub first.", LORA_REPO)
# ---------------------------------------------------------------------------
# Startup
# ---------------------------------------------------------------------------
def _start_ingestion_worker() -> None:
bluesky_handle = os.environ.get("BLUESKY_HANDLE", "")
youtube_key = os.environ.get("YOUTUBE_API_KEY", "")
if not bluesky_handle and not youtube_key:
log.info("No connector credentials — ingestion worker not started.")
return
try:
from src.ingestion.direct_runner import run as run_ingestion
t = threading.Thread(target=run_ingestion, daemon=True, name="ingestion-worker")
t.start()
log.info("Ingestion worker started in background thread.")
except Exception as exc:
log.warning("Ingestion worker failed to start: %s", exc)
@asynccontextmanager
async def lifespan(app: FastAPI):
from src.models.classifier import load as load_classifier
from src.intelligence.rag import preload_embedder
log.info("Starting ImmuniWatch ML Service v%s", MODEL_VERSION)
_download_model_files()
load_classifier(
onnx_path= str(ONNX_PATH),
thresholds_path=str(THRESHOLDS_PATH),
config_path= str(CONFIG_PATH),
tokenizer_repo= LORA_REPO,
hf_token= HF_TOKEN or None,
)
preload_embedder()
_start_ingestion_worker()
from src.api.counter_narrative_store import init_db
init_db()
log.info("Service ready on port %d", PORT)
yield
log.info("Service shutting down.")
# ---------------------------------------------------------------------------
# App
# ---------------------------------------------------------------------------
app = FastAPI(
title="ImmuniWatch Nigeria — ML Service",
version=MODEL_VERSION,
lifespan=lifespan,
docs_url="/docs",
redoc_url=None,
)
# ---------------------------------------------------------------------------
# CORS — allow any origin so the local dashboard (file://) can call the API
# ---------------------------------------------------------------------------
app.add_middleware(
CORSMiddleware,
allow_origins=["*"],
allow_credentials=False,
allow_methods=["*"],
allow_headers=["*"],
)
# ---------------------------------------------------------------------------
# Authentication — per spec Section 4
# Never log the key value — only log presence or absence
# ---------------------------------------------------------------------------
def _check_rate_limit(key: str) -> tuple:
now = time.time()
with _rate_lock:
if key not in _rate_buckets:
_rate_buckets[key] = collections.deque()
bucket = _rate_buckets[key]
while bucket and bucket[0] < now - RATE_LIMIT_WINDOW_S:
bucket.popleft()
if len(bucket) >= RATE_LIMIT_REQUESTS:
retry_after = int(RATE_LIMIT_WINDOW_S - (now - bucket[0])) + 1
return False, retry_after
bucket.append(now)
return True, 0
async def require_api_key(x_ml_api_key: str = Header(default=None)):
if x_ml_api_key is None:
log.warning("Rejected — X-ML-API-Key header absent")
raise HTTPException(status_code=401,
detail="X-ML-API-Key header is required")
if x_ml_api_key != API_KEY:
log.warning("Rejected — X-ML-API-Key invalid")
raise HTTPException(status_code=401, detail="Invalid API key")
allowed, retry_after = _check_rate_limit(x_ml_api_key)
if not allowed:
raise HTTPException(
status_code=429,
detail="Rate limit exceeded",
headers={"Retry-After": str(retry_after)},
)
# ---------------------------------------------------------------------------
# GET / — root info endpoint, no auth required
# ---------------------------------------------------------------------------
@app.get("/")
async def root():
return {
"service": "ImmuniWatch Nigeria ML Service",
"version": MODEL_VERSION,
"status": "running",
"docs": "/docs",
"health": "/health",
"dashboard": "/dashboard",
"classify": "POST /classify",
"batch": "POST /classify/batch",
}
# ---------------------------------------------------------------------------
# GET /dashboard — serve the HTML dashboard, no auth required
# ---------------------------------------------------------------------------
@app.get("/dashboard", include_in_schema=False)
async def dashboard():
path = Path("dashboard.html")
if not path.exists():
raise HTTPException(status_code=404, detail="Dashboard not found")
return FileResponse(path, media_type="text/html")
# ---------------------------------------------------------------------------
# GET /health — no auth required, must respond in < 10ms
# ---------------------------------------------------------------------------
@app.get("/health")
async def health():
from src.models.classifier import is_loaded
if not is_loaded():
return JSONResponse(
status_code=503,
content={"status": "unavailable", "reason": "model loading"},
)
return {
"status": "ok",
"model_loaded": True,
"model_version": MODEL_VERSION,
"device": DEVICE,
"uptime_s": int(time.time() - _start_time),
}
# ---------------------------------------------------------------------------
# Register all other routes with API key authentication
# ---------------------------------------------------------------------------
from src.api.routes import router # noqa: E402
from src.api.counter_narrative_routes import cn_router # noqa: E402
from src.api.kb_routes import kb_router # noqa: E402
app.include_router(router, dependencies=[Depends(require_api_key)])
app.include_router(cn_router, dependencies=[Depends(require_api_key)])
app.include_router(kb_router, dependencies=[Depends(require_api_key)]) |