Update app.py

app.py

@@ -6,8 +6,9 @@ import secrets
 import logging
 import numpy as np
 from contextlib import asynccontextmanager
+from datetime import datetime, timezone
 
-from fastapi import FastAPI, HTTPException, Security, Depends
+from fastapi import FastAPI, HTTPException, Security, Depends, Request
 from fastapi.security.api_key import APIKeyHeader
 from fastapi.middleware.cors import CORSMiddleware
 from pydantic import BaseModel
@@ -17,7 +18,11 @@ import onnxruntime as rt
 
 from utils.tokenizer_base import Tokenizer
 
-logging.basicConfig(level=logging.INFO)
+logging.basicConfig(
+    level=logging.INFO,
+    format="%(asctime)s UTC | %(levelname)s | %(message)s",
+    datefmt="%Y-%m-%d %H:%M:%S"
+)
 logger = logging.getLogger(__name__)
 
 # ── Config ────────────────────────────────────────────────────────────────────
@@ -30,9 +35,9 @@ ORT_INTRA_THREADS = int(os.getenv("ORT_INTRA_THREADS", "1"))
 ORT_INTER_THREADS = int(os.getenv("ORT_INTER_THREADS", "1"))
 
 # ── Torch global optimizations ────────────────────────────────────────────────
-torch.set_grad_enabled(False)       # no autograd overhead on tensor ops
-torch.set_num_threads(1)            # don't compete with ORT threads
-torch.set_num_interop_threads(1)    # no inter-op parallelism from torch
+torch.set_grad_enabled(False)
+torch.set_num_threads(1)
+torch.set_num_interop_threads(1)
 
 # ── API Key Auth ──────────────────────────────────────────────────────────────
 api_key_header = APIKeyHeader(name="X-API-Key", auto_error=False)
@@ -63,18 +68,21 @@ session = rt.InferenceSession(
 input_name = session.get_inputs()[0].name
 logger.info("✅ ONNX model ready")
 
+# ── Request counter (per worker) ──────────────────────────────────────────────
+request_count = 0
+
 # ── Inference ─────────────────────────────────────────────────────────────────
 def preprocess(image: Image.Image) -> np.ndarray:
     image = image.convert("RGB").resize(IMG_SIZE, Image.BILINEAR)
     x = np.ascontiguousarray(image, dtype=np.float32)
     x = (x / 255.0 - 0.5) / 0.5
-    return x.transpose(2, 0, 1)[np.newaxis, :]   # [1, 3, H, W]
+    return x.transpose(2, 0, 1)[np.newaxis, :]
 
 
 def solve_image(image: Image.Image) -> str:
-    x      = preprocess(image)
-    logits = session.run(None, {input_name: x})[0]
-    probs  = torch.tensor(logits).softmax(-1)
+    x        = preprocess(image)
+    logits   = session.run(None, {input_name: x})[0]
+    probs    = torch.tensor(logits).softmax(-1)
     preds, _ = tokenizer.decode(probs)
     return preds[0]
 
@@ -117,6 +125,28 @@ app.add_middleware(
 )
 
 
+# ── Access logging middleware ─────────────────────────────────────────────────
+@app.middleware("http")
+async def log_requests(request: Request, call_next):
+    global request_count
+    request_count += 1
+
+    timestamp = datetime.now(timezone.utc).strftime("%Y-%m-%d %H:%M:%S")
+    ip        = request.headers.get("X-Forwarded-For", request.client.host)
+    req_num   = request_count
+
+    start    = time.time()
+    response = await call_next(request)
+    elapsed  = time.time() - start
+
+    logger.info(
+        f"[#{req_num}] {timestamp} UTC | IP={ip} | "
+        f"{request.method} {request.url.path} | {elapsed:.3f}s"
+    )
+
+    return response
+
+
 # ── Endpoints ─────────────────────────────────────────────────────────────────
 @app.get("/health")
 def health():
@@ -127,6 +157,7 @@ def health():
         "quantized":     True,
         "workers":       os.getenv("WEB_CONCURRENCY", "1"),
         "intra_threads": ORT_INTRA_THREADS,
+        "requests_served": request_count,
     }