python-multipart + logging + /health endpoint

Dockerfile

@@ -2,10 +2,10 @@ FROM python:3.12-slim
 WORKDIR /app
 COPY . .
 
-# Normal packages + accelerate (fixes the device_map error)
-RUN pip install --no-cache-dir fastapi uvicorn transformers pillow accelerate
+# All required packages in one go
+RUN pip install --no-cache-dir fastapi uvicorn transformers pillow accelerate python-multipart
 
-# PyTorch CPU version (required for the model)
+# PyTorch CPU (required for the model)
 RUN pip install --no-cache-dir torch torchvision torchaudio --index-url https://download.pytorch.org/whl/cpu
 
 EXPOSE 7860

app.py

@@ -1,36 +1,74 @@
+import logging
 from fastapi import FastAPI, UploadFile, File, Form
-from pydantic import BaseModel
 from transformers import AutoModelForCausalLM, AutoProcessor
-import torch
 from PIL import Image
 import io
+import torch
+
+# === LOGGING SETUP ===
+logging.basicConfig(
+    level=logging.INFO,
+    format="%(asctime)s | %(levelname)s | %(message)s"
+)
+logger = logging.getLogger(__name__)
 
 app = FastAPI(title="Qwen3.5-0.8B OCR API")
 
+# === MODEL LOADING WITH LOGS ===
+logger.info("=== STARTING MODEL LOAD ===")
 model_id = "Qwen/Qwen3.5-0.8B"
-model = AutoModelForCausalLM.from_pretrained(model_id, torch_dtype="auto", device_map="cpu")
-processor = AutoProcessor.from_pretrained(model_id)
+try:
+    model = AutoModelForCausalLM.from_pretrained(
+        model_id,
+        torch_dtype="auto",
+        device_map="cpu"
+    )
+    processor = AutoProcessor.from_pretrained(model_id)
+    logger.info("✅ Model loaded successfully!")
+except Exception as e:
+    logger.error(f"❌ Model load failed: {e}")
+    raise
+
 
+# === HEALTH CHECK ===
+@app.get("/health")
+async def health():
+    logger.info("Health check requested")
+    return {
+        "status": "healthy",
+        "model": model_id,
+        "message": "Qwen3.5-0.8B OCR API is running"
+    }
 
+
+# === OCR ENDPOINT ===
 @app.post("/ocr")
 async def ocr(image: UploadFile = File(...), prompt: str = Form("Extract all text from this image accurately")):
-    img = Image.open(io.BytesIO(await image.read()))
+    logger.info(f"📸 OCR request received | Filename: {image.filename} | Prompt: {prompt[:100]}...")
+
+    try:
+        img = Image.open(io.BytesIO(await image.read()))
+
+        messages = [{"role": "user", "content": [
+            {"type": "image", "image": img},
+            {"type": "text", "text": prompt}
+        ]}]
 
-    messages = [{"role": "user", "content": [
-        {"type": "image", "image": img},
-        {"type": "text", "text": prompt}
-    ]}]
+        text = processor.apply_chat_template(messages, tokenize=False, add_generation_prompt=True)
+        inputs = processor(text, images=img, return_tensors="pt").to("cpu")
 
-    text = processor.apply_chat_template(messages, tokenize=False, add_generation_prompt=True)
-    inputs = processor(text, images=img, return_tensors="pt").to("cpu")
+        outputs = model.generate(**inputs, max_new_tokens=512, temperature=0.3, do_sample=False)
+        response = processor.decode(outputs[0], skip_special_tokens=True)
 
-    outputs = model.generate(**inputs, max_new_tokens=512, temperature=0.7)
-    response = processor.decode(outputs[0], skip_special_tokens=True)
+        logger.info("✅ OCR completed successfully")
+        return {"extracted_text": response}
 
-    return {"extracted_text": response}
+    except Exception as e:
+        logger.error(f"❌ OCR error: {e}")
+        return {"error": str(e)}
 
 
-# Simple test endpoint
+# === ROOT ===
 @app.get("/")
 async def root():
-    return {"message": "Qwen3.5-0.8B OCR API is running! Use POST /ocr with image file."}
+    return {"message": "Qwen3.5-0.8B OCR API ready! Try /health or POST /ocr"}

requirements.txt

@@ -3,4 +3,5 @@ uvicorn
 transformers
 pillow
 accelerate
+python-multipart
 torch --index-url https://download.pytorch.org/whl/cpu