Update app.py

app.py

@@ -1,37 +1,73 @@
-# app.py — use Qwen2-VL via together.ai or openrouter (free tier)
+import torch
 import gradio as gr
-import base64
-from io import BytesIO
+from transformers import AutoProcessor, AutoModelForVision2Seq
 from PIL import Image
-from openai import OpenAI
+import os
 
-client = OpenAI(
-    api_key="your_free_api_key",  # openrouter.ai free tier
-    base_url="https://openrouter.ai/api/v1"
+os.environ["HF_HUB_ENABLE_HF_TRANSFER"] = "1"
+
+model_id = "HuggingFaceTB/SmolVLM-256M-Instruct"  # smallest possible - 256M params
+
+print("Loading processor...")
+processor = AutoProcessor.from_pretrained(model_id)
+
+print("Loading model...")
+model = AutoModelForVision2Seq.from_pretrained(
+    model_id,
+    torch_dtype=torch.float32,
+    device_map="auto"
 )
+model.eval()
+print("Model ready!")
+
 
 def extract_text(image: Image.Image) -> str:
-    buffered = BytesIO()
-    image.save(buffered, format="PNG")
-    img_b64 = base64.b64encode(buffered.getvalue()).decode()
+    if image is None:
+        return "Please upload an image."
 
-    response = client.chat.completions.create(
-        model="qwen/qwen2-vl-7b-instruct:free",  # free model
-        messages=[{
+    messages = [
+        {
             "role": "user",
             "content": [
-                {"type": "image_url", "image_url": {"url": f"data:image/png;base64,{img_b64}"}},
-                {"type": "text", "text": "Extract all text exactly as it appears. Preserve question numbers, options A B C D, tables, and equations."}
+                {"type": "image"},
+                {"type": "text", "text": (
+                    "Extract all text from this image exactly as it appears. "
+                    "Preserve question numbers, options A B C D, "
+                    "tables, and any mathematical or chemical expressions. "
+                    "Format clearly."
+                )}
             ]
-        }]
+        }
+    ]
+
+    prompt = processor.apply_chat_template(
+        messages,
+        add_generation_prompt=True
     )
-    return response.choices[0].message.content
+
+    inputs = processor(
+        text=prompt,
+        images=[image],
+        return_tensors="pt"
+    ).to(model.device)
+
+    with torch.no_grad():
+        outputs = model.generate(
+            **inputs,
+            max_new_tokens=1024,
+            do_sample=False
+        )
+
+    generated = outputs[0][inputs["input_ids"].shape[1]:]
+    return processor.decode(generated, skip_special_tokens=True)
+
 
 demo = gr.Interface(
     fn=extract_text,
     inputs=gr.Image(type="pil", label="Upload NEET Question Image"),
     outputs=gr.Textbox(label="Extracted Text", lines=20),
-    title="NEET Question Extractor"
+    title="NEET Question Extractor",
+    description="Upload a scanned NEET question paper image to extract text"
 )
 
 demo.launch()