app.py

import os
import torch
from transformers import AutoProcessor, AutoModelForCausalLM
from PIL import Image
from groq import Groq

def enhance_medical_text(text):
    try:
        client = Groq(
            api_key=os.getenv("GROQ_API_KEY")  # Get API key from environment variable
        )

        chat_completion = client.chat.completions.create(
            messages=[
                {
                    "role": "system",
                    "content": "You are a medical prescription expert. Correct OCR errors in medicine names, dosages and medical terms..."
                },
                {
                    "role": "user",
                    "content": f"Correct this medical prescription OCR output:\n{text}"
                }
            ],
            model="llama3-8b-8192",
            temperature=0.1,
            max_tokens=1024
        )
        return chat_completion.choices[0].message.content
    except Exception as e:
        print(f"Groq enhancement error: {str(e)}")
        return text

# Get Hugging Face token securely from environment variables
HF_TOKEN = os.getenv("HF_TOKEN")

device = "cuda:0" if torch.cuda.is_available() else "cpu"
torch_dtype = torch.float16 if torch.cuda.is_available() else torch.float32

# Load model and processor with token
model = AutoModelForCausalLM.from_pretrained(
    "microsoft/Florence-2-large",
    token=HF_TOKEN,
    torch_dtype=torch_dtype,
    trust_remote_code=True
).to(device)

processor = AutoProcessor.from_pretrained(
    "microsoft/Florence-2-large",
    token=HF_TOKEN,
    trust_remote_code=True
)

def run_ocr(image, task_prompt="<OCR>"):
    inputs = processor(text=task_prompt, images=image, return_tensors="pt").to(device, torch_dtype)
    generated_ids = model.generate(
        input_ids=inputs["input_ids"],
        pixel_values=inputs["pixel_values"],
        max_new_tokens=1024,
        num_beams=3,
        do_sample=False
    )
    return processor.batch_decode(generated_ids, skip_special_tokens=True)[0]

import gradio as gr

def process_single_image(image):
    image = Image.fromarray(image)
    result = run_ocr(image, "<OCR>")
    corrected_text = enhance_medical_text(result)
    return result, corrected_text

if __name__ == "__main__":
    demo = gr.Interface(
        fn=process_single_image,
        inputs=gr.Image(label="Upload Prescription"),
        outputs=[
            gr.Textbox(label="Raw OCR Output"),
            gr.Textbox(label="Enhanced Medical Report")
        ],
        title="Medical Prescription OCR",
        description="Upload a medical prescription image for OCR processing and enhancement"
    )
    demo.launch(server_name="0.0.0.0", server_port=7860)