app.py

import gradio as gr
import spaces
from transformers import AutoModelForCausalLM, AutoTokenizer

# Define the medical assistant system prompt
SYSTEM_PROMPT = """
You are a knowledgeable medical assistant. Follow these steps in order:

1) INITIAL ASSESSMENT: First, warmly greet the user and ask about their primary concern.

2) ASK FOLLOW-UP QUESTIONS: For any health concern mentioned, systematically gather information by asking 1-2 specific follow-up questions at a time about:
   - Detailed description of symptoms
   - Duration (when did it start?)
   - Severity (scale of 1-10)
   - Aggravating or alleviating factors
   - Related symptoms
   - Medical history
   - Current medications and allergies
   - Family history of similar conditions

3) SUMMARIZE FINDINGS: Once you have gathered sufficient information (at least 4-5 exchanges with the user), organize what you've learned into clear categories:
   - Symptoms
   - Duration
   - Severity
   - Possible Causes
   - Medications/Allergies
   - Family History

4) PROVIDE RECOMMENDATIONS: Only after gathering comprehensive information, suggest:
   - One specific OTC medicine with proper adult dosing
   - One practical home remedy
   - When they should seek professional medical care

5) END WITH DISCLAIMER: Always end with a clear medical disclaimer that you are not a licensed medical professional and your suggestions are not a substitute for professional medical advice.

IMPORTANT: Do not skip ahead to recommendations without gathering comprehensive information through multiple exchanges. Your primary goal is information gathering through thoughtful questions.
"""

# Define model options
MODELS = {
    "TinyLlama-1.1B": "TinyLlama/TinyLlama-1.1B-Chat-v1.0",
    "Llama-2-7b": "meta-llama/Llama-2-7b-chat-hf"
}

# Global variables to store loaded models and tokenizers
loaded_models = {}
loaded_tokenizers = {}

def load_model(model_name):
    """Load model and tokenizer if not already loaded"""
    if model_name not in loaded_models:
        print(f"Loading {model_name}...")
        model_path = MODELS[model_name]
        tokenizer = AutoTokenizer.from_pretrained(model_path)
        model = AutoModelForCausalLM.from_pretrained(
            model_path,
            torch_dtype="auto",
            device_map="auto"  # Use GPU if available
        )
        loaded_models[model_name] = model
        loaded_tokenizers[model_name] = tokenizer
        print(f"{model_name} loaded successfully!")
    return loaded_models[model_name], loaded_tokenizers[model_name]

# Pre-load the smaller model to start with
print("Pre-loading TinyLlama model...")
load_model("TinyLlama-1.1B")

@spaces.GPU  # Required by ZeroGPU!
def generate_response(message, history, model_choice):
    """Generate a response from the selected model"""
    # Load the selected model if not already loaded
    model, tokenizer = load_model(model_choice)
    
    # Format the prompt based on the history and system prompt
    formatted_prompt = SYSTEM_PROMPT + "\n\n"
    
    # Add conversation history
    for human, assistant in history:
        formatted_prompt += f"User: {human}\nAssistant: {assistant}\n"
    
    # Add the current message
    formatted_prompt += f"User: {message}\nAssistant:"
    
    # Generate the response
    inputs = tokenizer(formatted_prompt, return_tensors="pt").to(model.device)
    outputs = model.generate(
        inputs["input_ids"],
        max_new_tokens=512,
        temperature=0.7,
        top_p=0.9,
        do_sample=True,
    )
    response = tokenizer.decode(outputs[0][inputs["input_ids"].shape[1]:], skip_special_tokens=True)
    
    return response.strip()

# Create the Gradio interface
with gr.Blocks() as demo:
    gr.Markdown("# Medical Assistant Chatbot")
    gr.Markdown("This chatbot uses LLM models to provide medical information and assistance. Please note that this is not a substitute for professional medical advice.")
    
    with gr.Row():
        model_dropdown = gr.Dropdown(
            choices=list(MODELS.keys()),
            value="TinyLlama-1.1B",
            label="Select Model"
        )
    
    chatbot = gr.ChatInterface(
        fn=lambda message, history, model_choice: generate_response(message, history, model_choice),
        additional_inputs=[model_dropdown],
    )

if __name__ == "__main__":
    demo.launch()