app.py

import gradio as gr
from transformers import pipeline

# Initialize the Hugging Face pipeline with a more advanced model
# Replace "EleutherAI/gpt-neo-2.7B" with other models like "mosaicml/mpt-7b-chat" or "OpenAssistant/oasst-sft-4-pythia-12b-epoch-3.5"
generation_pipeline = pipeline(
    "text-generation", 
    model="EleutherAI/gpt-neo-2.7B",  
    device=-1  # Use CPU explicitly
)

def dental_chatbot_response(message, history):
    """
    Responds to user queries with a focus on dental terminology.
    - Dynamically generates responses using an advanced LLM.
    - Designed to address dental-related questions or provide general responses.
    """
    print(f"User Input: {message}")
    print(f"Chat History: {history}")

    # Add a prompt to guide the LLM's focus on dental terminology
    prompt = (
        f"You are a highly knowledgeable and friendly dental expert chatbot. "
        f"Provide detailed and accurate explanations of dental terms, procedures, and treatments. "
        f"If the query is not dental-related, respond helpfully and informatively.\n\n"
        f"User: {message}\n\n"
        f"Chatbot:"
    )

    # Generate a response using the LLM
    generated = generation_pipeline(
        prompt,
        max_length=200,      # Increase max_length for more detailed responses
        num_return_sequences=1,
        do_sample=True,
        top_p=0.9,           # Nucleus sampling for diverse responses
        top_k=50             # Top-k sampling for quality control
    )
    
    # Extract the chatbot's response
    ai_response = generated[0]["generated_text"].split("Chatbot:")[1].strip()
    print(f"Dental Chatbot Response: {ai_response}")
    return ai_response

# Gradio ChatInterface
demo = gr.ChatInterface(
    fn=dental_chatbot_response,
    title="Advanced Dental Terminology Chatbot",
    description=(
        "Ask me anything about dental terms, procedures, and treatments! "
        "This chatbot is powered by an advanced LLM for detailed and accurate answers."
    )
)

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