app.py

import torch
import gradio as gr
from transformers import pipeline, AutoTokenizer
from datetime import datetime
from db import chat_history_collection

# Load Model & Tokenizer
MODEL_NAME = "TinyLlama/TinyLlama-1.1B-Chat-v1.0"

pipe = pipeline(
    "text-generation",
    model=MODEL_NAME,
    torch_dtype=torch.bfloat16,
    device_map="auto"
)

tokenizer = AutoTokenizer.from_pretrained(MODEL_NAME)  # Load tokenizer

def clean_response(response_text):
    """Removes unwanted system, user, and assistant tags from the response."""
    if "<|assistant|>" in response_text:
        return response_text.split("<|assistant|>")[-1].strip()
    return response_text.strip()

def chatbot_response(user_message, session_id="default_session", user_id="user_123"):
    """Generate a chatbot response using context from past chats."""
    
    try:
        # Fetch last 10 messages
        past_chats = list(chat_history_collection.find(
            {"session_id": session_id}).sort("timestamp", -1).limit(10)
        )

        messages = [{"role": "system", "content": "You are a friendly chatbot."}]

        # Add past messages to maintain context
        for chat in reversed(past_chats):
            messages.append({"role": "user", "content": chat["message"]})
            messages.append({"role": "assistant", "content": chat["response"]})

        # Append new user message
        messages.append({"role": "user", "content": user_message})

        # Generate prompt for model
        prompt = tokenizer.apply_chat_template(messages, tokenize=False, add_generation_prompt=True)

        # Generate response
        outputs = pipe(prompt, max_new_tokens=150, do_sample=True, temperature=0.7, top_k=50, top_p=0.95)
        raw_response = outputs[0]["generated_text"]

        # Clean response
        cleaned_response = clean_response(raw_response)

        # Save to database
        chat_data = {
            "session_id": session_id,
            "user_id": user_id,
            "message": user_message,
            "response": cleaned_response,
            "timestamp": datetime.utcnow()
        }
        chat_history_collection.insert_one(chat_data)

        return cleaned_response

    except Exception as e:
        return f"Error: {str(e)}"

# Gradio UI
iface = gr.Interface(
    fn=chatbot_response,
    inputs=gr.Textbox(label="User Message"),
    outputs=gr.Textbox(label="Chatbot Response"),
    title="TinyLlama Chatbot",
    description="Chat with an AI-powered assistant.",
    live=True
)

# Run Gradio app
if __name__ == "__main__":
    iface.launch(share=True)