app.py

import streamlit as st
from unsloth import FastLanguageModel
from transformers import TextStreamer

# Page Configuration
st.set_page_config(page_title="AI Traffic Law Advisor", layout="wide")

# Load the LoRA model
MODEL_PATH = "./lora_model"

@st.cache_resource(show_spinner=False)
def load_model():
    # Load model and tokenizer
    model, tokenizer = FastLanguageModel.from_pretrained(
        MODEL_PATH,
        device_map="auto"
    )
    # Enable inference mode
    model = FastLanguageModel.for_inference(model)
    return model, tokenizer

model, tokenizer = load_model()

st.title("AI Traffic Law Advisor")

user_query = st.text_area("Enter your legal question about traffic rules in India:", "")

if st.button("Get Advice"):
    if user_query.strip():
        messages = [{"role": "user", "content": user_query}]
        # Tokenize input
        inputs = tokenizer.apply_chat_template(
            messages,
            tokenize=True,
            add_generation_prompt=True,
            return_tensors="pt"
        ).to(model.device)
        
        # Stream response
        text_streamer = TextStreamer(tokenizer, skip_prompt=True)
        
        st.markdown("**AI Response:**")
        with st.spinner("Generating response..."):
            model.generate(input_ids=inputs, streamer=text_streamer, max_new_tokens=1048, temperature=0.7)
    else:
        st.warning("Please enter a query.")