app.py

import gradio as gr
from datasets import load_dataset
from sentence_transformers import SentenceTransformer
import faiss
import numpy as np
from transformers import pipeline

# Load dataset
dataset = load_dataset("lex_glue", "scotus")
corpus = [doc['text'] for doc in dataset['train'].select(range(200))]  # just 200 to keep it light

# Embedding model
embedder = SentenceTransformer('sentence-transformers/all-MiniLM-L6-v2')
corpus_embeddings = embedder.encode(corpus, convert_to_numpy=True)

# Build FAISS index
dimension = corpus_embeddings.shape[1]
index = faiss.IndexFlatL2(dimension)
index.add(corpus_embeddings)

# Text generation model
gen_pipeline = pipeline("text2text-generation", model="facebook/bart-large-cnn")

# RAG-like query function
def rag_query(user_question):
    question_embedding = embedder.encode([user_question])
    _, indices = index.search(np.array(question_embedding), k=3)
    context = " ".join([corpus[i] for i in indices[0]])

    prompt = f"Question: {user_question}\nContext: {context}\nAnswer:"
    result = gen_pipeline(prompt, max_length=250, do_sample=False)[0]['generated_text']
    return result

# Gradio UI
def chatbot_interface(query):
    return rag_query(query)

iface = gr.Interface(fn=chatbot_interface,
                     inputs="text",
                     outputs="text",
                     title="🧑‍⚖️ Legal Assistant Chatbot",
                     description="Ask legal questions based on case data (LexGLUE - SCOTUS subset)")

iface.launch()