app.py

import streamlit as st
import os
from pinecone import Pinecone, ServerlessSpec
from sentence_transformers import SentenceTransformer
import numpy as np
from datasets import load_dataset

# ✅ Step 1: Fetch Pinecone API key from Hugging Face secrets
if "PINECONE_API_KEY" not in st.secrets:
    st.error("🚨 Pinecone API key not found! Please set it in Hugging Face secrets.")
    st.stop()

pinecone_api_key = st.secrets["PINECONE_API_KEY"]  # ✅ Now it's properly defined

# ✅ Step 2: Initialize Pinecone client
pc = Pinecone(api_key=pinecone_api_key)

# ✅ Step 3: Connect to your existing Pinecone index
index_name = "legal-docs-index"
index = pc.Index(index_name)

# ✅ Step 4: Load embedding model
model = SentenceTransformer("text-embedding-ada-002")

# ✅ Step 5: Load dataset (for reference)
dataset = load_dataset("macadeliccc/US-LegalKit", split="train")
law_texts = [item['text'] for item in dataset if 'text' in item]

# ✅ Step 6: Function to search Pinecone index
def search_pinecone(query, top_k=5):
    query_embedding = model.encode([query]).tolist()
    results = index.query(query_embedding, top_k=top_k, include_metadata=True)
    return [match['metadata']['text'] for match in results['matches']]

# ✅ Step 7: Streamlit UI
st.title("🔍 Legal AI Assistant (US-LegalKit)")

query = st.text_input("📌 Enter your legal query:")

if query:
    results = search_pinecone(query)
    st.write("### 📄 Relevant Legal Documents:")
    for i, doc in enumerate(results, 1):
        st.write(f"**{i}.** {doc[:500]}...")