app.py

import gradio as gr
import os
from langchain_huggingface import HuggingFaceEmbeddings
from langchain_community.vectorstores import FAISS
import requests

# Paths
PROCESSED_DATA_DIR = "data/preprocessed/"
HUGGINGFACE_KEY_FILE = "configs/huggingface_api_key.txt"
API_URL = "https://api-inference.huggingface.co/models/gpt2"

# Function to read the Hugging Face API key
def read_huggingface_api_key(key_file):
    try:
        with open(key_file, "r") as file:
            return file.readline().strip()
    except FileNotFoundError:
        raise FileNotFoundError(f"API key file not found: {key_file}")
    except Exception as e:
        raise Exception(f"Error reading API key: {e}")

HUGGINGFACE_API_KEY = read_huggingface_api_key(HUGGINGFACE_KEY_FILE)
headers = {"Authorization": f"Bearer {HUGGINGFACE_API_KEY}"}

# Load FAISS Index
def load_faiss_index(processed_data_dir):
    embedding_model = HuggingFaceEmbeddings(model_name="sentence-transformers/all-MiniLM-L6-v2")
    vector_store = FAISS.load_local(
        processed_data_dir, embedding_model, allow_dangerous_deserialization=True
    )
    return vector_store

vector_store = load_faiss_index(PROCESSED_DATA_DIR)

# Query GPT-2 Model via Hugging Face API
def query_huggingface_api(prompt):
    response = requests.post(API_URL, headers=headers, json={"inputs": prompt})
    if response.status_code == 200:
        return response.json()[0]["generated_text"]
    else:
        return f"Error {response.status_code}: {response.text}"

# Generate Response
def generate_response(query):
    retriever = vector_store.as_retriever()
    retrieved_chunks = retriever.get_relevant_documents(query)

    if not retrieved_chunks:
        return "No relevant documents found."

    context = "\n\n".join([doc.page_content for doc in retrieved_chunks[:3]])[:1500]
    prompt = (
        f"You are a legal expert specializing in business laws and the legal environment. "
        f"Using the following context, answer the question concisely and accurately.\n\n"
        f"Context:\n{context}\n\nQuestion: {query}\n\nAnswer:"
    )
    return query_huggingface_api(prompt)

# Gradio Interface for QA Bot
def qa_bot(query):
    return generate_response(query)

# Define Gradio Interface
demo = gr.Interface(
    fn=qa_bot,
    inputs=gr.Textbox(label="Enter your question about business laws:"),
    outputs=gr.Textbox(label="Answer"),
    title="Business Law QA Bot",
    description="Ask questions about business laws and the legal environment. Powered by FAISS and GPT-2.",
)

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