app.py

import gradio as gr
import numpy as np
import faiss
from transformers import pipeline, AutoTokenizer, AutoModelForQuestionAnswering
from sentence_transformers import SentenceTransformer
from utils import get_wikipedia_content, split_text, create_faiss_index

# Load models
embedding_model = SentenceTransformer("sentence-transformers/all-mpnet-base-v2")
qa_model_name = "deepset/roberta-base-squad2"
qa_tokenizer = AutoTokenizer.from_pretrained(qa_model_name)
qa_model = AutoModelForQuestionAnswering.from_pretrained(qa_model_name)
qa_pipeline = pipeline("question-answering", model=qa_model, tokenizer=qa_tokenizer)

# Function to handle user query
def answer_question(topic, question):
    # Fetch Wikipedia content
    document = get_wikipedia_content(topic)
    if not document:
        return "Could not retrieve information from Wikipedia."

    # Split content into chunks
    chunks = split_text(document)
    
    # Create FAISS index
    index, embeddings = create_faiss_index(chunks)
    
    # Encode query and search relevant chunks
    query_embedding = embedding_model.encode([question])
    k = 3
    distances, indices = index.search(np.array(query_embedding), k)
    retrieved_chunks = [chunks[i] for i in indices[0]]
    
    # Generate answer using QA model
    context = " ".join(retrieved_chunks)
    answer = qa_pipeline(question=question, context=context)
    
    return f"**Answer:** {answer['answer']}"

# Create Gradio Interface
iface = gr.Interface(
    fn=answer_question,
    inputs=[gr.Textbox(label="Enter a Wikipedia Topic"), gr.Textbox(label="Ask a Question")],
    outputs="text",
    title="Wikipedia Q&A System",
    description="Ask questions about any Wikipedia topic using AI-powered retrieval and QA models.",
)

if __name__ == "__main__":
    iface.launch(share=True)