app.py

import streamlit as st
import PyPDF2
from transformers import RagTokenizer, RagRetriever, RagTokenForGeneration

# Load PDF and extract text
def load_pdf(uploaded_file):
    reader = PyPDF2.PdfReader(uploaded_file)
    text = ""
    for page in reader.pages:
        if page.extract_text():  # Ensure text extraction is valid
            text += page.extract_text() + "\n"
    return text

# Initialize RAG model
def initialize_rag_model():
    # Load the tokenizer and model
    tokenizer = RagTokenizer.from_pretrained("facebook/rag-token-nq")
    
    # Use a dummy retriever for testing purposes
    retriever = RagRetriever.from_pretrained("facebook/rag-token-nq", use_dummy_dataset=True)
    
    model = RagTokenForGeneration.from_pretrained("facebook/rag-token-nq")
    return tokenizer, retriever, model

# Process user query
def generate_answer(query, context, tokenizer, retriever, model):
    # Tokenize the input question
    inputs = tokenizer(query, return_tensors="pt")

    # Prepare inputs for the model with a dummy context
    inputs["context_input_ids"] = retriever(context, return_tensors="pt")["input_ids"]
    
    # Generate the answer
    outputs = model.generate(**inputs)
    answer = tokenizer.batch_decode(outputs, skip_special_tokens=True)
    return answer[0]

# Streamlit UI
st.title("PDF Question-Answer Chatbot")

uploaded_file = st.file_uploader("Upload a PDF file", type=["pdf"])
if uploaded_file is not None:
    text = load_pdf(uploaded_file)
    st.write("PDF loaded successfully. You can now ask questions.")
    
    # Initialize the RAG model
    tokenizer, retriever, model = initialize_rag_model()

    user_query = st.text_input("Ask a question about the PDF:")
    if user_query:
        answer = generate_answer(user_query, text, tokenizer, retriever, model)
        st.write(f"Answer: {answer}")  # Display the answer