app.py

import gradio as gr
from PyPDF2 import PdfReader
from transformers import AutoTokenizer, AutoModelForCausalLM, pipeline
from langchain.text_splitter import RecursiveCharacterTextSplitter
from langchain.vectorstores import FAISS
from langchain.embeddings import HuggingFaceEmbeddings
from langchain.chains import RetrievalQA
from langchain.llms import HuggingFacePipeline
model_id = "ibm-granite/granite-3.3-2b-instruct"
tokenizer = AutoTokenizer.from_pretrained(model_id)
model = AutoModelForCausalLM.from_pretrained(
    model_id,
    torch_dtype="auto",
    device_map="auto"
)

# 2. Extract text from PDF
def extract_text(pdf_file):
    reader = PdfReader(pdf_file)
    text = ""
    for page in reader.pages:
        if page.extract_text():
            text += page.extract_text() + "\n"
    return text
def build_qa_chain(pdf_text):
    splitter = RecursiveCharacterTextSplitter(chunk_size=1000, chunk_overlap=200)
    chunks = splitter.split_text(pdf_text)
    
    embeddings = HuggingFaceEmbeddings()
    vectorstore = FAISS.from_texts(chunks, embeddings)
    retriever = vectorstore.as_retriever()
    
    llm_pipeline = pipeline(
        "text-generation", 
        model=model, 
        tokenizer=tokenizer,
        max_length=512,
        temperature=0
    )
    llm = HuggingFacePipeline(pipeline=llm_pipeline)
    
    return RetrievalQA.from_chain_type(llm=llm, retriever=retriever)

# Store QA chain globally
qa_chain = None

def process_pdf(file):
    global qa_chain
    text = extract_text(file)
    qa_chain = build_qa_chain(text)
    return
def answer_question(question):
    if qa_chain is None:
        return "❌ Please upload a PDF first."
    return qa_chain.run(question)

# UI with Gradio
with gr.Blocks() as demo:
    gr.Markdown("## 📄 Granite 2B — PDF Q&A")

    with gr.Row():
        pdf_input = gr.File(label="Upload PDF", type="filepath")
        status_output = gr.Textbox(label="Status")
    pdf_input.upload(process_pdf, pdf_input, status_output)

    with gr.Row():
        question_input = gr.Textbox(label="Ask a question")
        answer_output = gr.Textbox(label="Answer")
        ask_btn = gr.Button("Ask")
    ask_btn.click(answer_question, question_input, answer_output)
demo.launch()