app.py

import os
import gradio as gr

from langchain.document_loaders import PyPDFLoader
from langchain.text_splitter import RecursiveCharacterTextSplitter
from langchain.embeddings import HuggingFaceEmbeddings
from langchain.vectorstores import FAISS
from langchain.llms import HuggingFaceHub
from langchain.chains import RetrievalQA
from langchain.prompts import PromptTemplate
from langchain.chains import LLMChain

# -- Set Hugging Face API token from secret --
os.environ["HUGGINGFACEHUB_API_TOKEN"] = os.getenv("HF_API_TOKEN")

# -- Shared LLM and Embedder --
embedder = HuggingFaceEmbeddings(model_name="sentence-transformers/all-MiniLM-L6-v2")
llm = HuggingFaceHub(
    repo_id="google/flan-t5-large",
    model_kwargs={"temperature": 0.3, "max_length": 512}
)

# -- Internal state variables --
global_chunks = []

def build_qa_chain(pdf_file):
    global global_chunks

    loader = PyPDFLoader(pdf_file.name)
    docs = loader.load()

    chunks = RecursiveCharacterTextSplitter(
        chunk_size=500, chunk_overlap=50
    ).split_documents(docs)

    global_chunks = chunks  # store for summarizer

    db = FAISS.from_documents(chunks, embedder)
    return RetrievalQA.from_chain_type(llm=llm, retriever=db.as_retriever())

def summarize_document():
    if not global_chunks:
        return "⚠️ Please upload and index a document first."

    # Merge all text from chunks
    full_text = " ".join(chunk.page_content for chunk in global_chunks)

    # Use a prompt template for summarization
    prompt = PromptTemplate(
        input_variables=["doc"],
        template="Please summarize the following document in a few concise bullet points:\n\n{doc}"
    )
    chain = LLMChain(llm=llm, prompt=prompt)
    summary = chain.run(full_text[:3000])  # Trim to fit model input
    return summary


# --- Gradio UI ---
with gr.Blocks() as demo:
    gr.Markdown("# AskYourDocs 📄🤖\nUpload a PDF, then ask questions or get a summary.")

    pdf_input = gr.File(label="Upload PDF", file_types=[".pdf"])
    build_btn = gr.Button("Index Document")
    summarize_btn = gr.Button("Summarize Document")

    status = gr.Markdown("")
    qa_state = gr.State()

    question = gr.Textbox(label="Ask a question")
    answer = gr.Textbox(label="Answer", interactive=False)
    summary_output = gr.Textbox(label="Summary", lines=10, interactive=False)

    # Step 1: Build vector index
    def _index(pdf):
        if pdf is None:
            return "⚠️ Please upload a PDF first.", None
        chain = build_qa_chain(pdf)
        return "✅ Document indexed!", chain

    build_btn.click(_index, inputs=pdf_input, outputs=[status, qa_state])

    # Step 2: Ask questions
    def _ask(q, chain):
        if chain is None:
            return "⚠️ Upload and index a PDF first."
        if not q.strip():
            return "⚠️ Ask something..."
        return chain.run(q)

    question.submit(_ask, inputs=[question, qa_state], outputs=answer)

    # Step 3: Summarize document
    summarize_btn.click(fn=summarize_document, outputs=summary_output)

demo.launch()