app.py

import streamlit as st
import os
import tempfile
import faiss
import fitz  # PyMuPDF for PDFs
import docx
import openpyxl
from langchain_community.embeddings import HuggingFaceEmbeddings
from langchain.vectorstores import FAISS
from langchain.text_splitter import RecursiveCharacterTextSplitter
from langchain.docstore.document import Document
from langchain_community.llms import Groq
from langchain.chains import RetrievalQA
from langchain.schema import Document as LCDocument

# Initialize LLM
llm = Groq(
    model="llama3-8b-8192",
    api_key=os.getenv("GROQ_API_KEY")  # Put this in Hugging Face secrets
)

# Embeddings model
embedding_model = HuggingFaceEmbeddings(model_name="sentence-transformers/all-MiniLM-L6-v2")

# File processors
def read_pdf(file_path):
    text = ""
    doc = fitz.open(file_path)
    for page in doc:
        text += page.get_text()
    return text

def read_docx(file_path):
    doc = docx.Document(file_path)
    return "\n".join([p.text for p in doc.paragraphs])

def read_excel(file_path):
    wb = openpyxl.load_workbook(file_path, data_only=True)
    text = ""
    for sheet in wb.sheetnames:
        ws = wb[sheet]
        for row in ws.iter_rows(values_only=True):
            text += " ".join([str(cell) for cell in row if cell is not None]) + "\n"
    return text

def process_file(uploaded_file):
    suffix = uploaded_file.name.split(".")[-1]
    with tempfile.NamedTemporaryFile(delete=False, suffix="." + suffix) as tmp_file:
        tmp_file.write(uploaded_file.read())
        tmp_path = tmp_file.name

    if suffix.lower() == "pdf":
        return read_pdf(tmp_path)
    elif suffix.lower() in ["docx"]:
        return read_docx(tmp_path)
    elif suffix.lower() in ["xlsx"]:
        return read_excel(tmp_path)
    else:
        return "Unsupported file type."

# Streamlit UI
st.title("📄 RAG Document QA with Faiss + LLaMA3")

uploaded_file = st.file_uploader("Upload a PDF, Word or Excel file", type=["pdf", "docx", "xlsx"])

if uploaded_file:
    st.success("✅ File uploaded successfully.")
    raw_text = process_file(uploaded_file)

    # Split text into chunks
    splitter = RecursiveCharacterTextSplitter(chunk_size=500, chunk_overlap=50)
    texts = splitter.split_text(raw_text)
    docs = [Document(page_content=t) for t in texts]

    # Embed and create vector store
    with st.spinner("Indexing document..."):
        db = FAISS.from_documents(docs, embedding_model)
        retriever = db.as_retriever(search_type="similarity", search_kwargs={"k": 4})
        qa = RetrievalQA.from_chain_type(llm=llm, chain_type="stuff", retriever=retriever)

    st.success("✅ Document indexed! Ask your questions below:")

    user_query = st.text_input("❓ Ask a question about your document")
    if user_query:
        with st.spinner("Generating answer..."):
            answer = qa.run(user_query)
            st.markdown(f"**💬 Answer:** {answer}")