app.py

import streamlit as st

import transformers
import pdfplumber
from langchain.vectorstores import FAISS
from langchain.embeddings import HuggingFaceEmbeddings
from langchain.text_splitter import CharacterTextSplitter
from langchain.docstore.document import Document
from langchain.chains import RetrievalQA
from langchain.prompts import PromptTemplate
from langchain.chains.combine_documents.base import BaseCombineDocumentsChain
from langchain.chains import LLMChain

# Use Groq LLM from langchain-community
from langchain_community.llms import Groq

# Load PDFs from a folder using pdfplumber
def load_pdfs_from_folder(folder_path):
    docs = []
    for fname in os.listdir(folder_path):
        if fname.endswith(".pdf"):
            full_path = os.path.join(folder_path, fname)
            text = ""
            with pdfplumber.open(full_path) as pdf:
                for page in pdf.pages:
                    text += page.extract_text() or ""
            docs.append(Document(page_content=text, metadata={"source": fname}))
    return docs

# Answer a question via Groq LLM (fallback)
def ask_groq(question):
    llm = Groq(api_key=st.secrets["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"], model="mixtral-8x7b-32768")
    return llm.predict(question)

def main():
    st.set_page_config(page_title="RAG Chatbot with Groq", layout="centered")
    st.title("📚 RAG Chatbot with Groq Fallback")

    uploaded_files = st.file_uploader("Upload one or more PDF files", type=["pdf"], accept_multiple_files=True)
    question = st.text_input("Ask a question:")
    process_btn = st.button("Process and Answer")

    if process_btn and uploaded_files and question:
        with st.spinner("📂 Loading documents..."):
            docs = load_uploaded_pdfs(uploaded_files)

        # Chunking
        text_splitter = CharacterTextSplitter(chunk_size=1000, chunk_overlap=200)
        chunks = text_splitter.split_documents(docs)

        # Embedding
        embeddings = HuggingFaceEmbeddings(model_name="sentence-transformers/all-MiniLM-L6-v2")
        db = FAISS.from_documents(chunks, embeddings)

        retriever = db.as_retriever(search_kwargs={"k": 3})

        # Check if relevant documents exist
        relevant_docs = retriever.get_relevant_documents(question)
        if not relevant_docs:
            st.warning("❓ No relevant documents found. Asking Groq LLM...")
            answer = ask_groq(question)
        else:
            llm = Groq(api_key=st.secrets["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"], model="mixtral-8x7b-32768")
            qa = RetrievalQA.from_chain_type(llm=llm, retriever=retriever, return_source_documents=True)
            result = qa({"query": question})
            answer = result['result']

        st.success("💬 Answer:")
        st.write(answer)

if __name__ == "__main__":
    main()