app.py

import streamlit as st
from PyPDF2 import PdfReader
from langchain.text_splitter import RecursiveCharacterTextSplitter
import os
from langchain_google_genai import GoogleGenerativeAIEmbeddings
from langchain_community.vectorstores import FAISS
from langchain_google_genai import ChatGoogleGenerativeAI
from langchain.chains.question_answering import load_qa_chain
from langchain.prompts import PromptTemplate
#from dotenv import load_dotenv


#load_dotenv()
API_KEYS = [os.getenv("APIKEY1"), os.getenv("APIKEY2")]
current_key_index = -1




def switch_api_key():
    global current_key_index
    current_key_index = (current_key_index + 1) % len(API_KEYS)
    return API_KEYS[current_key_index]




def get_pdf_text(pdf_docs):
    text = ""
    for pdf in pdf_docs:
        pdf_reader = PdfReader(pdf)
        for page in pdf_reader.pages:
            text += page.extract_text()
    return text




def get_text_chunks(text):
    text_splitter = RecursiveCharacterTextSplitter(chunk_size=10000, chunk_overlap=1000)
    return text_splitter.split_text(text)




def get_vector_store(text_chunks):
    api_key = switch_api_key()
    embeddings = GoogleGenerativeAIEmbeddings(model="models/embedding-001", google_api_key=api_key)
    vector_store = FAISS.from_texts(text_chunks, embedding=embeddings)
    vector_store.save_local("faiss_index")




def get_conversational_chain():
    api_key = switch_api_key()
    prompt_template = """
    You are a helpful assistant that only answers based on the context provided from the PDF documents.
    Do not use any external knowledge or assumptions. If the answer is not found in the context below, reply with "I don't know."


    Context:
    {context}


    Question:
    {question}


    Answer:
    """
    model = ChatGoogleGenerativeAI(model="gemini-2.0-flash", temperature=0, google_api_key=api_key)
    prompt = PromptTemplate(template=prompt_template, input_variables=["context", "question"])
    chain = load_qa_chain(model, chain_type="stuff", prompt=prompt)
    return chain




def user_input(user_question):
    api_key = switch_api_key()
    embeddings = GoogleGenerativeAIEmbeddings(model="models/embedding-001", google_api_key=api_key)
    new_db = FAISS.load_local("faiss_index", embeddings, allow_dangerous_deserialization=True)
    docs = new_db.similarity_search(user_question)
    chain = get_conversational_chain()


    response = chain({"input_documents": docs, "question": user_question}, return_only_outputs=True)
    st.write("Reply: ", response["output_text"])




# Streamlit application
def main():
    st.markdown(
        """
        <style>
        .header {font-size: 20px !important;}
        .subheader {font-size: 16px !important;}
        </style>
        """,
        unsafe_allow_html=True
    )
    st.markdown('<h1 class="header">CSC 121: Computers and Scientific Thinking (Chatbot)</h1>', unsafe_allow_html=True)
    st.markdown('<h2 class="subheader">Ask a question ONLY from the CSC 121 textbook of Dr. Reed</h2>', unsafe_allow_html=True)

    user_question = st.text_input("Ask a question")

    if user_question:
        user_input(user_question)




if __name__ == "__main__":
    main()