PDF Chatbot with Streamlit and LangChain

This project creates a PDF chatbot using Streamlit and LangChain. The chatbot can read and understand uploaded PDF files, allowing users to ask questions and receive answers based on the content of those documents.

Prerequisites

Ensure you have the following libraries installed:

• os
• streamlit
• langchain
• langchain_chroma
• langchain_core
• langchain_openai

You can install the required libraries using pip:

pip install streamlit langchain langchain_chroma langchain_core langchain_openai

Project Structure

• Main functionality:
  • Upload PDF files.
  • Load and read the content of PDF files.
  • Process the content and create embeddings.
  • Ask questions based on the content and get responses from the chatbot.

Code Explanation

1. Load ChatGPT Model

The ChatOpenAI model is loaded to generate responses based on user queries.

llm = ChatOpenAI(model="gpt-3.5-turbo-0125")

2. Read Documents

A function to read documents from a specified directory.

def read_doc(directory):
    file_loader = PyPDFDirectoryLoader(directory)
    documents = file_loader.load()
    return documents

3. Format Documents

A function to format the documents into a single string.

def format_docs(docs):
    return "\n\n".join(doc.page_content for doc in docs)

4. Initialize Session State

Initialize the session state to keep track of messages.

if "messages" not in st.session_state:
    st.session_state.messages = []

5. Sidebar Layout for File Upload

Create a sidebar for file uploads.

st.sidebar.title("Upload PDF")
uploaded_files = st.sidebar.file_uploader("Choose PDF file/files", accept_multiple_files=True)
for uploaded_file in uploaded_files or []:
    if uploaded_file is not None:
        file_name = uploaded_file.name
        save_directory = "documents/"
        if not os.path.exists(save_directory):
            os.makedirs(save_directory)
        save_path = os.path.join(save_directory, file_name)
        with open(save_path, "wb") as f:
            f.write(uploaded_file.getbuffer())
        st.sidebar.success(f"File '{file_name}' uploaded successfully!")

6. Main Content Layout

Display the chat history and handle user inputs.

st.title("🤖PDF Chatbot🤖")

all_messages = st.session_state.messages
for message in all_messages:
    with st.container():
        with st.chat_message(message["role"]):
            st.write(message["prompt"])
        with st.chat_message("assistant"):
            st.write(message["response"])

7. Main Functionality

Load documents, create embeddings, and handle user queries.

status = "Failed"
if uploaded_files:
    status = "Success"
    docs = read_doc('documents/')
    text_splitter = RecursiveCharacterTextSplitter(chunk_size=1000, chunk_overlap=200)
    splits = text_splitter.split_documents(docs)
    vectorstore = Chroma.from_documents(documents=splits, embedding=OpenAIEmbeddings())
    retriever = vectorstore.as_retriever()
    prompt = hub.pull("rlm/rag-prompt")
    rag_chain = (
        {"context": retriever | format_docs, "question": RunnablePassthrough()}
        | prompt
        | llm
        | StrOutputParser()
    )
    prompt = st.text_input("Ask your question")
    if prompt:
        response = rag_chain.invoke(prompt)
        st.session_state.messages.append({"role": "user", "prompt": prompt, "response": response})

        with st.container():
                st.markdown(
                    f"<div style='display: flex; align-items: flex-start;'>"
                    f"<div>{ai_icon}</div>"
                    f"<div>{response}</div>"
                    f"</div>",
                    unsafe_allow_html=True
                )
        rag_chain.invoke(" ")

8. CSS for Chat Bar

CSS to pin the chat bar to the bottom of the screen.

st.markdown(
    """
    <style>
    .stInput {
        position: fixed;
        bottom: 0;
        left: 0;
        width: 100%;
        z-index: 1;
    }
    </style>
    """,
    unsafe_allow_html=True
)

How to Run

1. Ensure all prerequisites are installed.
2. Save the script as pdf_chatbot.py.
3. Run the script using Streamlit:

   streamlit run pdf_chatbot.py

4. Upload your PDF files through the sidebar.
5. Ask questions using the text input at the bottom of the page.

Conclusion

This project demonstrates how to create a chatbot that can understand and respond to questions based on the content of uploaded PDF documents. By leveraging Streamlit and LangChain, we can build an interactive and user-friendly application.