Project Setup

Python

To run the project, you will need Python setup locally. We have designed the project to work with Python v3.11.

If you do not have Python installed, you can follow the installation instructions on Python.org.

You can verify your Python version by running the following command.

python --version

Setting up the Project

The project repository is hosted on Github.

You can use the GitHub UI or CLI to clone the repository or download a ZIP archive. We recommend forking the repository so you have a personal copy for future reference.

To clone the repository using the git CLI, you can run the following command.

git clone https://github.com/neo4j-graphacademy/llm-chatbot-python

Installing Dependencies

The project has four dependencies, streamlit, langchain, openai and neo4j-driver. To install the dependencies, you can run the pip install command.

pip install -r requirements.txt

Starting the App

The bot.py file in the root folder contains the main application code.

To start the app, run the streamlit run command.

streamlit run bot.py

This command command will start a server on http://localhost:8501. Once you have run the command, you should see the following interface.

When you send a message, the message will be rendered with a red icon to represent a user message. The app will wait for one second, and then render the same message next to an orange robot icon, which represents an assistant message.

Page Config

The code sets out by calling the st.set_page_config() to configure the title and icon used on the page.

# Page Config
st.set_page_config("Ebert", page_icon=":movie_camera:")

App Session State

The next block of code checks the session state for the current user. The session is used to save a list of messages between the user and the LLM.

If an array of messages has not already been set, a list of messages containing a default greeting from the assistant message is set.

# Set up Session State
if "messages" not in st.session_state:
    st.session_state.messages = [
        {"role": "assistant", "content": "Hi, I'm the GraphAcademy Chatbot!  How can I help you?"},
    ]

The session state will persist for as long as the user keeps their browser tab open.

As the app state changes, certain sections of the UI may be re-rendered. Storing a list of messages to the session state ensures that no information is lost during the re-rendering.

Chat Messages

Within a container, any messages held in the session state are written to the screen using the write_message() helper function.

# Set up Session State
if "messages" not in st.session_state:
    st.session_state.messages = [
        {"role": "assistant", "content": "Hi, I'm the GraphAcademy Chatbot!  How can I help you?"},
    ]

For berevity, the write_message() helper function has been abstracted into the utils.py file.

def write_message(role, content, save = True):
    """
    This is a helper function that saves a message to the
     session state and then writes a message to the UI
    """
    # Append to session state
    if save:
        st.session_state.messages.append({"role": role, "content": content})

    # Write to UI
    with st.chat_message(role):
        st.markdown(content)

The function accepts two positional arguments; the role of the author, either human or assistant, and the message. An additional save parameter can be passed to instruct the function to append the message to the session state.

The block concludes by setting a prompt variable that will contain the user input. When the user sends their message, the write_message() function is used to save the message to the session state and render the message in the UI.

Handling Submissions

The handle_submit() mocks an interaction by calling the sleep() method to pause for a second, before repeating the user’s input.

# Submit handler
def handle_submit(message):
    """
    Submit handler:

    You will modify this method to talk with an LLM and provide
    context using data from Neo4j.
    """

    # Handle the response
    with st.spinner('Thinking...'):
        # # TODO: Replace this with a call to your LLM
        from time import sleep
        sleep(1)
        write_message('assistant', message)

You will modify this function to add interactions with the LLM.