PaLM2 Tutorial: Create your own AI-powered Virtual Assistant

Introduction

PaLM2 is next generation large language model that builds on Google’s legacy of breakthrough research in machine learning and responsible AI. It outperforms previous models, including PaLM, in various reasoning tasks such as code and math, classification, question answering, translation, and natural language generation.

Prerequisites

To use Google’s PaLM2 API you should join the waitlist To deploy our app on Streamlit, we need create an account. Go to Streamlit and create an account. It's also free! 🎉. I recommend, to use your GitHub account to create an account on Streamlit, so later you can deploy your app on Streamlit Sharing Cloud.

Getting Started

0. Create a new project directory

Create a new project directory and navigate to it in your terminal:

mkdir palm2-tutorial
cd palm2-tutorial

Create and activate virtual environment:

MacOS/Linux:
python3 -m venv .venv
source venv/bin/activate

Windows:
py -3 -m venv .venv
.venv\Scripts\activate

Install Streamlit and Google's PaLM API dependencies:

pip install streamlit
pip install streamlit_chat
pip install google-generativeai

1. Create a new Streamlit app

Create a new file called app.py. Import Streamlit and Google's PaLM API packages:

import streamlit as st
from streamlit_chat import message
import google.generativeai as palm

Add a title to the app:

st.title("PaLM Tutorial")

Initialize the messages state:

if "messages" not in st.session_state:
    st.session_state["messages"] = [{"role": "assistant", "content": "Say something to get started!"}]

Now, let's create a form to get user input and sent it to the Google’s PaLM API. We will use Streamlit's st.form() to create a form and st.columns() to create two columns. The first column will be used for user input, and the second column Send button.

with st.form("chat_input", clear_on_submit=True):
    a, b = st.columns([4, 1])

    user_prompt = a.text_input(
        label="Your message:",
        placeholder="Type something...",
        label_visibility="collapsed",
    )

    b.form_submit_button("Send", use_container_width=True)

Make user input at the left side of the screen, so it looks like a chat app:

for msg in st.session_state.messages:
    message(msg["content"], is_user=msg["role"] == "user")  # display message on the screen

Configure Google's PaLM API with your API key and get response from the API:

if user_prompt and palm_api_key:

    palm.configure(api_key=palm_api_key)  # set API key

    st.session_state.messages.append({"role": "user", "content": user_prompt})
    
    message(user_prompt, is_user=True)

    response = palm.chat(messages=[user_prompt])  # get response from Google's PaLM API

    msg = {"role": "assistant", "content": response.last}  # we are using dictionary to store message and its role. It will be useful later when we want to display chat history on the screen, to show user input at the left and AI's right side of the screen.

    st.session_state.messages.append(msg)  # add message to the chat history

    message(msg["content"])  # display message on the screen

Optionally, implement function to clear the chat history:

def clear_chat():
    st.session_state.messages = [{"role": "assistant", "content": "Say something to get started!"}]

if len(st.session_state.messages) > 1:
    st.button('Clear Chat', on_click=clear_chat)

2. Run the app

Run the app:

streamlit run app.py

You should see something like this:

Perfect! Now, let's deploy our app on Streamlit Sharing Cloud. Checkout this tutorial for detailed guide to deploy your app on Streamlit Sharing Cloud.

Conclusion

In this tutorial, we learned how to build AI-powered Virtual Assistant using Google's PaLM2 model. We used Streamlit to build the app, it is a great tool to build any kind of apps in seconds using pure Python.

View full implementation here.

Thank you for following along with this tutorial.

If you have any questions, feel free to reach out to me on LinkedIn or Twitter. I'd love to hear from you!

made with 💜 by abdibrokhim for lablab.ai tutorials.