A simple Streamlit app to chat with JCVD (powered by Claude v2) using Amazon Bedrock

jcvd_chat.py

# pylint: disable=line-too-long
"""
JCVD Chat 🕺💬

A simple Streamlit app to chat with JCVD (powered by Claude v2) using Amazon Bedrock
https://aws.amazon.com/bedrock/claude/

> Adapted from Streamlit's Build conversational apps
https://docs.streamlit.io/knowledge-base/tutorials/build-conversational-apps

Want to learn more about Anthropic Claude? Check out
https://www.anthropic.com/index/introducing-claude

⠀⢰⣶⣶⣶⣶⣶⣶⣶⣶⣶⣶⣶⣶⣶⣶⣶⣶⣶⣶⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀
⠀⢸⣿⣿⣿⣿⣿⣿⣿⣿⣿⣿⣿⣿⣿⣿⣿⣿⣿⣿⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀
⠀⢸⣿⣿⣿⣿⣿⣿⣿⣿⣿⣿⣿⣿⣿⣿⣿⣿⣿⣿⠀⣶⣶⣶⣶⣶⣶⣶⡆⠀
⠀⢸⣿⣿⣿⣿⣿⣿⣿⣿⣿⣿⣿⣿⣿⣿⣿⣿⣿⣿⠀⣿⣿⣿⣿⣿⣿⣿⡇⠀
⠀⢸⣿⣿⣿⣿⣧⣤⣿⣿⣧⣼⣿⣿⣤⣿⣿⣿⣿⣿⠀⣿⣿⣿⣿⣿⣿⣿⡇⠀
⠀⢸⣿⣿⣿⣿⣿⣿⣿⣿⣿⣿⣿⣿⣿⣿⣿⣿⣿⣿⠀⣿⣿⣿⣿⣿⣿⣿⡇⠀
⠀⠈⠉⠉⠉⢹⣿⣿⡿⠋⣉⣉⣉⣉⣉⣉⣉⣉⣉⣉⣀⣿⣿⣿⣿⣿⣿⣿⡇⠀
⠀⠀⠀⠀⠀⠈⣿⡟⠁⠀⣿⣿⣿⣿⣿⣿⣿⣿⣿⣿⣿⣿⣿⣿⣿⣿⣿⣿⡇⠀
⠀⠀⠀⠀⠀⠀⠛⠀⠀⠀⠉⠉⠉⠉⠉⠉⠉⠉⠉⠉⠉⢿⣿⣿⡏⠉⠉⠉⠁⠀
⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠈⢻⣿⠁⠀⠀⠀⠀⠀
⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠛
"""

import json
import time

import boto3
import streamlit as st

st.title("JCVD Chat 🕺💬")
st.subheader("I am the Fred Astaire of chatbots!")

# Initialize chat history
if "messages" not in st.session_state:
    st.session_state.messages = []

###########################
# Inference configuration #
###########################

st.sidebar.title("Inference Configuration")

if st.sidebar.button('Clear chat history'):
    st.session_state.messages = []

st.sidebar.subheader("Randomness & Diversity")

st.sidebar.slider(
    label='Temperature',
    key='temperature',
    help='Amount of randomness injected into the response',
    value=1.0,
    min_value=0.1,
    max_value=1.0,
    step=0.1,
    format='%.1f',
)

st.sidebar.slider(
    label='Top P',
    key='top_p',
    help='Use nucleus sampling. In nucleus sampling, we compute the cumulative distribution over all the options for each subsequent token in decreasing probability order and cut it off once it reaches a particular probability specified by top_p',
    value=0.999,
    min_value=0.0,
    max_value=1.0,
    step=0.001,
    format='%.3f',
)

st.sidebar.slider(
    label='Top K',
    key='top_k',
    help='Only sample from the top K options for each subsequent token',
    value=250,
    min_value=0,
    max_value=500,
    step=1,
    format='%d',
)

st.sidebar.subheader("Length")

st.sidebar.slider(
    label='Length',
    key='length',
    help='The maximum number of tokens to generate before stopping',
    value=300,
    min_value=0,
    max_value=2048,
    step=1,
    format='%d',
)

###########
# Chat UI #
###########

# Initialize Bedrock client
session = boto3.Session()
bedrock_runtime = session.client('bedrock-runtime')

# Display messages
for message in st.session_state.messages:
    with st.chat_message(message["role"]):
        st.markdown(message["content"])

# Prompt user for input
if prompt := st.chat_input("What's up, doc?"):
    # Add user prompt to chat history
    st.session_state.messages.append(
        {
            "role": "user",
            "content": prompt
        }
    )

	# Display user prompt
    with st.chat_message("user"):
        st.markdown(prompt)

	# Display assistant response
    with st.chat_message("assistant"):
        message_placeholder = st.empty()
        full_response = ""  # pylint: disable=invalid-name
        response = bedrock_runtime.invoke_model_with_response_stream(
            accept="*/*",
            body=json.dumps(
                {
                    "prompt": f"Human: {prompt}\\nInstructions: Imagine you are JCVD\\nAssistant:",
                    "max_tokens_to_sample": st.session_state.length,
                    "temperature": st.session_state.temperature,
                    "top_k": st.session_state.top_k,
                    "top_p": st.session_state.top_p,
                    "stop_sequences": [
                        "\\n\\nHuman:"
                    ],
                    #"anthropic_version": "bedrock-2023-05-31",
				}
			),
            contentType="application/json",
            modelId="anthropic.claude-v2"
		)

        stream = response.get('body')
        if stream:
            for event in stream:
                chunk = event.get('chunk')
                if chunk:
                    chunk_obj = json.loads(chunk.get('bytes').decode())
                    text = chunk_obj['completion']
                    full_response += text
                    # Add a blinking cursor to simulate typing
                    time.sleep(0.05)
                    message_placeholder.markdown(full_response + "▌")
                message_placeholder.markdown(full_response)
        message_placeholder.markdown(full_response)

    # Add assistant response to chat history
    st.session_state.messages.append(
        {
            "role": "assistant",
            "content": full_response
        }
    )

# Display dancing JCVD
if not st.session_state.messages:
    st.markdown('![](https://media.tenor.com/CVp9l7g3axwAAAAj/jean-claude-van-damme-jcvd.gif)')