Add shiny.ui.Chat

[cpsievert]

Adds a shiny.ui.Chat class, designed to support any LLM provider (i.e., response assistant) of your choosing (e.g., OpenAI, Anthropic, Ollama, LangChain, Google, etc). To get started, consider this bare-bones Chat example that uses no provider at all. It just displays a starting message, and then for each input message, just adds "You said: " in it's response:

App code

from shiny.express import ui

ui.page_opts(title="Hello Shiny Chat")

# Create a chat instance, with an initial message
chat = ui.Chat(
    id="chat",
    messages=[
        {"content": "Hello! How can I help you today?", "role": "assistant"},
    ],
)

# Display the chat
chat.ui()

# Define a callback to run when the user submits a message
@chat.on_user_submit
async def _():
    user_msg = chat.get_user_input()
    await chat.append_message(f"You said: {user_msg}")

An "actual" LLM-powered chat bot might look something more like this (this example requires an OpenAI API key to run). Note also that, by default, responses are interpreted as markdown strings, and code blocks are rendered with code highlighting + copy/paste:

App code

# ------------------------------------------------------------------------------------
# A basic Shiny Chat example powered by OpenAI via LangChain.
# To run it, you'll need OpenAI API key.
# To get one, follow the instructions at https://platform.openai.com/docs/quickstart
# To use other providers/models via LangChain, see https://python.langchain.com/v0.1/docs/modules/model_io/chat/quick_start/
# ------------------------------------------------------------------------------------
import os

from langchain_openai import ChatOpenAI

from shiny.express import ui

# Provide your API key here (or set the environment variable)
llm = ChatOpenAI(api_key=os.environ.get("OPENAI_API_KEY"))

# Set some Shiny page options
ui.page_opts(
    title="Hello LangChain Chat Models",
    fillable=True,
    fillable_mobile=True,
)

# Create and display an empty chat UI
chat = ui.Chat(id="chat")
chat.ui()


# Define a callback to run when the user submits a message
@chat.on_user_submit
async def _():
    # Get messages currently in the chat
    messages = chat.get_messages()
    # Create a response message stream
    response = llm.astream(messages)
    # Append the response stream into the chat
    await chat.append_message_stream(response)

A whole collection of other examples are available under the examples/chat directory. See the basic and enterprise sub-directories for getting started with various providers.

Follow up tasks

• Tests