st-chat-message's Introduction

st-chat-message

Description

This is a simple chat message component for streamlit. It is based on the streamlit-chat component, trying to be as compatible as possible, but it adding a few features:

Markdown support
LaTeX support
Tables

Installation

pip install st-chat-message

poetry add st-chat-message

Usage

import streamlit as st
from st_chat_message import message

message("Hello world!", is_user=True)
message("Hi")

Buiding from source

Prerequisites

nodejs >= 18.x
yarn >= 1.22.x
poetry >= 1.2.x
python >= 3.8.x

Building

./build.sh

Publishing

poetry publish

License

This project is licensed under the MIT License - see the LICENSE file for details

st-chat-message's People

Contributors

Stargazers

Watchers

st-chat-message's Issues

Stream Problem

Hello, thank you for providing such a great chat-message library. I have understood how to use it correctly according to your example and can also use stream.

But when I use it with LangChain's CallBack, I directly use st.experimental_rerun() will cause the entire script to run again, making the langchain callBack function unable to continue executing, making it permanently stuck in the first place. Here is my code:

class StreamHandler(BaseCallbackHandler):
    def __init__(self, stream_key: str, initial_text=""):
        # self.text = initial_text
        self.key = stream_key

    def on_llm_start(
        self,
        *args,
        **kwargs: Any,
    ) -> Any:
        st.session_state.current_text = ''
        st.session_state.partial = True
        message(st.session_state.current_text, key=key, partial=st.session_state.partial)


    def on_llm_new_token(self, token: str, **kwargs) -> None:
        st.session_state.current_text += token
        st.experimental_rerun()

    def on_llm_end(
        self,
        response: LLMResult,
        **kwargs
    ):
        st.session_state.partial = False

key = f"message_{len(st.session_state.langchain_messages) + 1}"
stream_handler = StreamHandler(key)
response = agent.run(prompt, callbacks=[stream_handler])

So I'm thinking, is there a way for messages to control internal state updates through Python? Just like:

class StreamHandler(BaseCallbackHandler):
    def __init__(self, container, initial_text=""):
        self.container = container
        self.text = initial_text

    def on_llm_new_token(self, token: str, **kwargs) -> None:
        self.text += token
        self.container.markdown(self.text)

with st.chat_message("assistant"):
            stream_handler = StreamHandler(st.empty())
            response = st.session_state.agent.run(prompt, callbacks=[stream_handler])

undo76 / st-chat-message Goto Github PK