Option/Button to download the chat

[thesherzad]

Assuming someone is using the chat for Exploratory Data Analysis (EDA) they might want to download the chat and use as supporting document. Given the fact that when app closes, the previous chats won't be available, it'd be good if user has that option to download the chat.

[chendaniely]

querychat returns a QueryChat object which gives you access to the underlying chatlas object. In the examples you see it's using chat.df() to get the filtered df, but you can use chat.chat() to get the chatlas object.

Once you have the chatlas object, you have access to all the underlying chat data, including the .get_turns() method or .export() method

I patched together the code from .export() to export out the chat into a markdown file:

In the UI:

ui.download_button("download", "Download"),

In the server:

    # https://github.com/posit-dev/chatlas/blob/8871240172dd470f7eee10028c3be8ac9f64a5eb/chatlas/_chat.py#L1471C13-L1471C22
    @render.download(
        filename=lambda: f"chat.md"
    )
    def download():
        turns = chat.chat().get_turns(include_system_prompt=True)

        content_arr = []
        for turn in turns:
            turn_content = "\n\n".join(
                [
                    str(x).strip()
                    for x in turn.contents
                ]
            )

            turn_content = f"## {turn.role.capitalize()}\n\n{turn_content}"
            content_arr.append(turn_content)
        contents = "\n\n".join(content_arr)
        yield contents

If you only want the last response (as suggest in #4) you would use .get_last_turn() instead of get_turns().
You can then process the data however you need.

As far as a specific type of output that formatted the way you want, that might be up to you to manage, or something that would have to be in the chatlas package.

Full working app example:

from pathlib import Path

import chatlas
from seaborn import load_dataset
from shiny import App, render, ui, reactive

import querychat as qc

titanic = load_dataset("titanic")

def use_github_models(system_prompt: str) -> chatlas.Chat:
    return chatlas.ChatGithub(
        model="gpt-4.1",
        system_prompt=system_prompt,
    )

querychat_config = qc.init(
    titanic,
    "titanic",
    create_chat_callback=use_github_models,
)

app_ui = ui.page_sidebar(
    ui.sidebar(
        qc.ui("chat"),
        ui.download_button("download", "Download"),
        width="400px"
    ),

    ui.card(
        ui.output_data_frame("data_table"),
        fill=True,
    ),

    title="querychat with Python",
    fillable=True,
)

def server(input, output, session):
    chat = qc.server("chat", querychat_config)

    @render.data_frame
    def data_table():
        return chat.df()


    # https://github.com/posit-dev/chatlas/blob/8871240172dd470f7eee10028c3be8ac9f64a5eb/chatlas/_chat.py#L1471C13-L1471C22
    @render.download(
        filename=lambda: f"chat.md"
    )
    def download():
        turns = chat.chat().get_turns(include_system_prompt=True)

        content_arr = []
        for turn in turns:
            turn_content = "\n\n".join(
                [
                    str(x).strip()
                    for x in turn.contents
                ]
            )

            turn_content = f"## {turn.role.capitalize()}\n\n{turn_content}"
            content_arr.append(turn_content)
        contents = "\n\n".join(content_arr)
        yield contents

app = App(app_ui, server)