Skip to main content

Chatbot

An intuitive chatbot card for basic prompting, uselful when showcasing LLMs. The message text is interpreted as Markdown.

Check the full API at ui.chatbot_card.

Basic chatbot​

chatbot-0

from h2o_wave import data

q.page['example'] = ui.chatbot_card(
    box='1 1 5 5',
    name='chatbot',
    data=data(fields='content from_user', t='list', rows=[
        ['Hello, buddy. Can you help me?', True],
        ['Sure, what you need?', False],
    ]),
)

With a stop button​

Chatbot card supports text streaming out of the box. Specifying the generating attribute renders a stop button that notifies the app about the user wishing to stop the stream. See full example to learn more.

chatbot-1

from h2o_wave import data

q.page['example'] = ui.chatbot_card(
    box='1 1 5 5',
    name='chatbot', 
    data=data(fields='content from_user', t='list', rows=[
        ['Hello, buddy. Can you help me?', True],
        ['Sure, what you need?', False],
    ]),
    generating=True,
    events=['stop']
)

With infinite scroll​

Some chats can get lengthy very quickly. Use scroll_up event to avoid loading the whole chat history in a single go for better performance (loading smaller chunks is faster then loading a big one) and stability (too many messages can break the browser + user may not even want to see them all). See this example to learn more.

chabot-infinite-scroll

from h2o_wave import data

q.page['example'] = ui.chatbot_card(
    box='1 1 5 5',
    name='chatbot', 
    data=data(fields='content from_user', t='list', rows=[
        ['Hello, buddy. Can you help me?', True],
        ['Sure, what you need?', False],
    ]),
    events=['scroll']
)

Collect feedback​

Add the thumbs up and thumbs down buttons below the chatbot response to capture user feedback by configuring the feedback event. See full example to learn more.

chatbot-2

from h2o_wave import data

q.page['example'] = ui.chatbot_card(
    box='1 1 5 5',
    name='chatbot', 
    data=data(fields='content from_user', t='list', rows=[
        ['Hello, buddy. Can you help me?', True],
        ['Sure, what you need?', False],
    ]),
    events=['feedback']
)

With suggestions​

Use suggestions to provide user with pre-defined prompt options. Use in combination with suggestion event. See full example to learn more.

chatbot-suggestions

from h2o_wave import data

q.page['example'] = ui.chatbot_card(
    box='1 1 5 5',
    name='chatbot', 
    data=data(fields='content from_user', t='list'),
    events=['suggestion'],
    suggestions=[
        ui.chat_suggestion('sug1', label="Write a poem", caption="about H2O Wave", icon="Edit"),
        ui.chat_suggestion('sug2', label="Plan a trip", caption="to Europe", icon="Airplane"),
        ui.chat_suggestion('sug3', label="Give me ideas", caption="for a new project", icon="Lightbulb"),
        ui.chat_suggestion('sug4', label="Explain me", caption="CSS preprocessors", icon="Code")
    ],
)

Disable input​

Disable user input with disabled property. This can be handy e.g. to limit user input to suggestions only.

chatbot-3

from h2o_wave import data

q.page['example'] = ui.chatbot_card(
    box='1 1 5 5',
    name='chatbot', 
    data=data(fields='content from_user', t='list', rows=[
        ['Hello, buddy. Can you help me?', True],
        ['Sure, what you need?', False],
    ]),
    disabled=True
)