Customizing Instant Learning

Teaching the RunLLM agent new information, based on previous conversations, is a critically important feature for many customers. One feature we offer out of the box in Slack is that instant learning is triggered, when an admin in your RunLLM organization tags a Slack message with the :bulb: emoji.

With the SDK, you can configure the emoji triggers to do whatever you want. In this case, we trigger instant learning on the :brain: emoji instead.

from runllm import Agent, AnswerCategory, Emoji, Event, Mention, SlackListener
from runllm.decorators import entrypoint

@entrypoint(
    listeners=[
        SlackListener(
            team_id=INTERNAL_SLACK_TEAM_ID,

            # By default, we trigger on @RunLLM mentions, but we can configure the workflow
            # to also trigger on the brain emoji.
            default_trigger=[
                Mention(),
                Emoji(shortcode=":brain:", exclude_replies=False)
            ]
        )
    ],
)
def custom_learning_wf(agent: Agent, event: Event):
    # Learn when triggered by the brain emoji.
    if isinstance(event.trigger , Emoji):
        agent.learn(event.conversation)
        return

    # Otherwise, respond normally.
    answer = agent.answer(convo)
    agent.send_to_slack_thread(answer, to=event.conversation.surface)