Skip to main content

New features designed for Slack apps sending AI responses

We've introduced a new suite of features to help Slack apps provide an end-user experience typical of LLM tools:

  • Slack apps can now stream in their responses to the end user using three new API methods.
  • There are new Block Kit components to allow end users to quickly interact with AI responses.
  • The Node and Python SDKs have new utilities to streamline integration of these new features.

Read on for more info!

Streaming messages​

Three Web API methods work together to enable Slack apps to provide end users a text streaming experience similar to typical LLM tools:

Interacting with AI responses​

Three new Block Kit components allow more user interaction with agent responses:

  • The feedback_buttons block element allows users to provide feedback on whether an AI response was positive or negative.

  • The icon_button provides a way for users to quickly trigger actions like deleting AI responses.

  • The context_actions block provides a container for these interactive elements.

Supported by Slack SDKs and Bolt framework​

The Python Slack SDK and Node Slack SDK each fully support these new features — with a new streamer helper utility to aid developers in tying them all together.

Here's a simplified example in Python:

streamer = client.chat_stream(
    channel=channel_id,
    recipient_team_id=team_id,
    recipient_user_id=user_id,
    thread_ts=thread_ts,
)

for event in returned_message:
    streamer.append(markdown_text=f"{chunk-received-from-llm}")

streamer.stop(blocks=feedback_block)

This utility can also be used in the Bolt frameworks based upon these SDKs. Read more in the Using AI in apps section of your preferred Bolt flavor: Python or JavaScript.

We've also updated our app agent template repos to use these new features. Pick your flavor of Bolt and give it a try.

Get started with the Bolt for Python agent template:

slack create --template slack-samples/bolt-python-assistant-template
slack run

Get started with the Bolt for JavaScript agent template:

slack create --template slack-samples/bolt-js-assistant-template
slack run

Read all the details in the Bolt Python release notes and Bolt JS release notes.