Tutorial: Send Slack message

Preview

This feature is in Public Preview.

In this tutorial, you create a SQL UDF operator for Lakeflow Designer that posts messages to Slack channels. SQL UDFs are the right choice when a function needs to call external APIs over HTTP. For a broader overview, see User-defined operators in Lakeflow Designer.

Overview

This operator sends messages to Slack using:

• SQL UDF: Written in SQL rather than Python.
• Unity Catalog HTTP connection: Securely manages Slack API credentials.
• Preview mode support: Prevents actual Slack API calls during workflow preview.
• Expression parameters: Allows dynamic message content from DataFrame columns.

Why SQL UDF?

For operators that need to call external APIs (like Slack, REST endpoints, webhooks), you must use SQL UDFs. Python UDFs and UDTFs cannot make HTTP requests. SQL UDFs have access to the http_request() function which works with Unity Catalog connections.

Step 1: Set up the Unity Catalog HTTP connection

Before creating the UDF, you need to set up a Unity Catalog HTTP connection to securely store your Slack API credentials. Replace <xoxb-your-slack-bot-token> with your actual Slack Bot Token. You can get this from your Slack app settings. You can use this same connection across multiple UDFs. To learn more, see Connect to external HTTP services.

SQL

-- Create a connection to store Slack credentials securely
CREATE CONNECTION my_slack_connection TYPE HTTP OPTIONS (
  host 'https://slack.com',
  port '443',
  base_path '/api/',
  bearer_token '<xoxb-your-slack-bot-token>'
);

Step 2: Create the operator YAML

Now you will create the YAML for the operator. For details about the schema, see User-defined operator YAML reference.

The YAML for this operator includes:

• Expression parameter (msg): Allows dynamic message content from dataframe columns.
• String parameter (channel): Static channel name/ID.
• Preview mode (is_preview): A config property with format: is_preview that enables preview mode to prevent actual API calls during testing.

YAML

schema: user-defined-operator-v0.1.0
type: uc-udf
name: Send Slack Message
id: send_msg
version: '1.0.0'
description: Send Slack Message to a Channel
config:
  type: object
  properties:
    msg:
      type: string
      format: expression
      title: Message
      examples:
        - 'Select message column or expression'
      x-ui:
        widget: expression
        port: input_data
    channel:
      type: string
      title: Channel
    is_preview:
      type: boolean
      format: is_preview
      default: false
  required:
    - msg
    - channel
  additionalProperties: false
ports:
  input:
    - name: input_data
      title: Input Data
  output:
    - name: output
      title: Send Response Data

This includes:

Config Key	Widget	Purpose
msg	expression	Dynamic message content from input data.
channel	text	Slack channel to send to (e.g., #alerts).
is_preview	n/a	A boolean config property with format: is_preview that allows the operator to behave differently during a preview (in this case, avoid actually creating a Slack message).

Step 3: Create the Unity Catalog function

When creating SQL UDFs, there are a few things that are uncommon compared to most SQL queries:

• Use the RETURN syntax instead of AS $$.
• Embed the YAML configuration in a SQL comment block (/* ... */).
• Can use the http_request function for API calls.

SQL

CREATE OR REPLACE FUNCTION main.my_schema.send_slack_msg(
    msg STRING,
    channel STRING,
    is_preview BOOLEAN
)
RETURNS STRING
RETURN (/*
    schema: user-defined-operator-v0.1.0
    type: uc-udf
    name: Send Slack Message
    id: send_msg
    version: "1.0.0"
    description: Send Slack Message to a Channel
    config:
      type: object
      properties:
        msg:
          type: string
          format: expression
          title: Message
          examples:
            - "Select message column or expression"
          x-ui:
            widget: expression
            port: input_data
        channel:
          type: string
          title: Channel
        is_preview:
          type: boolean
          format: is_preview
          default: false
      required:
        - msg
        - channel
      additionalProperties: false
    ports:
      input:
        - name: input_data
          title: Input Data
      output:
        - name: output
          title: Send Response Data
    */
  CASE
    WHEN NOT is_preview THEN
      http_request(
        conn => 'my_slack_connection',
        method => 'POST',
        path => 'chat.postMessage',
        json => to_json(named_struct('channel', channel, 'text', msg)),
        headers => map('Content-Type', 'application/json;charset=utf-8')
      ).text
    ELSE 'Preview mode - no message sent to ' || channel
  END
);

This SQL function includes the following features:

Feature	Purpose
http_request()	Makes HTTP calls to external APIs.
conn => 'my_slack_connection'	References the UC connection for authentication.
to_json() and named_struct()	Constructs the JSON payload for the Slack API.
YAML comment block	Used by Lakeflow Designer to create the operator.
CASE WHEN	Implements preview mode logic.

Step 4: Test the function

Next, test the function to make sure it works before registering it as an operator.

Test in preview mode first, to avoid sending a Slack message:

SQL

-- Test in preview mode (won't send real message)
SELECT main.my_schema.send_slack_msg(
  'Hello from Lakeflow Designer!',
  '#test-channel',
  true  -- is_preview = true
) AS result;

-- Expected result: "Preview mode - no message sent to #test-channel"

Test with external API call (will send a message to Slack):

SQL

-- Test with real API call (USE WITH CAUTION!)
SELECT main.my_schema.send_slack_msg(
  'Hello from Lakeflow Designer!',
  '#test-channel',
  false  -- is_preview = false
) AS result;

-- Expected: Slack API response JSON

Step 5: Register the operator

Add the operator to your .user_defined_operators.yaml file:

YAML

operators:
  - catalog: main
    schema: my_schema
    functionName: send_slack_msg

note

If you define this file in your user folder, it only appears for you. For more information, see Make your operator discoverable.

Step 6: Set up permissions

For SQL UDFs that use Unity Catalog connections, users need an additional permission:

SQL

-- Schema and function access
GRANT USE SCHEMA ON SCHEMA main.my_schema TO `<user>`;
GRANT EXECUTE ON FUNCTION main.my_schema.send_slack_msg TO `<user>`;

-- Connection access (required for API calls)
GRANT USE CONNECTION ON CONNECTION my_slack_connection TO `<user>`;

important

Without the USE CONNECTION permission, users won't be able to make API calls even if they can execute the function.

Using the operator in Lakeflow Designer

After it's registered, the operator will appear in Lakeflow Designer with:

• An input port to connect your data source.
• An expression picker to select which column contains the message content.
• A text input for the Slack channel.

Users can send notifications based on their data — for example, alerting when certain thresholds are exceeded.

Common use cases

• Alerts — Send notifications when data quality issues are detected.
• Notifications — Notify teams when workflows complete.
• Webhooks — Call external APIs to trigger downstream processes.
• Logging — Send audit messages to external systems.

Best practices for building API-calling operators

1. Always use preview mode — Add an is_preview config property with format: is_preview to prevent accidental API calls.
2. Use Unity Catalog connections — Never hardcode credentials in your UDF. Unity Catalog connections are only available in SQL UDFs.
3. Handle errors gracefully — API calls can fail; consider what to return on error.
4. Test thoroughly — Use preview mode during development.
5. Document the connection setup — Users need to know what connection to create.