How to Thread Summarization with Power Automate

My flows > New flow > Automated cloud flow

Create automated cloud flow

Navigate to make.powerautomate.com and sign in with your Microsoft account. Click 'My flows' in the left sidebar, then 'New flow' at the top. Select 'Automated cloud flow' from the dropdown. Name your flow 'Thread Summarizer' and search for 'Slack' in the trigger list.

1. 1Click 'My flows' in the left navigation
2. 2Select 'New flow' button
3. 3Choose 'Automated cloud flow'
4. 4Type 'Thread Summarizer' as flow name
5. 5Search for 'Slack' in trigger search box

Triggers > Slack > When a reaction is added to a message

Configure Slack reaction trigger

Select 'When a reaction is added to a message' from the Slack triggers. You'll need to create a new connection to Slack by clicking 'Sign in' and authorizing Power Automate access. Choose your workspace from the dropdown. Set the Channel to 'Any channel' and Reaction to '📝' (memo emoji).

1. 1Click 'When a reaction is added to a message'
2. 2Click 'Sign in' and authorize Slack access
3. 3Select your Slack workspace from dropdown
4. 4Set Channel field to 'Any channel'
5. 5Enter '📝' in the Reaction field

New step > Slack > Get messages from a channel

Add thread retrieval step

Click 'New step' below the trigger. Search for 'Slack' and select 'Get messages from a channel'. This action will fetch all messages in the thread. Set the Channel to the dynamic content 'Channel' from the trigger. Set the Message timestamp to 'Thread timestamp' from the trigger dynamic content.

1. 1Click 'New step' button
2. 2Search for 'Slack' in the connector search
3. 3Select 'Get messages from a channel'
4. 4Click in Channel field and select 'Channel' from dynamic content
5. 5Set Message timestamp to 'Thread timestamp'

New step > Data Operation > Compose

Parse thread messages

Add another step and search for 'Data Operation'. Select 'Compose' to format the thread messages for OpenAI. In the Inputs field, you'll build an expression to combine all message text. Click the expression tab and enter a formula to concatenate the message texts with sender names.

1. 1Click 'New step' and search 'Data Operation'
2. 2Select 'Compose' action
3. 3Click in the Inputs field
4. 4Select 'Expression' tab
5. 5Enter the message parsing expression

New step > HTTP

Connect to OpenAI

Add a new step and search for 'HTTP' since Power Automate doesn't have a native OpenAI connector. Select 'HTTP' action and configure it as a POST request to https://api.openai.com/v1/chat/completions. You'll need to add your OpenAI API key in the headers and format the request body as JSON.

1. 1Click 'New step' and search for 'HTTP'
2. 2Select 'HTTP' action
3. 3Set Method to 'POST'
4. 4Enter 'https://api.openai.com/v1/chat/completions' as URI
5. 5Add Authorization header with your API key

HTTP > Body

Configure OpenAI request

In the HTTP step's Body field, create a JSON payload for the OpenAI API. Include the model (gpt-3.5-turbo), messages array with system and user prompts, and parameters like max_tokens. Use dynamic content from the Compose step as the user message containing the thread text.

1. 1Click in the Body field of HTTP step
2. 2Select 'Expression' tab
3. 3Build JSON object with model and messages
4. 4Set system prompt for summarization
5. 5Use Compose output as user content

New step > Data Operation > Parse JSON

Parse OpenAI response

Add a 'Parse JSON' step from Data Operations. In the Content field, select 'body' from the HTTP response dynamic content. You'll need to provide a schema that matches OpenAI's response format. The key field is choices[0].message.content which contains the summary.

1. 1Add 'Parse JSON' step
2. 2Set Content to HTTP response body
3. 3Click 'Generate from sample' for schema
4. 4Paste a sample OpenAI API response
5. 5Verify the schema includes choices array

New step > Slack > Post message to a channel

Post summary to Slack

Add a final Slack step 'Post message to a channel'. Use the same Channel from the trigger. In the message text, format the summary with context like 'Thread Summary:' followed by the parsed OpenAI content. Set Thread timestamp to reply in the same thread.

1. 1Search for Slack connector
2. 2Select 'Post message to a channel'
3. 3Set Channel to trigger's Channel value
4. 4Format message with summary prefix
5. 5Set Thread timestamp from trigger

Save > Test > My flows > Run history

Test the flow

Save your flow and test it by going to any Slack channel and reacting to a message with the 📝 emoji. Check the Flow run history in Power Automate to see if all steps executed successfully. Look for any errors in the HTTP request to OpenAI or the JSON parsing steps.

1. 1Click 'Save' at the top right
2. 2Go to Slack and add 📝 reaction to any message
3. 3Return to Power Automate
4. 4Click 'Run history' to view execution
5. 5Check each step for green success indicators

Add this expression to the Compose step to format thread messages with usernames and clean up Slack's formatting artifacts.

join(
  map(
    body('Get_messages_from_a_channel')?['messages'],

join(
  map(
    body('Get_messages_from_a_channel')?['messages'],
    concat(
      item()?['user'], ': ',
      replace(
        replace(
          replace(item()?['text'], '<@', '@'),
          '>', ''
        ),
        '&lt;', '<'
      )
    )
  ),
  '\n'
)