How to Clean Up Meeting Notes with Pipedream

Dashboard > New Workflow

Create New Workflow

Go to pipedream.com and click New Workflow in the top right. You'll land on the workflow builder with an empty canvas. The interface shows Sources on the left for triggers and Steps below for actions.

1. 1Click the blue 'New Workflow' button
2. 2Select 'Start from scratch' option
3. 3Name your workflow 'Meeting Notes Cleanup'

Workflow > Add Trigger > Slack

Add Slack Message Trigger

Click Add a trigger and select Slack. Choose the New Message trigger type. This fires instantly when someone posts in your designated channel. You'll need to connect your Slack workspace and select the specific channel.

1. 1Click 'Add a trigger'
2. 2Search for 'Slack' in the apps list
3. 3Select 'New Message' from trigger options
4. 4Connect your Slack account

Trigger Configuration > Channel Selection

Configure Channel and Filter

Select your meeting notes channel from the dropdown. Add a filter condition to only process messages that start with a specific keyword like 'notes:' or contain more than 50 characters. This prevents the workflow from running on short messages or reactions.

1. 1Choose your target channel from the dropdown
2. 2Set 'Message Length' condition to greater than 50 characters
3. 3Add text filter for messages starting with 'Meeting:' or 'Notes:'

Workflow > Add Step > Code

Add Node.js Code Step

Click Add Step below the trigger and select Code. Choose Node.js runtime. This step will extract the message text and prepare it for OpenAI formatting. You'll write code to clean up the raw text and remove any Slack formatting artifacts.

1. 1Click the '+' button to add a step
2. 2Select 'Code' from the step types
3. 3Choose 'Run Node.js code'
4. 4Name the step 'Format Message Text'

This Node.js code cleans up Slack message formatting artifacts and extracts key meeting details before sending to OpenAI. Paste it in the code step between Slack trigger and OpenAI.

export default defineComponent({
  async run({ steps, $ }) {
    const message = steps.trigger.event;

export default defineComponent({
  async run({ steps, $ }) {
    const message = steps.trigger.event;
    
    // Clean up Slack formatting
    let cleanText = message.text
      .replace(/<@U[A-Z0-9]+>/g, '') // Remove user mentions
      .replace(/<#C[A-Z0-9]+\|[^>]+>/g, '') // Remove channel links
      .replace(/[\s\S]*?/g, '') // Remove code blocks
      .replace(/`([^`]+)`/g, '$1') // Remove inline code
      .replace(/\s+/g, ' ') // Collapse whitespace
      .trim();
    
    // Extract meeting metadata
    const meetingMatch = cleanText.match(/^Meeting:\s*(.+?)\./i);
    const meetingTitle = meetingMatch ? meetingMatch[1] : 'Team Meeting';
    
    // Remove common filler words for cleaner AI processing
    const fillerWords = /\b(um|uh|like|you know|basically|actually)\b/gi;
    cleanText = cleanText.replace(fillerWords, '');
    
    return {
      cleaned_text: cleanText,
      meeting_title: meetingTitle,
      original_user: message.user,
      message_length: cleanText.length,
      timestamp: message.ts
    };
  }
});

Workflow > Add Step > OpenAI

Add OpenAI Chat Completion Step

Add another step and search for OpenAI. Select the Chat Completions action. This sends your cleaned meeting notes to GPT-4 with specific formatting instructions. You'll configure the model, temperature, and system prompt for consistent output.

1. 1Click '+' to add another step
2. 2Search for 'OpenAI' in the apps list
3. 3Select 'Chat Completions' action
4. 4Connect your OpenAI account with API key

OpenAI Configuration > Messages

Configure AI Prompt

Set up the system message to format notes consistently. Use a temperature of 0.3 for predictable output. The prompt should specify exact formatting rules: bullet points for key discussion items, numbered action items with owners, and a next steps section.

1. 1Set Model to 'gpt-4-turbo-preview'
2. 2Set Temperature to 0.3
3. 3Add system message with formatting instructions
4. 4Reference the cleaned text from step 4 in user message

Workflow > Add Step > Slack > Send Message

Add Slack Reply Step

Add a final Slack step to post the formatted notes back to the channel. Select Send Message and configure it to reply in the same thread as the original message. This keeps conversations organized and shows the before/after clearly.

1. 1Add another Slack step
2. 2Choose 'Send Message' action
3. 3Set Channel to same as trigger channel
4. 4Set Thread TS to original message timestamp

Send Message > Message Configuration

Map OpenAI Response to Message

In the message content field, reference the OpenAI response text. Add a prefix like 'Formatted Notes:' and include the AI-generated content. Test the mapping by expanding the data panel to see the exact field path from the OpenAI step.

1. 1Click in the Message Text field
2. 2Add prefix text 'Formatted Meeting Notes:'
3. 3Reference the OpenAI completion content
4. 4Add emoji or formatting for visual separation

Workflow > Test

Test the Workflow

Click Test in the top right to run the workflow with sample data. If you don't have test data, go to your Slack channel and post a sample meeting note to trigger the workflow. Check each step's output to verify data flows correctly.

1. 1Click the green 'Test' button
2. 2Check trigger data appears correctly
3. 3Verify code step cleans text properly
4. 4Confirm OpenAI returns formatted notes

Workflow > Deploy

Deploy Workflow

Click Deploy to make the workflow live. It now runs automatically when new messages appear in your configured channel. The workflow processes messages in real-time with typical response times under 10 seconds for standard meeting notes.

1. 1Click the blue 'Deploy' button
2. 2Confirm deployment in the popup
3. 3Test with a real message in Slack
4. 4Monitor the execution log for any errors