Building a Slack AI Assistant

Overview

This guide provides a step-by-step process to build a Slack AI assistant using tools like Make, Slack, ChatGPT, and Airtable. The assistant can answer common, complex questions and is integrated with Slack and a vector store database for data storage.

Key Tools and Setup

• Slack Workspace: Create channels for client queries.
• Make (Integromat): Connects Slack with the assistant.
• Airtable: Database to track messages and responses.
• ChatGPT Assistants: Develop and train based on uploaded content in vector store databases.

Implementation Steps

1. Setup Airtable Database

• Create an Airtable base from scratch.
• Delete default columns and add required columns:
  • Message ID (auto-number)
  • Message, User, Type, Timestamp (single line text)
  • Thread Timestamp, Client Message ID, Team, Channel, Channel Type
  • Assistant Thread ID, Assistant Response (long text)

2. Connect Make with Slack

• Webhook Creation:
  • Go to Make, create a Webhook, name it, and save.
  • In Slack API, create a new app, turn on event subscriptions, and connect with Make.
• Subscribe to Bot Events:
  • Add bot user events for message history in private channels.

3. Create Slack App

• Install the app in the Slack workspace.
• Add the app to a Slack channel.

4. Setup OpenAI for ChatGPT

• Create Project in OpenAI:
  • Generate API keys and create a ChatGPT assistant.
• Upload Content:
  • Add files and guides that the assistant should use for responses.

5. Build Make Automation

• Search Airtable for Threads:
  • Use search records to find existing threads based on thread timestamp.
• Create a New Record:
  • Record new messages in Airtable.
• Respond to User:
  • Send initial acknowledgment message in Slack.
• Integrate ChatGPT Assistant:
  • Use the assistant to generate responses to messages.
• Update Airtable:
  • Log assistant responses and update threads in Airtable.

6. Advanced Setup

• Filter Messages:
  • Prevent infinite loops by filtering bot messages.
• Error Handling:
  • Add error handlers in Make to manage bot errors gracefully.
• Multiple Assistants:
  • Route messages to different assistants based on Slack channels.

Community Resources

• No Code Architects Community:
  • Access templates and support.
  • Participate in daily calls for further guidance.

Conclusion

Following these steps allows you to set up a functional AI assistant within Slack, automated with Make, and supported by a robust database in Airtable. This setup can handle specific, niche tasks efficiently, enhancing support within your community.