Skip to content

ChatGPT on Slack's next-gen modular platform

License

Notifications You must be signed in to change notification settings

seratch/chatgpt-on-deno

Repository files navigation

ChatGPT on Slack's next-gen platform

This app contains a sample TypeScript project for use on Slack's next-generation hosted platform. The project demonstrates a simple ChatGPT app that runs in a Slack channel.

Guide Outline:


Supported Workflows

  • Configure: Configure what channels the app can respond to messages in
  • Quick Reply: Runs when a user starts a discussion with ChatGPT or asks a simple question by mentioing the app's bot user
  • Discuss: Runs when a user send a follow-up message in a discussion thread with ChatGPT

Setup

Before getting started, make sure you have a development workspace where you have permissions to install apps. If you don’t have one set up, go ahead and create one. Also, please note that the workspace requires any of the Slack paid plans.

Create Your OpenAI API Account

Also, this app needs a valid Open AI access key for ChatGPT API calls. Head to your OpenAI account page to create a new API key. We will use the token string later.

Install the Slack CLI

To use this template, you first need to install and configure the Slack CLI. Step-by-step instructions can be found in our Quickstart Guide.

Clone the Template

Start by cloning this repository:

# Clone this project onto your machine
$ slack create chatgpt-on-deno -t seratch/chatgpt-on-deno

# Change into this project directory
$ cd chatgpt-on-deno

Save Env Values

Copy the OpenAI API key string and save the value in .env file:

OPENAI_API_KEY=sk-...

When you deploy your app later, you can set the same value by running the following command:

slack env add OPENAI_API_KEY sk-...

Also, if you already have the API access to GPT-4 model, you can switch to gpt-4 by having OPENAI_MODEL=gpt-4 in the .env file for dev app and running slack env add OPENAI_MODEL gpt-4 for deployed app.

Create a link trigger for configuring your app

Triggers are what cause workflows to run. These triggers can be invoked by a user, or automatically as a response to an event within Slack.

A link trigger is a type of trigger that generates a Shortcut URL which, when posted in a channel or added as a bookmark, becomes a link. When clicked, the link trigger will run the associated workflow.

Link triggers are unique to each installed version of your app. This means that Shortcut URLs will be different across each workspace, as well as between locally run and deployed apps. When creating a trigger, you must select the Workspace that you'd like to create the trigger in. Each Workspace has a development version (denoted by (dev)), as well as a deployed version.

To create a link trigger for the workflow that enables end-users to configure the ChatGPT workflow in this template, run the following command:

$ slack trigger create --trigger-def triggers/configure_link.ts

After selecting a Workspace, the output provided will include the link trigger Shortcut URL. Copy and paste this URL into a channel as a message, or add it as a bookmark in a channel of the Workspace you selected.

Note: this link won't run the workflow until the app is either running locally or deployed! Read on to learn how to run your app locally and eventually deploy it to Slack hosting.

Running Your Project Locally

While building your app, you can see your changes propagated to your workspace in real-time with slack run. In both the CLI and in Slack, you'll know an app is the development version if the name has the string (dev) appended.

# Run app locally
$ slack run

Connected, awaiting events

Once running, click the previously created Shortcut URL associated with the (dev) version of your app. This should start the included sample workflow.

To stop running locally, press <CTRL> + C to end the process.

When you click the link trigger URL in Slack, you can configure the channel list as below:

Once the translator is added to a channel, you can mention @ChatGPT on Deno (you can rename it by editing manfest.ts) with a simple question. The app will respond to your question quickly.

If you have follow-up questions or requests to ChatGPT, you can reply to the bot's message in its thread:

Deploying Your App

Once you're done with development, you can deploy the production version of your app to Slack hosting using slack deploy. Also, please don't forget setting the OpenAI API key for the deployed app.

$ slack deploy
$ slack env add OPENAI_API_KEY (your key here)

After deploying, create a new link trigger for the production version of your app (not appended with (dev)). Once the trigger is invoked, the workflow should run just as it did in when developing locally.

Viewing Activity Logs

Activity logs for the production instance of your application can be viewed with the slack activity command:

$ slack activity

Project Structure

manifest.ts

The app manifest contains the app's configuration. This file defines attributes like app name and description.

slack.json

Used by the CLI to interact with the project's SDK dependencies. It contains script hooks that are executed by the CLI and implemented by the SDK.

/functions

Functions are reusable building blocks of automation that accept inputs, perform calculations, and provide outputs. Functions can be used independently or as steps in workflows.

/workflows

A workflow is a set of steps that are executed in order. Each step in a workflow is a function.

Workflows can be configured to run without user input or they can collect input by beginning with a form before continuing to the next step.

/triggers

Triggers determine when workflows are executed. A trigger file describes a scenario in which a workflow should be run, such as a user pressing a button or when a specific event occurs.

Resources

To learn more about developing with the CLI, you can visit the following guides:

To view all documentation and guides available, visit the Overview page.