🟣 Getting started with OpenAI GPT-3 and Make

Hey Makers :wave:

I cannot wait to share this one with you.
Today, we’ll cover the basics of getting started with OpenAI GPT-3 on Make.

Let’s dive right in!


:fire: Chat GPT-3 has been taking over the internet recently! :fire:

This is a free chatbot powered by OpenAI, which does a great job intelligently answering any question you ask. What’s more interesting is that the technology behind this chatbot, GPT-3, can be accessed via the OpenAI API!

If you are familiar with Make and the power that APIs can bring to the world of process automation, then I’m sure you are already excited!

In this article, you’ll learn how to

  • sign up for an OpenAI account
  • generate an API key
  • use the API key to establish a connection with Make’s OpenAI GT3 module

We will then walk you through an example scenario you can set up to quickly generate SEO keyword suggestions for a list of article topics.

Let’s get started!

:white_check_mark: Creating an OpenAI account :white_check_mark:


  • Navigate over here and click sign up at the top right corner to create a new account.
  • In order to be able to use the API you need to set up a paid account by visiting the Billing → Overview page.
  • At the time of writing this article, OpenAI gives you $18 in free credits so you can experiment. If you need more credits, you can check the pricing.

Calculating the cost of each API call can be a little complex. It’s easy to lose control and end up with a high bill.

Therefore, we recommend that you set up both a hard and a soft Usage Limit under Billing → Usage Limit, before you start experimenting with GPT-3.



:make: The Make GPT-3 app :make:


Once you have your OpenAI account created, you can start using the Make GPT-3 app to bring the power of AI to your scenarios.

Currently, the app offers 4 modules that can be used in a range of different use cases such as

  • generating blog post ideas
  • moderating your community comments
  • generating super-realistic images of anything you can imagine.


However, there is a little caveat…

In order to get the best results out of GPT-3, you need to provide prompts that are very clear and specific. After all, you are talking to a computer :robot: .

A great way to get prompt ideas is by spending some time on sites like FlowGPT. Where you can find examples of prompts people are using to achieve useful results.

Our suggestion is also to practice some of your prompts on the OpenAI Playground. Simply enter your prompt and then click ‘Submit’ to generate a response.


It is also useful to experiment with some of the settings on the right sidebar.

For example

  • ‘Maximum Length’ allows you to choose the length of the response GTP-3 will give to your prompt.
  • ‘Model’ allows you to choose the language model that will be used to generate your response (‘text-davinci-003’ is the most advanced but also the most expensive).


Once you find a prompt that consistently generates the quality of results you are looking for, you can transfer it over to your Make scenario.

Configuring a GPT-3 module in Make

The first step is to establish a connection with your OpenAI account. This is pretty straightforward.

Simply click on the Add button to create a new connection and then navigate to your OpenAI account to create an API key under API Keys.



Click ‘Create new secret key’.


Once your API key is created you can paste it into the ‘API Key’ field of your Make app connection.

You will also need your Organization ID which can be accessed under the Settings page of your OpenAI account.


With these credentials, your Make app should now be able to establish a connection with your OpenAI account so you can start using the API.

Please note - if you do not set up a paid plan with OpenAI, you will not be able to use the API and Make will throw an error.



:brain: Generating SEO keywords with Make and GPT-3 :brain:


If you are a content writer or SEO specialist, you know how time-consuming it can be to come up with SEO keyword ideas for different topics.

GPT-3 makes this easy. All we have to do is ask… nicely :slightly_smiling_face:

To test this out, we’ve built a Google Sheet with a list of article topics that we want to generate keywords for.


Now, with the help of Make and the GPT-3 app, we will generate keywords for each one of them.


Here’s a simple scenario for this use case

The Google Sheets > Watch Rows module will retrieve all the new rows from the spreadsheet.

Then for each row, we’re using the OpenAI GPT-3 > Create a Completion module to generate the SEO keywords.

The prompt we’re using is very simple:

give me some keywords to use for SEO for a new saas platform targeted to {{article topic}}

The article topic will dynamically be populated for each spreadsheet row by retrieving the value in the ‘Article Topic’ column.

The ‘Max Tokens’ value is set to 100. This will allocate enough tokens for GPT-3 to generate about 10 keywords. If you need more, you can increase this number. However, you need to be cautious as this will result in higher usage costs.


Once the keywords are generated, they will be added to the SEO Keyword Ideas column of the Google Sheet.

Let’s execute this scenario to see the results.

As you can see in the screenshot below, 10 keywords have been generated for each topic.


These keywords might be a little competitive though…

Let’s change the prompt and ask GPT-3 to generate some long-tail keywords instead.


Here are the results this time:


Now, this is something we could easily work with, right?

As you can imagine, this is just scratching the surface of what can be achieved with GPT-3 and Make scenarios.


:nerd_face: We can’t wait to see what you come up with! :nerd_face:

If you have any questions or any ideas on useful content we can create to help you get started with GPT-3, please let us know in the comments :arrow_down:

9 Likes

Thanks for this guide!
I have a problem with image generation via OpenAI module.
After one day all images start to show the message: "access denied. error 403 "


Heya @ECORN welcome to the community :wave:

Most likely the reason this is happening is that the links that Chat GPT-3 generates for the images, expire after some time.

The solution will be to download the images from Chat GPT-3 as soon as they are created and store them in cloud storage such as Google Drive or Dropbox. To automatically download an image from a URL in Make you can use the HTTP > Get a file module.

Hope this helps

1 Like

@Michaela I tried to use Make’s GPT-3 module to summarize blogposts, but the summary it returns is for something completely different.

I made it summarize a blogpost about Wordpress and the summary was about “why diabetes is such a prevalent problem in the US” ?! :upside_down_face:

And this happens every single time. But when I try out the prompt in the Open AI playground it works like expected.

Yesterday I used the GPT-3 edit tool in a Make integration to fix grammar and spelling mistakes => It replaced the text with something completely unrelated (Some type of C Programming snippet :hushed:)

Does anybody know why this is happening?

The only part of the module that worked for me was the Dall-E image generation…

This is a great use-case for GPT in Google Sheets. However, if you want a little more straight forward process, you can always use SheetGPT to accomplish the same thing with a free Google Sheets plugin.

I understand that’s not the point of this article, but wanted to throw it out there anyway!

1 Like

A post was split to a new topic: Using Make to post from RSS to Twitter; ChatGPT results are always just 1 sentence

Great post. This is awesome.

1 Like

Stoked to hear you like the post @PJLai Hope you’ll have tons of fun working with ChatGPT on Make :sunglasses:

A post was split to a new topic: ChatGPT and Google Sheets