Integration Service user guide

In the following example, we build a connector for OpenAI DALL-E from a blank canvas, using Personal Access Token authentication.

This example shows how to use OpenAI model DALL-E to generate an image based on a text prompt.

First, let's see how generating an image through API works.

With DALL-E, you supply a prompt, the model automatically rewrites it for safety reasons, and adds more details. Detailed prompts generally result in higher quality output. Then, the model outputs the generated image either through a link ("responseformat": "URL") or a Base64 string ("responseformat": "b64_json"). The URL is the default value, per the API specification. For more details on the API parameters, refer to API Reference - OpenAI API.

Here is an example cURL to generate an image using this API:

There are three key things you need to be aware of in building the first API resource for your connector:

Dissecting the previous cURL, you can observe the following components:

The Base URL provides the location or endpoint of the API resource for which the calls are to be made.

The Authentication type tells the connector how the user is authenticated before being permitted to utilize the API resource. In this example, we are using Personal Access Token as a Bearer Token in the Authorization header, using the OpenAI API key.

The API routes – the operation and resource – provide the path to a specific resource available on the API, for example /generations. The resource is appended to the base URL, in which that resource can be fetched/listed (GET), added to/inserted (POST ), updated/upserted (PATCH/PUT), or deleted (DELETE).

You can add as many API resources and operations as you want, based on the activities that you need for your workflow. For example, you can generate images (POST /generations), list all the generated images with GET /generations (if the GET operation is available for the resource), or edit an existing image (POST /edits).

Before you publish and use the connector in your workflow, test it first.

Now that you are able to test your connector, you can configure the design of the workflow component for this specific activity, through Activity designer.

Activity Designer helps curate input and output fields that you want to expose to users for this activity. This represents the "look and feel" when the API connector is used in UiPath Studio or Studio Web. For more information, refer to Activity designer.

You can change the display name of fields, mark input fields as required or not, whether they are a primary or secondary parameter etc. The input fields are auto-generated from the request fields or body once you perform a Send Request action.

As for the response or output fields, you can update the response root key so you can curate specific output attributes from the API response, like revised_prompt or URL, as captured in the following image.

Next, publish your connector in your tenant of choice. For more information, refer to Publishing your connector. Then you can use it in your Studio Desktop or Studio Web workflows.