Studio Web User Guide

DELIVERY:

Last updated Aug 7, 2025

Building your first API workflow

The following tutorial provides hands-on experience with key concepts behind the API workflow interface through a simple example using the Petstore public API.

In this example, you request "Pet information" from a language model (LLM) and use that data to add pets to the Swagger Petstore. The Swagger Petstore (https://petstore.swagger.io) offers accessible and easy-to-use API endpoints, available with or without authentication.

Step 1: Calling the LLM

On your API workflow designer canvas, select Add (the plus + icon).
Select Connector > UiPath GenAI Activities > Content Generation.
Set up or select an existing connection in the Properties panel.
Configure these properties:
- Model—gpt-4o-mini-2024-07-18
- Prompt
  
  "For the swagger Petstore, create the details of a new fantasy creature that can be added to their inventory. Respond only with valid json. Return an array of 5 of these."
Test your workflow up until this point.
See the results in the Output panel. You should see the raw input and output of the activity call. The response should look similar to the following:
```
{
  "cacheReadInputTokens": 0,
  "created": 1745444601,
  "usage": {
    "total_tokens": 741,
    "completion_tokens": 686,
    "prompt_tokens": 55,
    "cache_read_input_tokens": 0
  },
  "contextGroundingCitationsString": "[]",
  "totalTokens": 741,
  "promptTokens": 55,
  "model": "gpt-4o-mini-2024-07-18",
  "id": "chatcmpl-BPcADRRpy7ZDZpBOxp6XJYk0HOpaa",
  "text": "```json\n[\n    {\n        ....  \"A stealthy creature that blends into the shadows, highly elusive.\"\n    }\n]\n```",
  "choices": [
    {
      "index": 0,
      "finish_reason": "stop",
      "message": {
        "content": "```json\n[\n  ...ws, highly elusive.\"\n    }\n]\n```",
        "role": "assistant"
      }
    }
  ],
  "completionTokens": 686,
  "object": "chat.completion"
}{
  "cacheReadInputTokens": 0,
  "created": 1745444601,
  "usage": {
    "total_tokens": 741,
    "completion_tokens": 686,
    "prompt_tokens": 55,
    "cache_read_input_tokens": 0
  },
  "contextGroundingCitationsString": "[]",
  "totalTokens": 741,
  "promptTokens": 55,
  "model": "gpt-4o-mini-2024-07-18",
  "id": "chatcmpl-BPcADRRpy7ZDZpBOxp6XJYk0HOpaa",
  "text": "```json\n[\n    {\n        ....  \"A stealthy creature that blends into the shadows, highly elusive.\"\n    }\n]\n```",
  "choices": [
    {
      "index": 0,
      "finish_reason": "stop",
      "message": {
        "content": "```json\n[\n  ...ws, highly elusive.\"\n    }\n]\n```",
        "role": "assistant"
      }
    }
  ],
  "completionTokens": 686,
  "object": "chat.completion"
}
```

Step 2: Using Script to properly format the response

The information you need sits within the content.text property, which is not properly formatted.

To your current API workflow, add the Script activity.

Open the Expression editor and write the following:

const cleanedJsonStr = $context.outputs.v2_sub_generateChatCompletion_1.content.text
  .replace(/^```json\n/, '')
  .replace(/\n```$/, '');

// Step 2: Parse into JSON
let parsedObj;
parsedObj = JSON.parse(cleanedJsonStr);
return { aipet: parsedObj };const cleanedJsonStr = $context.outputs.v2_sub_generateChatCompletion_1.content.text
  .replace(/^```json\n/, '')
  .replace(/\n```$/, '');

// Step 2: Parse into JSON
let parsedObj;
parsedObj = JSON.parse(cleanedJsonStr);
return { aipet: parsedObj };

This JavaScript code parses the content.text object and returns it in a clean format.

Test your workflow again. Observe the properly formatted response.

Step 3: Iterating over the response array

The LLM returned multiple pet examples as an array, as indicated in the prompt at Step 1.

To your current API workflow, add the Loop > ForEach activity.
Configure the For Each activity as follows:
- In—
```
$context.outputs.Javascript_1.aipet$context.outputs.Javascript_1.aipet
```
- Item name—currentItem
- Accumulate results—On
This command iterates through every item of the response array.

Step 4: Adding the returned response to Petstore

Inside the For Each activity body, add the HTTP activity, and configure it as follows:

Method—POST
Request URL—https://petstore.swagger.io/v2/pet

Request body—Open the Expression editor and prompt the Autopilot field with: "Within this foreach, transform each object so it can be posted to the Swagger Petstore Pet creation. Find the values for each property in the previous step."

The Autopilot response should look like the following (you can also copy this snippet):

{
  id: $currentItem.id,
  name: $currentItem.name,
  category: $currentItem.category,
  photoUrls: $currentItem.photoUrls,
  tags: $currentItem.tags,
  status: $currentItem.status,
  age: $currentItem.age,
  properties: $currentItem.properties
} {
  id: $currentItem.id,
  name: $currentItem.name,
  category: $currentItem.category,
  photoUrls: $currentItem.photoUrls,
  tags: $currentItem.tags,
  status: $currentItem.status,
  age: $currentItem.age,
  properties: $currentItem.properties
}

Test your workflow. At this point, your API workflow should return a Successful status. This means that the pet data has been posted correctly to Pet store.

Step 5: Returning a workflow response

This step exposes the final results of the workflow to external consumers in a clean, simplified format.

To your current API workflow, add the Response activity, and configure it as follows:

Type—Success

Details—Open the Expression editor and write the following:

$context.outputs.For_Each_1.results.map(result => ({
  id: result.content.id,
  name: result.content.name,
  description: result.content.description
}))$context.outputs.For_Each_1.results.map(result => ({
  id: result.content.id,
  name: result.content.name,
  description: result.content.description
}))

This snippet returns a custom JSON with the mentioned details.

Test your workflow. Notice the final response with the three details.

Step 6: Defining Input and Output schemas

This step makes the workflow objects available to external consumers.

For your current API workflow, open the Data manager panel.
For the Input tab:
- Add a new property and named it "Genre".
- Set the type to String.
- Mark it as required.
For the Output tab:
- Add three properties, and name them "id", "name", and "type". These are the properties returned by the workflow.
- Set their type to String.
Select the Content Generation activity in your workflow.
1. Update the Prompt field to:
  
  "For the swagger Petstore, create the details of a new " + $workflow.input.Genre + " creature that can be added to their inventory. Respond only with a json object holding the pet information."
  
  This new prompt uses the $workflow.input.Genre property defined in the Input schema.
Define a Test configuration and provide a value for the Genre property:
```
{ Genre: "Fantasy" }{ Genre: "Fantasy" }
```

Step 7: Publishing and executing

You have reached the end of a successful workflow build.

Publish the workflow to your Personal Workspace folder in Orchestrator.
Navigate to the subfolder where the corresponding process was created and select Start job.

Orchestrator reads the workflow input schema and requires you to input a Genre. Once you provide a value, the job starts.

On this page