Build and Run an Automation

This tutorial is designed to walk you through the process of building your first automation. Whereas integrations consist of predefined workflows that end-users can easily activate via a web form, automations are a powerful tool for enhancing your internal business operations. They allow you to link applications and services used within your organization or to launch bespoke business logic that meets your unique requirements.

Automation Overview

In this tutorial, we'll construct an automation designed to gather customer feedback via Typeform, relay each piece of feedback as a Slack notification, and automatically generate a GitHub issue based on the customer's response to the question: "What improvements or additional features would you like to see in future updates?". Utilizing OpenAI's ChatGPT, the generated ticket will be structured as a comprehensive user story, encompassing title, role, goal, reason, acceptance criteria, and test case.

Let's delve into the specifics of the automation flow:

Here's the first piece of customer feedback collected through our Typeform:

As a consequence, we received a notification in Slack:

Furthermore, a new GitHub Issue was created containing this content:

A Step-by-step Guide

The upcoming sections will walk you through the steps required to build the aforementioned automation from start to finish.

Select a trigger

Every automation begins with a trigger, which initiates the automation flow. Typically, triggers can capture data from external sources—like a Webhook trigger or a google.email.NewEmail trigger—or they can be based on time events, such as a Scheduler or Timer. This setup allows you to program your automations to run at predetermined times, for instance, "every day" or "each week on Monday at 2pm". A unique trigger type is the "OnStart" trigger, which activates immediately when your automation flow is started. This is particularly useful for batch operations that you intend to run only once or for debugging purposes.

To start building an automation, go to the Automations page and select "Create Automation". This will lead you to the Automation Designer. The first item you'll encounter there is the Trigger Selector. Look up "Typeform" and choose the "NewEntry" trigger.

Select a trigger in your Automation

In the Configuration Inspector on the right, authenticate with Typeform and select one of your forms:

Authenticate and configure your trigger

Add Actions

Drag the Slack module from the left panel and choose the "SendChannelMessage" action in the Inspector panel on the right. Authenticate with your Slack account and select a Slack channel to direct your notifications:

Click the "+" button next to the Slack message field to include data placeholders for each Typeform question. These placeholders will be substituted with real data once the automation executes and retrieves customer feedback from Typeform:

Similarly, add the "OpenAI.CreateChatCompletion" action. Configure the model, set the response type to json_object, and define the prompt as follows:

The goal is to have ChatGPT generate a JSON comprising two fields: "title" and "content." This JSON will then be seamlessly integrated into the "Github.CreateIssue" action. The structure of our output JSON should resemble the following:

{
    "title": CHATGPT_INFERED_USER_STORY_TITLE,
    "content": CHATGPT_INFERED_USER_STORY_DESCRIPTION_IN_MARKDOWN
}

Create new variables and transform data with Modifiers

When your automation flow includes a component that outputs data you need to alter and then utilize the modified data across various configuration fields, employing the Control.SetVariable component proves beneficial. This component enables you to define custom variables, which can subsequently be referenced by name in other linked components. This approach is particularly advantageous when you aim to use transformed data in several locations, as it obviates the need to replicate the same modifications for a single variable across multiple fields.

In our scenario, we intend to capture the output of the ChatGPT completion (ChatGPT's response) in a variable named result. This allows us to conveniently reuse this data for both the title and description fields when we're ready to create our GitHub issue.

Given our interest in only the first choice, and considering that the message content of this choice is a JSON string (as specified in our prompt to ChatGPT), we need to manipulate the Choices variable to extract our desired JSON. To transform data within Appmixer, simply click on a variable to access the Modifiers panel and then sequentially apply the necessary modifiers, akin to how formulas are utilized in an Excel sheet. In our case, we will employ the sequence of modifiers: First Item -> JSON Path -> Parse. This sequence will allow us to select the first item from the Choices list, extract the message.content from this item, and finally parse the message.content into a JSON object for later use. It is crucial to note that ChatGPT outputs a JSON string, not a JSON object, necessitating the Parse modifier to convert the text into structured data.

Finalize our Workflow Automation

The final step involves creating a GitHub Issue using the title and content (in Markdown) generated by our AI. Navigate through the interface to locate the GitHub connector, then drag and drop it onto the canvas. Select the "CreateIssue" action and assign response.title and response.content (the JSON fields generated by ChatGPT) to the corresponding fields within the "Github.CreateIssue" action. As an additional step, choose the "enhancement" label from the selection box to categorize our GitHub issues as enhancements.

Start and Monitor Your Automation

Your automation is now set up and ready to be launched. Begin by giving it a distinctive name; this can be done by double-clicking on the default "New flow" title and renaming it to something more descriptive, like "Track Customer Feature Requests". To activate your automation, click the "Start flow" button located in the top right corner. You'll notice the log panel opens at the bottom displaying the initial log entry: "Flow started". As customer feedback starts coming in through our Typeform web form, the log panel will populate with entries detailing the actions taken within the flow. These logs provide insights into each component's input, the data transmitted between components, and the output generated, offering a comprehensive view of the automation's operation:

Clicking any log entry will reveal a panel that offers additional details.

Additionally, you can visit the Insights page to access enhanced filtering options and gain improved visibility into your logs. For instance, if you're specifically interested in logs pertaining to the CreateIssue component, you can utilize the Flows selector in the left panel. This allows you to select and view logs exclusively for the component of interest.

This action will produce a filtered list of logs:

Additionally, the Time range picker allows you to choose any date interval you're interested in for viewing the logs.

Lastly, you can search for specific strings using the Search filter to find particular entries in the logs.