Transforming Automated Conversations Using Chat GPT

Transforming Automated Conversations Using Chat GPT

You’ve probably heard of ChatGPT, the latest breakthrough in Natural Language Processing. In just five days, it has gained over a million users and is taking the internet by storm. The free AI chatbot has become very popular quickly, even though it has only been out for a few weeks.

This is because it can give clear answers and explanations and do complicated things like write an article or tell a joke. Since ChatGPT can write test cases in different frameworks and languages, it has much to do with automated testing for mobile and web developers. Let’s explore the possibilities for Transforming Automated Conversations Using Chat GPT.

What is ChatGPT?

ChatGPT is a large language model made by OpenAI. It was fine-tuned by using supervised and reinforcement teaching on large datasets. ChatGPT looks at the data using algorithms to find patterns that assist it in understanding how words are used in natural language and context.

It is a conversation model, which means it is meant to be used like a back-and-forth chat. You can tell ChatGPT to say anything you want as long as it’s not offensive.

ChatGPT can give creative and on-topic answers, and it often gives details about why it gives a certain answer. It can also remember what it said before so it can have a conversation that makes sense.

So as per this Transforming Automated Conversations Using Chat GPT article, People have asked ChatGPT to do everything from writing folk songs regarding beer to answering questions about soil physics.

ChatGPT and Automated Testing

One of the most interesting things about ChatGPT for those who work in software is that it can make relevant code based on a simple request in natural language. It can make code in many different languages and use a lot of built-in packages in those languages.

So the obvious question is whether ChatGPT can be used to make code for automated testing. At this point, the answer is “Yes, kind of.” Nikolay Advolodkin of Sauce Labs showed in his video, ChatGPT Can Code Improve Automation Than Me, that ChatGPT can write Selenium code in more than one language.

But it’s not enough to be capable of writing code that looks right. In a perfect world, you would tell ChatGPT what test to make, it would know everything about the version of the website you are testing, and it would make perfect, executable code that doesn’t need any changes.

So based on this Transforming Automated Conversations Using Chat GPT article, ChatGPT can’t do that right now. Even so, what it is capable of is still pretty cool. Let’s start with an example of how ChatGPT could be used not as a replacement for testers but as a new low-coding method.

What exactly is Low-Code Testing?

Low-code development lets people write code even if they have little or no experience with coding. They can do this by using drag-and-drop platforms or plain English. Low-code testing solutions make it easier for development organizations to grow, which makes it easier to write test code.

Low code makes it easier for teams to start writing test automation code because tests can be written without technical skills. This reduces test debt. ChatGPT can be a powerful tool for writing test cases with little or no code.

ChatGPT works with natural language, so users can write in their style and still be understood. This differs from template-based models, which often depend on the certain grammatical structure or key phrases. As we’ll show, ChatGPT makes scripts, classes, and functions for test automation that are very good.

A Low-Code Language: Cucumber

So considering this Transforming Automated Conversations Using Chat GPT article, ChatGPT can make a lot of different languages and libraries, but its best feature is that it can make Cucumber code.

Behavior-driven development is used in Cucumber, which is a testing framework. Scenarios are written in plain English, with keywords like “Given,” “When,” and “Then” in a feature file. Then, the code is linked to these natural language phrases in step definitions.

Cucumber makes it easier to keep up with tests because its scenarios combine the natural language meaning of a test with the automation code that runs it.

This makes it easier for testers who aren’t as acquainted with test code to see the connection between the test intent (written in natural language) and the test code. This leads to ChatGPT’s strength: it can make code that sounds like natural language.

Test cases for ChatGPT and Cucumber

As the next example shows, ChatGPT’s ability to make code is impressive. Cucumber can make both the scenario and the step definitions simultaneously with a simple prompt. It knows on its own what Cucumber needs to run and doesn’t need to be told.

Even though the prompt isn’t clear about what to test on the website, ChatGPT still makes a script to test the search function, which is the most crucial part of the Google website. In this case, it correctly says that the search bar on Google is called “q.”

A generic and easy-to-use script

This is a good example of how ChatGPT can write code generally, but not every website is as popular as Google. Even though it was right about the identifier “q,” that doesn’t mean it will always be right.

So according to this Transforming Automated Conversations Using Chat GPT article, And we know from testing ChatGPT that it will make up an accurate element locator if it doesn’t know one.

It’s a lot of work to go through all of the generated code and find and replace all of the element locators, but the problem can be significantly simplified by separating the code about page from the code for the test cases.

A page object model is what we use to do this. Using a page object model lets us change the code if element locators or the app’s structure change between tests. This makes the testing script easier to keep up-to-date.

In the next example, we create a generic test for a website’s login page using Cucumber in Python and tell ChatGPT to employ a page object model and class variables as element locators.

ChatGPT correctly made the LoginPage object, which has the right name. The element locators are represented as class variables, and the object has methods for common login page tasks, like entering a username.

At the end of the example, the step definitions show how to interact with the website using the methods on the login page.

Now, using the page model object and the step definitions, it seems like a user can either automate (with ChatGPT) or manually start writing scenarios to test the webpage and have a working test code!

Changing and fixing code with ChatGPT

Looking closely, you can see that all test input values, like the URL, username, and password, are hard-coded into the step implementation.

It’s unlikely that the website you’re testing will be “http://www.example.com/login” and that “username” and “password” will work as username and password, respectively.

And these values don’t have to be hard-coded; Cucumber can get variables from the feature file and use them in the scenarios. But what if you don’t recognize how to format this or don’t want to take the time to change all the code?

You can ask ChatGPT to help you fix it. We can fix the problem by asking ChatGPT to update the code it already made.

Instead of hard-coding the most likely wrong values, the updated step implementations now read in the values we asked for.

This is one of the great things about ChatGPT: because the model is conversational, you can tell the system exactly what you wish to change in the code, and it is very good at listening to your requests and carrying them out.

Conclusion – Transforming Automated Conversations Using Chat GPT

ChatGPT is a very powerful natural language model with a lot of potentials. It can do a lot and will probably be a leader in low-code testing solutions to several problems.

It could be used for testing, but to use ChatGPT as it is now, you still need to know a fair amount about both the language and the app you are testing. But we shouldn’t be too skeptical about it.

ChatGPT is an impressive way to turn natural language into code, which was impossible with earlier models. So this concludes the topic for Transforming Automated Conversations Using Chat GPT.