Rapise v8.3 Mega Release - AI Autonomous Testing, AWS Bedroc

AI Assistant With AWS Bedrock Support

In the previous version of Rapise we introduced the groundbreaking AI Assistant that helps you create automated tests from manual test scenarios, generate data lists and data tables, and provide an interactive assistant to help you with everyday questions about test automation. The previous version was limited to using the OpenAI GPT4o LLM either from Microsoft Azure or OpenAI. In the latest update to Rapise, this same functionality can now be used with AWS Bedrock, utilizing the Anthropic Claude family of LLMs:

This integration with AWS Bedrock lets you connect Rapise to your AWS Bedrock account and ensures that all of your confidential data lives within your own AWS account, making use of key AWS AI safety features such as Guardrails.

Autonomous Exploratory Testing with AI Robot

The new AiRobot feature uses the Anthropic Claude 3.5 Sonnet V2 Computer Use feature to enable automated exploratory testing. This also allows execution of some manual tests directly with minimal or no conversion from a manual test case.

The AiRobot makes interactions with an application completely automatic. It understands the screen contents, moves the mouse and presses the keyboard keys, just like a human tester. In addition, the AiRobot may interact with the Web Browser, with a window or with the Desktop. It can also handle hard-to-automate environments like Citrix, Remote Desktop, and SAP.

To see the new functionality in action, please watch this introductory video.

Note that Rapise already has an AI Assistant command in RVL which look very similar to the new AiRobot functionality. The key difference here is runtime token usage. While the AI Assistant generates a code once and then uses it for all test executions without accessing the AI and spending any tokens, the AiRobot is interactive - it is spending tokens during the execution, all the time.

Advanced Visual Testing Capabilities with AI Tester

AiTester enhances the testing process by introducing advanced visual testing capabilities, simulating the keen observational skills of a manual tester. This feature enables automated detection of visual discrepancies, such as layout shifts, missing elements, or subtle changes in design, that traditional functional tests might overlook.

With AiTester, you can send text and image-augmented queries to AI models hosted by OpenAI, Azure OpenAI, and Amazon Bedrock. Example applications include generating data, performing image-based verifications (e.g., detecting discrepancies or counting visual elements), and more.

By incorporating AI-powered analysis, AiTester ensures a more thorough review of the user interface, identifying issues with the precision and adaptability of human judgment, all while maintaining the speed and consistency of automated execution.

For more details on using AiTester in your environment, and examples of the different use cases, please review KB883 - AiTester Public Module, but here are some of the most common uses:

Comparing Images

This functionality will automatically analyze the image from the original test recording with that at playback and describe any differences found. In the example below, you can see that the testing engine has found the various differences between the two pictures (number in the clouds, sun vs. butterfly in the sky, color of the umbrella, etc.)

Assess the State of an Application

The general image query lets you ask open-ended questions about the image in question. In this example, we are asking the engine if the tester has already logged in or not (spoiler alert, the answer is No):

Comparing Screenshots

The same image comparison feature we illustrated above can also be used with the specific task of comparing screenshots. In this case we are able to get discrete, concrete differences between the two screenshots. They are in fact two versions of the same application, one written in React, the other in Angular.

Perform AI-based Assertions

Finally, you can use the AI Tester to ask opinionated questions such as whether a page looks "OK" and have the AI give its reasoning as to why (or why not!). In this example, all of the UI elements are clearly visible and district from each other.

Whereas by comparison, the following screen has at least one overlapping element:

Other Enhancements

In addition to these three major new features, we have implemented the following other enhancements:

Add OCR global object [IN:8095]
Implement auto-wait for Web actions (mouse move, click, set text, etc.) [IN:9867]
Implement Appium Windows Driver support [IN:9747]
Add support for Selenium 4 Relative Locators [IN:10018]
Import manual test cases directly from Spira [IN:10116]
Add ability to run Playwright code on a connected browser instance [IN:10367]

This new version also includes other new features, improvements, bug fixes, and new Global objects. For more details, please check out the Rapise 8.3 Release Notes.