AI-powered no-code testing platform

01The context

QA teams spent large amounts of time writing and repairing end-to-end tests. The goal was to let them describe a test in natural language and have the system generate, run, and maintain it.

02The challenges

• Natural language to reliable scripts

  Plain-English descriptions had to become executable tests that cope with dynamic pages, changing DOM structures, and loading states.

• Test maintenance

  Traditional automated tests break whenever the UI changes. The platform needed to recover from minor changes on its own.

• 48-hour constraint

  A working MVP had to be built and demonstrated within the hackathon, without blocking later scaling.

03What we did

1. 01

   AI test generation

   The OpenAI API, combined with DOM analysis of the target application, generates Puppeteer scripts with accurate selectors from natural-language descriptions.

2. 02

   Self-healing selectors

   A multi-strategy locator falls back through CSS selectors, XPath, text content, and visual matching, and updates the test definition when the primary selector fails.

3. 03

   Modular services

   Independent parser, executor, and reporter services connected by a message queue allowed parallel development during the hackathon and easy extension afterwards.

04Key features

• Natural-language test authoring
• Screenshot and video capture of every run
• HTML reports with failure analysis
• Parallel execution across browsers
• REST API for CI/CD integration