LLMs and Test Automation: A New Era of Self Healing
summary
System pulls test cases from Jira and converts JIRA steps to Playwright code, then runs them against the live DOM so LLMs see current state. Two models (Mistral and Lamma) generate and patch JavaScript/Playwright locators; when a locator changes (username → amazing username) the framework rewrites the test and reruns it. A five-attempt verification loop uses the test runner to reject bad outputs (attempt 1 fails, attempt 2 fixes and passes), reducing manual maintenance.
AI-written summary · about
chapters (10)
- 0:00Introduction to LLMs and QA
- 1:17Understanding test automation tools
- 2:39What are large language models
- 3:21Pros and cons of LLMs
- 4:26Solving outdated information
- 5:48Addressing hallucination issues
- 6:54Context limitations in LLMs
- 8:43Live demo of self-healing test automation
- 12:49Behind the scenes of the system
- 14:56Final integration and Q&A
related talks
33mShray Sharma: Leveraging AI for Self-Healing Tests: Current Capabilities and Future Prospects
33mExplore LLMs and Multimodal Approaches in Test Automation
48mSupercharged Testing: AI-Powered Workflows with Playwright + MCP
14mAutonomous Testing Agents with LLMs in React
25mAI-Powered Debugging & Browser Automation with Playwright MCP
47mValidating LLMs – A True Story
41mHow continuous testing keeps your LLM on track
29mPlaying your tests wright