[Python] Looking for ideas to improve my AI-augmented Playwright + Behave + Allure framework

I've been working on an end-to-end testing framework using below tools:

• Playwright for browser automation
• Behave for BDD-style test execution
• Allure for reporting
• AI integration using Ollama API for Selector Healing

Details for AI Selector Healing:


The framework includes an intelligent AI-powered selector healing system that automatically recovers from selector failures using the Ollama AI model.

Features

• Intelligent Recovery: AI analyzes page structure and suggests optimal selectors
• Visual Analysis: Uses screenshots for better element identification
• Confidence Scoring: AI provides confidence levels for suggested selectors
• Historical Learning: Maintains selector mapping for reuse and learning
• Multiple Selector Types: Supports XPath, CSS, and text-based selectors
• Automatic Integration: Seamlessly integrated into Page Object Model

Benefits

• Self-Healing Tests: Tests automatically recover from selector changes
• Reduced Maintenance: Less manual selector updates required
• Higher Reliability: AI suggests robust, context-aware selectors
• Continuous Learning: Improves over time with historical data
• Faster Development: Reduces debugging time for selector issues

Notes

• I'm very new to AI
• My target is to know more about AI and its usage in QA Automation
• Repo URL: playwright-behave-allure-ai-framework

Thanks!

What I have done till now

• Automatic Detection: When a selector fails (throws exception), the system automatically triggers AI healing
• Context Capture: Captures current page screenshot and HTML content
• AI Analysis: Uses Ollama (devstral:24b model) to analyze the page and suggest new selectors
• Validation: Validates AI-suggested selectors before using them
• Learning: Maintains a selector_map.json file for future reference

What I’m looking for


I'm looking to evolve this into something more powerful or genuinely helpful for QA/dev teams.

• Feature ideas that could benefit from AI
• Suggestions on improving the current structure or performance
• Cool/unique ways to use AI in a test automation workflow
• Anything that could make this more useful or developer-friendly

Continue reading...