Self-Healing & Learning

UI changes should not mean broken tests. When your application evolves, RoostGPT automatically detects, diagnoses, and repairs failing tests — and gets smarter with every run.

How Self-Healing Works

When a test fails, RoostGPT determines the root cause and takes the appropriate action:

Diagnosis	What Happened	Action Taken
Selector drift	A UI element moved or was renamed	Test is regenerated with updated selectors
Flaky failure	Network hiccup or environment issue	Identified as non-issue — no code change needed
Code bug	Test logic issue, not a UI change	Agentic fixer patches the test code
Workflow changed	The underlying flow changed significantly	You are notified — the flow needs review
Navigation failed	The UI changed too much to navigate	You are notified — a new test suite is needed

The system re-validates tests before any fix is attempted, filtering out false positives caused by environmental flakiness. Only consistent failures trigger the repair pipeline.

What Gets Healed

• Changed selectors — page object selectors updated to match new DOM structure
• Modified workflows — test steps reordered or updated to follow the new flow
• Timing issues — wait conditions adjusted for slower or asynchronous responses
• Layout changes — assertions updated for new element positions or content

Knowledge Accumulation

The more you use RoostGPT for UI testing, the better it performs. The system builds domain expertise for each application:

• Learns from success and failure — reliable selector patterns are remembered and reused; failed approaches are avoided automatically
• Faster over time — repeat runs benefit from accumulated knowledge, reducing exploration time and improving test accuracy
• Cross-team knowledge sharing — optionally share learnings across your organization so one team's insights benefit everyone testing the same application
• Self-maintaining — accumulated knowledge is automatically organized and kept relevant without manual curation

Enable site memory via ENABLE_SITE_MEMORY=true in your configuration.

Resume & Extend Previous Runs

Never start from scratch. Pick up exactly where you left off:

Operation	Description
Explore more	Discover additional workflows by targeting new focus areas, building on everything from previous runs
Re-run specific workflows	Re-execute only the workflows affected by application updates
Regenerate tests	Produce fresh test code from existing scenario data when generation settings change
Run history	Browse previous runs and select exactly what to re-process

All operations update the same metadata, preserving complete history across sessions.

Import Existing Playwright Tests

Already have a Playwright test suite? Don't start over — import it.

• Existing test analysis — reads your current Playwright tests and understands what they cover
• Coverage gap detection — identifies what is missing so new generation focuses only on gaps
• No duplicates — imported tests are tracked to prevent redundant generation
• Build on your investment — the system enhances your existing suite rather than replacing it

Configure SCAN_EXISTING=true in your .env file to scan before exploration begins.

Next Steps

• Autonomous Discovery — How initial test discovery works
• Page Object Model — How POM selectors are structured for resilience
• CI/CD Integration — Run self-healing in your pipeline
• Configuration Reference — Healing and resume settings