AI Check Recovery Agent

supaguard includes an autonomous AI Check Recovery Agent that helps you maintain your synthetic monitors with minimal effort. When a check fails consistently, the agent analyzes the failure, generates a fix, and presents it to you for approval.

How it Works

1. Failure Detection: If a check fails 3 times in a row (configurable) for the same semantic reason, the Recovery Agent is triggered.
2. Semantic Analysis: We use LLMs to verify that the failures are fundamentally the same root cause (e.g., a changed CSS selector) and not just transient network flakiness.
3. Fix Generation: The agent analyzes the Playwright script and the error logs to generate a repaired version of the code.
4. Human Review: A "Pull Request" style notification is sent to your Slack channel or Email.
5. Approval: You can review the side-by-side diff on the supaguard Dashboard and apply the fix with a single click.

Configuration

The Recovery Agent can be configured via your organization settings.

• Failure Threshold: Default is 3 consecutive failures.
• Sensitivity: Adjust how strict the semantic matching should be.
• Muting: Choose whether to mute further alerts while a fix is pending.

Benefits

• Reduced Maintenance: Spend less time manually fixing broken selectors.
• Faster Recovery: Get back to green faster with AI-assisted debugging.
• High Confidence: Every fix is reviewed by a human before being applied to production monitors.