Updated the pg_dump commands in the migration-safety workflow to use `sed` for removing restrict/unrestrict lines. This ensures consistent schema comparison by ignoring irrelevant metadata.
Introduces a GitHub Actions workflow to ensure Alembic migrations are safe and reversible. The workflow validates schema consistency by testing upgrade and downgrade operations and comparing schemas before and after the roundtrip.
Introduced a mechanism to detect and handle when a target is unreachable, including a detailed offline state message with host and port information. Updated the UI to display a card notifying users of the target's offline status and styled the card accordingly in CSS.
Previously, the loop did not consider the time spent on `collect_once`, potentially causing delays. By adjusting the sleep duration dynamically, the poll interval remains consistent as intended.
Introduced an exponential backoff mechanism with a configurable base, max delay, and jitter factor to handle retries for target failures. This improves resilience by reducing the load during repeated failures and avoids synchronized retry storms. Additionally, stale target cleanup logic has been implemented to prevent unnecessary state retention.
Replaced all inline error messages with the standardized `api_error` helper for consistent error response formatting. This improves clarity, maintainability, and ensures uniform error structures across the application. Updated logging for collector failures to include error class and switched to warning level for target unreachable scenarios.
Introduced standardized error response formats for API errors, including middleware for consistent request IDs and exception handlers. Updated the frontend to parse and process these error responses, and documented the error format in the README for reference.
