Troubleshooting

Common guardrail issues and practical fixes

Use this guide to diagnose common issues during guardrail setup, deployment, and runtime operations.

Quick Triage Flow

Creation and Setup Issues

”Add Guardrail” button not visible

Likely cause: Missing required permission.

Fix:

1. Verify your role has guardrail management permissions.
2. Refresh session and retry.

Provider list is empty or missing expected provider

Likely cause: Provider access not configured or unavailable.

Fix:

1. Confirm provider credentials/licensing.
2. Validate provider availability in your environment.
3. Retry after configuration update.

Probe Selection Issues

Cannot find expected probe

Likely cause: Active filters/search narrowing results.

Fix:

1. Clear search and status filters.
2. Switch to broader tags (for example, All).
3. Verify provider selection in the workflow.

Deployment Issues

Deployment remains pending

Likely cause: Incomplete target selection or backend processing delay.

Fix:

1. Confirm project and endpoint are valid.
2. Re-open deployment details and check status transitions.
3. Retry deployment if no progress occurs.

Guardrail not applied to endpoint

Likely cause: Wrong deployment type or mapping mismatch.

Fix:

1. Confirm you selected Add to Existing Deployment if needed.
2. Check deployment mapping in guardrail detail and project tabs.
3. Redeploy with corrected target.

Runtime Quality Issues

Too many false positives

Fix:

• Reduce severity slightly.
• Disable or refine over-sensitive rules.
• Validate changes in a limited rollout.

Unsafe content still passing

Fix:

• Increase severity for key probes.
• Add missing probes for uncovered categories.
• Ensure both input and output guard types are enabled.

Escalation Checklist

Before escalating, collect:

• Guardrail profile name and ID.
• Selected provider and probes.
• Deployment target and current status.
• Example prompts/responses showing the issue.
• Timestamp of the most recent failure.