Timestamp, method, path, status, duration, provider, model, API key name, tokens (prompt + completion), cost, client IP, and trace ID. Sort on any of them.
Click a row’s trace ID to jump straight to the matching trace.
Open Monitoring → Traces (/dashboard/monitoring/traces).
A trace is a timeline of the steps a request went through inside the gateway. Each step has an icon:
Icon
Step
Shield
Guardrail evaluation (PII, hate speech, prompt injection, keyword, IP).
Key
Authentication and quota check.
Router
Routing rule match + provider selection.
Cloud
Upstream model call.
Cache
Cache hit or miss.
Click a step to see its latency, tokens consumed, cost contribution, and any error. The entire trace is time-aligned, so a slow guardrail stands out immediately.
Audit logs describe administrative actions on a specific gateway (config changes, key rotations, failed auth). They stay on the gateway and are not routed through the cloud monitoring pipeline. View or export them from the gateway’s detail page while the dashboard is connected to it.
From any of the three views, click Export to download the current filter as CSV or JSON. Exports are assembled locally by the gateway; the cloud dashboard only forwards the query.
