Insights — proxifai Docs

Insights is proxifai’s observability surface. It’s a tenant-aware façade over a standard Grafana-stack backend: Loki for logs, Mimir for metrics, and Tempo for traces. In the app it lives at /insights with tabs for Dashboards, Logs, and Audit.

Insights mounts only when INSIGHTS_ENABLED is set and the backends (LOKI_URL, etc.) are configured. Deployments without the observability stack simply don’t show the routes — no broken panels.

Logs, metrics & traces

The Insights API is a thin, multi-tenant proxy that injects org/service scoping and forwards to the backend:

GET /api/v1/insights/logs?q=<LogQL>&service=&env=&from=&to=      # → Loki
GET /api/v1/insights/metrics?q=<PromQL>&service=&env=&from=&to=&step=   # → Mimir
GET /api/v1/insights/traces/{traceID}?service=                  # → Tempo

Logs — write a LogQL query with label selectors and line filters; results stream back scoped to your org (and to the picked service when one is in scope).
Metrics — write a PromQL query with an optional step for range resolution.
Traces — look up a distributed trace by ID.

Because the façade enforces tenant isolation, you only ever see your organization’s telemetry.

Scope awareness

When a service is selected in the scope picker, the Logs view passes ?service=<id> and the backend filters to that service’s streams. With no service picked, you see the whole workspace.

Dashboards

The Dashboards tab hosts user-creatable dashboards. Build one in the visual builder at /dashboards/new (or edit at /dashboards/{id}/edit) from chart primitives backed by SQL, PromQL, or REST panels. A few system dashboards ship out of the box (team overview, sprint progress, project health). Dashboards are gated by the dashboards feature flag.

Audit log

The Audit tab (/insights/audit) surfaces the organization audit log — who did what, when — backed by GET /api/v1/audit-log. It’s also the basis of the per-org SOC-2 compliance export (/api/v1/compliance).

Retention

Log/metric retention is configurable per environment (see Environments) via PUT /api/v1/projects/{projectId}/environments/{envId}/retention, so noisy dev environments don’t have to retain data as long as prod.