Skip to main content

Documentation Index

Fetch the complete documentation index at: https://langwatch.ai/docs/llms.txt

Use this file to discover all available pages before exploring further.

The /governance route is the org governance daily home: the read-mostly oversight dashboard an admin opens to answer “is the gateway healthy, are we within budget, did anything anomalous happen overnight, are users actually using this?”. It’s the counterpart of /me for the admin persona. It is intentionally distinct from the admin-authoring surface at /settings/governance/*, see Control plane → Daily-use vs admin-authoring for why the split exists.
Pairs with: Control plane & integration tiers for the surface map, Anomaly rules for what populates the anomalies strip, and Cost centers for how the Spend by cost center card attributes spend across the org.
Visibility: the /governance entry only appears in the main sidebar once the org’s setup-state OR-of-flags trips (see When the entry shows up below). A brand-new org sees the Settings → AI Governance admin pages instead until they have data flowing.

​
What you see

The dashboard groups data into four strips, top to bottom.

​
1. KPI strip

Five at-a-glance numbers for the selected window (default: last 30 days):
KPIWhat it shows
Total spendUSD across every gateway request the org saw, summed
Active usersDistinct principal IDs that hit the gateway in the window
Active sourcesIngestionSources in active state (have received at least one event)
Anomaly countAnomalyAlert rows in the window, broken down by severity (anomalyBreakdown.{critical,warning,info})
Recent activityCount of governance-tagged events in the last 24h
The window selector at the top right adjusts every strip simultaneously.

​
2. Spend over time

A daily bar chart of org-wide spend. Click a bar to drill into the per-user breakdown for that day. The chart is the unified view across:
  • Personal Virtual Keys minted via /me.
  • Project Virtual Keys minted from project pages.
  • Anything else that hit the gateway with a vk-lw-* Bearer.
It does not include Ingestion Source events (those are governance-tagged but don’t carry cost), those land in Ingestion-source health below.

​
3. Ingestion-source health

A status row per active IngestionSource:
  • Status pill: active, awaiting_first_event, error.
  • Source type: otel_generic, claude_cowork, workato, etc.
  • Last event: relative timestamp of the most recent successful ingest.
  • Event count: 24h, 7d, 30d, picked from the same ingestionSources.healthMetrics tRPC procedure that langwatch ingest health uses (so the CLI and the web UI are byte-for-byte identical on --json).
A red dot on a source means the receiver hasn’t seen events in the configured window, see the per-source detail page for the timeline + the receiver-side error log.

​
4. Anomalies

The list of AnomalyAlert rows in the window, sorted newest first. Each row shows:
  • Severity chip (critical, warning, info).
  • Rule name that fired.
  • Triggered at timestamp.
  • Current vs baseline values for spend-spike anomalies.
  • Scope (organization, source_type, source).
Click a row to expand the AnomalyAlert detail (the rule definition that matched, the event window that triggered it, the per-actor breakdown). Click the rule name to navigate to the anomaly-rule editor.

​
When the entry shows up

/governance is gated by an OR-of-flags signal that the org has activated governance. The signal is the api.governance.setupState({ organizationId }) tRPC procedure (read-only, cheap, server-side): 
{
  hasPersonalVKs:      boolean,  // any user has minted a personal VK
  hasRoutingPolicies:  boolean,  // at least one routing policy exists
  hasIngestionSources: boolean,  // at least one source exists
  hasAnomalyRules:     boolean,  // at least one rule defined
  hasRecentActivity:   boolean,  // any governance event in the last 30d
}
Governance active = any 3 of 5. After the admin walkthrough you’ll have at least 4 (routing policies + ingestion sources + recent activity from the dev pass + the budget you create implies personal VKs once a developer signs in). The same signal drives:
  • The Govern (Preview) entry visibility in the main sidebar.
  • The CLI langwatch governance status output (mirrors the same booleans).
  • The post-onboarding redirect, when an admin lands on /, the persona-2 home picker sends them to /governance if active, else to /settings/governance to keep configuring.

​
Permissions

The /governance page itself is gated to governance:view (granted to ADMIN by default). Underlying procedures:
  • governance.kpis(organizationId, window), KPI strip.
  • governance.spendOverTime(organizationId, window), bar chart.
  • ingestionSources.healthMetrics(organizationId), health strip.
  • activityMonitor.recentAnomalies(organizationId, window), anomaly list (this lives in the activityMonitor router, not governance, for routing-through-service reasons).
The activity monitor procedures cap at activityMonitor:view, the recommended permission to grant a security-analyst custom role that should see anomalies without managing rules.

​
What you can do FROM here

The dashboard is read-mostly, but every section deep-links into the admin-authoring surface where you’d actually change something:
  • KPI strip → click “Total spend” KPI → drills into per-user breakdown table.
  • Spend over time chart → click a day → opens that day’s per-user breakdown filtered to ±2 hours of the spike.
  • Ingestion-source health row → click into the per-source detail page.
  • Anomalies row → click the rule name → opens the anomaly-rule editor at that rule.
If the dashboard surfaces something you want to act on (rotate a key, edit a rule threshold, archive a source), follow the deep link into the corresponding admin-authoring page under /settings/governance/*.

​
Where to next