Skip to Content
LYNKORIADocs

Metrics catalog

Every number the console shows, where it comes from, how it is aggregated and its limits; plus the metrics it cannot show yet.

The console shows only the metrics listed as available here, and its info buttons show these same definitions. The data architecture is described in ADR 025. Usage numbers are operational, not a billing ledger.

Available

Calls created

Calls (one-to-one, group and livestream) created by your backend. Repeating a create for an active productCallId returns the same call and is not counted again.

KeycallCount
KindHistorical
Sourceproject_usage_hourly and project_usage_daily rollups, incremented by the API as events happen
Aggregationsum
BucketsHour (ranges up to 7 days) or UTC day
RangesLast 24 hours, 7, 30 or 90 days, or a custom range up to 400 days; hourly data covers the last 35 days
DelayAbout 15 seconds

Participant minutes

Time participants spent in calls, summed over participant sessions. Livestream viewers count.

KeyparticipantMinutes
KindHistorical
Sourceproject_usage_hourly and project_usage_daily rollups, incremented by the API as events happen
Aggregationsum
BucketsHour (ranges up to 7 days) or UTC day
RangesLast 24 hours, 7, 30 or 90 days, or a custom range up to 400 days; hourly data covers the last 35 days
DelayRecorded when each participant session ends
LimitationsMinutes are written when a participant session ends and are split across the hours and days in which they took place, so ongoing calls are not included until participants leave.

Peak participants

Highest number of participants connected at once in one environment during the bucket.

KeypeakParticipants
KindHistorical
Sourceproject_usage_hourly and project_usage_daily rollups, incremented by the API as events happen
Aggregationmax
BucketsHour (ranges up to 7 days) or UTC day
RangesLast 24 hours, 7, 30 or 90 days, or a custom range up to 400 days; hourly data covers the last 35 days
DelayAbout 15 seconds
LimitationsAcross several environments or projects the console shows the highest single-environment peak, not a sum.

Messages sent

Messages stored. Edits, deletions and reactions are not counted; idempotent retries count once.

KeymessagesSent
KindHistorical
Sourceproject_usage_hourly and project_usage_daily rollups, incremented by the API as events happen
Aggregationsum
BucketsHour (ranges up to 7 days) or UTC day
RangesLast 24 hours, 7, 30 or 90 days, or a custom range up to 400 days; hourly data covers the last 35 days
DelayAbout 15 seconds

Tokens issued

Messaging tokens plus call join tokens minted by your backend.

KeytokensIssued
KindHistorical
Sourceproject_usage_hourly and project_usage_daily rollups, incremented by the API as events happen
Aggregationsum
BucketsHour (ranges up to 7 days) or UTC day
RangesLast 24 hours, 7, 30 or 90 days, or a custom range up to 400 days; hourly data covers the last 35 days
DelayAbout 15 seconds

API requests

REST requests authenticated with a service key or a messaging token. Console and health requests are not counted.

KeyapiRequests
KindHistorical
Sourceproject_usage_hourly and project_usage_daily rollups, incremented by the API as events happen
Aggregationsum
BucketsHour (ranges up to 7 days) or UTC day
RangesLast 24 hours, 7, 30 or 90 days, or a custom range up to 400 days; hourly data covers the last 35 days
DelayAbout 15 seconds
LimitationsRequests rejected before authentication (bad or missing credentials) cannot be attributed to a project and are not counted.

API errors

Counted API requests answered with HTTP 4xx or 5xx.

KeyapiErrors
KindHistorical
Sourceproject_usage_hourly and project_usage_daily rollups, incremented by the API as events happen
Aggregationsum
BucketsHour (ranges up to 7 days) or UTC day
RangesLast 24 hours, 7, 30 or 90 days, or a custom range up to 400 days; hourly data covers the last 35 days
DelayAbout 15 seconds
Limitations4xx responses include your own client errors (validation, quota), not only failures.

Error rate

API errors divided by API requests over the whole bucket or range.

KeyerrorRate
KindHistorical
SourceDerived from the API requests and API errors sums (never an average of rates)
Aggregationratio
BucketsHour (ranges up to 7 days) or UTC day
RangesLast 24 hours, 7, 30 or 90 days, or a custom range up to 400 days; hourly data covers the last 35 days
DelayAbout 15 seconds

Active calls

Calls that have been created and not ended, by type.

KeyactiveCalls
KindLive (current value)
Sourcecall_sessions rows with status active
Aggregationcount of current rows
BucketsNone (current value)
RangesNow
DelayImmediate; the console refreshes every 15 seconds while visible
LimitationsA call LiveKit has already closed stays active until the reconciler or the room_finished webhook ends it (normally within a minute).

Connected participants

Participant sessions that have joined and not left, across active calls.

KeyactiveParticipants
KindLive (current value)
Sourcemedia_participant_sessions rows without left_at (from LiveKit webhooks)
Aggregationcount of current rows
BucketsNone (current value)
RangesNow
DelaySeconds (LiveKit webhook delivery)

Livestream audience

Connected participants in active livestream calls (hosts, speakers and viewers).

KeylivestreamAudience
KindLive (current value)
Sourcemedia_participant_sessions joined to livestream call_sessions
Aggregationcount of current rows
BucketsNone (current value)
RangesNow
DelaySeconds (LiveKit webhook delivery)
LimitationsRoles are not stored per session, so viewers and hosts are not split.

Concurrent room quota used

Active calls in an environment as a share of its maxConcurrentRooms limit, the same count the API uses to refuse new calls.

KeyroomUtilization
KindLive (current value)
Sourcecall_sessions (active) and the environment limits
Aggregationratio
BucketsNone (current value)
RangesNow
DelayImmediate

Possibly stale calls

Active calls with nobody connected that started more than 10 minutes ago. Normally ended by the reconciler; a growing number signals missed LiveKit webhooks.

KeystaleCalls
KindLive (current value)
Sourcecall_sessions and media_participant_sessions
Aggregationcount of current rows
BucketsNone (current value)
RangesNow
DelayImmediate

Webhook deliveries pending

Outbound webhook events not yet delivered and not dead-lettered.

KeywebhookPending
KindLive (current value)
Sourceoutbox_events
Aggregationcount of current rows
BucketsNone (current value)
RangesNow
DelayImmediate

Webhook deliveries dead-lettered

Events that exhausted their retries (24 hours). Admins can retry them.

KeywebhookDead
KindLive (current value)
Sourceoutbox_events
Aggregationcount of current rows
BucketsNone (current value)
RangesNow
DelayImmediate

Not available yet

These are not shown in the console. Each needs the listed prerequisite first.

MetricWould come fromPrerequisite
API latency (p50, p95, p99)PrometheusA monitoring backend scraping the API metrics port (PD-5), and a platform operator role to see it (PD-1).
Users onlineRedis presence leasesPresence is stored per user; an online counter per environment would have to be maintained by the presence service first.
Realtime connectionslynkoria_sockets_connected gauge per instanceA monitoring backend (PD-5); the gauge has no project label by design.
CPU, memory, network, LiveKit node healthPrometheus node and LiveKit exportersA monitoring backend (PD-5) and a platform operator role (PD-1).
Bandwidth and egressNot recordedLiveKit does not report bytes per room to the API today; would need Analytics or per-room stats collection.

On this page