Quota Usage & Alerts
The Quota Usage & Alerts screen is reached from the Identity Management menu. Like Limit Plans, it is listed under the AI Gateway menu group as well only for a project member who cannot open Identity Management; both entries open the same screen. It can be opened both from the project menu (that project's own consumers only) and from the Admin menu (every project); both menus show the same screen with a different scope.
Both tabs are governed by the Identity permission group: View to read the report and the alert rules, Manage to create, edit or delete an alert rule. The AI Budgets permission group is accepted in place of Identity, so a role holding only AI Budgets → View reaches the screen and both of its tabs — the same OR the Limit Plans screens use. The Admin menu's cross-project scope always requires a system administrator, whichever group the role carries. Two lookups on the Alerts tab belong to other permission groups: the LLM provider list (AI Development → View) and the e-mail connection list (Connections → View). A role that lacks them still opens the screen normally — no error is shown; only those two lists stay empty, so a provider-scoped rule displays the raw provider reference instead of the provider's name and the e-mail connection field in the rule detail stays blank. Adding the relevant View permission to the role makes the names resolve.
Quota Sources
Every row listed on this screen is a counter coming from a different quota mechanism of the platform. The Type column shows which source the counter comes from:
| Type | Source of the Counter |
|---|---|
| Plan | The product/portal subscription quota the consumer is linked to, including the counters of API Request Count limit plans applied to a consumer, an API client, an application or an organization — they show here as Plan rather than as their own type |
| Policy | An API Based Quota policy added on an API Proxy or Proxy Group |
| RLC | A counter coming from a legacy Rate Limit Control List definition. The gateway no longer evaluates these lists — their limits are now limit plans, shown as Plan — so no new counters of this type are created |
| AI Token | The token limits of an AI Token Limit limit plan applied to an identity, or the LLM Gateway Limit of a Token Rate Limit policy; limit and consumption are shown as token counts. Attribution follows the identity the plan is applied to: a consumer, an API client migrated from one, a native API client that never was a consumer, and a portal application each appear as their own rows |
| AI Cost | The USD limits (hour, day, month) of an AI Gateway limit; limit and consumption are shown as currency (USD) |
The screen shows the hour, day and month windows of the token and USD limits. Minute windows and the request-count limits (Requests/Minute, Hour, Day, Month) of an AI Token Limit plan are not shown here.
LLM Gateway rows come from the Token Rate Limit policy: one row for each undivided proxy limit, and one row for each per-value row that has a fixed value. A policy that is split by a variable counts each value of its default limit on its own, so the default limit has no single row.
When a Token Rate Limit policy is split by a variable, each value gets its own row for each LLM Gateway, and the value is shown as the identity. The same value on two LLM Gateways is two separate rows.
- Report
- Alert Definitions
Lists and filters the consumption within active quota windows. The Limits tab of a consumer, API client, application or organization does not show consumption; this screen is where the consumption of limit plans is followed (use the Plan and AI Token types, and AI Cost for USD budgets).
Filters
The report can be filtered by the following criteria:
- Organization, Consumer, API client: Narrow down to a specific organization, consumer or API client.
- API Proxy / Proxy Group: Narrow down to the API Proxy or Proxy Group the counter is enforced on.
- Type: Narrow down to one or more of the quota sources above.
- Show closed windows: Also lists expired windows (closed within the last 24 hours); only active windows are shown by default.
- Show other counters: Also lists counters whose format could not be recognized and are therefore classified as Other — this includes the counters of limit plans that are attributed to a subject other than a consumer, an API client, an application or an organization (an API proxy or a rule assignment); hidden by default.
The list can also be narrowed with a quick display filter to All, 70%+ (watch) or 90%+ (critical) usage level.
Summary Cards
Right below the filter fields, three summary cards are shown; they count active windows only (closed windows are not included in these numbers):
| Card | Meaning |
|---|---|
| Active Counters | Number of active quota windows matching the filter |
| Critical (90%+) | Number of active counters with usage at 90% or above |
| Watch (70-90%) | Number of active counters with usage between 70% and 90% |
Table Columns
| Column | Description |
|---|---|
| Type | The quota source of the counter (Plan / Policy / RLC / AI Token / AI Cost / Other) |
| Consumer | The subject of the counter. If the counter is definitively matched to a consumer (plan counters and policy/RLC counters whose identity matched a username), the consumer's name is shown. Otherwise, the value the counter is keyed by (IP, header, JWT claim, etc.) is shown masked. Rows are grouped by organization and consumer; the group header shows the organization (click it to open the organization's application → consumer → API client breakdown), the consumer and the API clients the counter counts |
| App | The portal application the consumer or the API client belongs to; empty for counters not linked to the portal |
| API Proxy/Group | The API Proxy or Proxy Group the counter is enforced on |
| Environment | The environment the counter is measured in |
| Interval | The length of the quota window: Hourly, Daily or Monthly |
| Limit | The actual request (or, for AI Cost, USD) limit enforced for this window; empty for unlimited counters |
| Consumed | The actual amount consumed so far in this window (live counter) |
| Window Reset | The date and time the current quota window resets (renews) |
| Usage | Consumption as a percentage of the limit; colour-coded healthy below 70%, watch between 70-90%, critical at 90% and above |
| Closed | Badge for an expired window; only shown when "Show closed windows" is enabled |
Report data is refreshed from the quota counters every minute; the consumption of the last few minutes may appear slightly delayed. AI Token and AI Cost rows come from the hourly AI billing aggregation, so they may lag real-time enforcement by up to ~1 hour.
Rules that send an automatic notification when quota consumption meets a defined condition.
Condition Types
| Condition | Triggers When |
|---|---|
| Usage percent above threshold | Usage reaches a defined percentage (e.g. 80%) |
| Remaining request count below threshold | The remaining requests/tokens in the window fall below a defined value |
| Quota exhausted | The window limit is reached (100% usage) |
| Consecutive windows above threshold | Usage exceeds the threshold in a defined number of consecutive windows (to catch persistent/repeated overuse) |
A rule can optionally be restricted to a specific quota interval (Hourly/Daily/Monthly) and specific counter types (Plan/Policy/RLC/AI Token/AI Cost); if none are selected, the rule covers only the gateway request counters (Plan/Policy/RLC) — AI Token/AI Cost coverage must always be selected explicitly. When an AI counter type is selected, the rule can also be optionally narrowed to a specific model or provider.
Scope and Recipients
A rule covers all organizations and consumers by default; it can optionally be restricted to specific organizations. The alert e-mail can be sent to one or more of three recipient types:
| Recipient Type | Description |
|---|---|
| Organization | The notification e-mails of the organization the consumer exceeding/approaching the quota belongs to |
| Portal Account | The e-mail of the portal account (for counters linked to the portal) |
| Static Recipients | Fixed e-mail addresses defined within the rule |
The Alert Thresholds of an AI Token Limit limit plan do not send a notification by themselves. When a threshold is passed, the traffic log entry is marked and the event appears in the Active alerts panel on the Overview tab of the AI view of Analytic → Dashboard. To be notified by e-mail or another channel, define a rule on this tab with AI Token or AI Cost selected as the counter type, or use the alert trigger types AI Quota Soft-Cap Hit Count and AI Quota Alert-Only Count.
The e-mail content is sent with a built-in template; a custom subject and content template can be defined if desired. Additional notification channels (Slack, webhook, etc.) can also be defined alongside the e-mail.
Reminders and Notification History
While the condition persists within the same quota window, at most a defined number of notifications are sent, with at least a defined number of minutes between consecutive notifications. The last 100 notifications sent for each rule can be viewed as history, together with their date, recipients and content.
Relationship Between Quota, Limit Plans and Throttling
The counters shown on this screen belong to long-window (hourly/daily/monthly) quota mechanisms: the consumer plan, the API Based Quota policy, the hourly/daily/monthly windows of limit plans, and AI Gateway limits.
Short-window (second/minute) throttling counters — such as the API Based Throttling policy, or the second/minute windows of limit plans — are not shown on this screen. These counters exist for instant protection (burst/DDoS prevention) and reset and renew very quickly on every request. The instantaneous state of these short-lived limits can instead be followed through the rate limit information reflected in the response headers of the related policy (remaining count, limit, reset time).
- Use this screen to monitor long-window quota consumption (monthly subscription limit, hourly/daily/monthly limit plan quota, AI limits) and to get alerted when a threshold is approached.
- To see the instantaneous state of short-window rate limiting, check the rate limit information in the response headers of the related API call.
- For historical analytics of AI Gateway usage by person/group/model/provider, see AI Reports; for general API traffic analytics, see the Traffic Metrics report.
Reading These Counters Through the Management API
Everything this screen shows can also be read programmatically, per subject, through the five read-only endpoints of the Usage Reports API — one each for an organization, a consumer, an application, an API client and an API proxy. Each answers the period's token and cost consumption, the gateway traffic behind it, the live quota windows (limit, consumed, remaining, percentage, reset time) and, for an organization or an application, the breakdown one level below.
Two differences from the screen are worth planning around:
- The budget rows are live, exactly as here: they describe the windows open right now and ignore the requested period. The response states how fresh they are — the counters are synchronized about once a minute.
- A Policy or RLC counter is answered as a single summary row per window rather than one row per identity, carrying the worst percentage observed. Those counters are keyed by a request identity (an IP address, a header value) that the platform stores encrypted, so it is never published on an automation surface.
A source that cannot be read is reported as a partial result with the unreachable environments named; it is never answered as zero consumption.
Related Pages
Configuring the long-window (hour/day/week/month) total request quota policy.
Configuring the short-window (second/minute) rate limiting policy.
Managing identity-based rate limit definitions.
Configuring the token and USD limits of an LLM Gateway.