The Usage tab: every API's consumption, day by day, against its limits

CallPages owners get a new Usage tab in the dashboard: every API’s consumption per day, week, and month, shown against that provider’s real limits, with OK / near / over / unknown status pills and per-page monthly AI-minutes bars. The counters live in D1 and record on a path that can never break a call — monitoring is evidence, not a risk.

Quota failures used to arrive as surprises. An owner learned a provider was exhausted the same way a caller did — by the call failing. There was no place to see which API was burning through its budget, how fast, or against what limit; no early warning between ‘everything fine’ and ‘quota exceeded.’ Usage was invisible until it was a crisis.

From surprise to early warning

The old failure mode had a shape: everything fine, everything fine, then a dead call and an owner asking what happened. The Usage tab removes the surprise. Consumption per API, per day, week, and month, against real limits — with a status pill that turns from OK to near while there is still time to act: top up a key, shift fallback priority in the Infrastructure tab, or move a heavy page to its own budget.

Per-page accountability

Aggregate numbers hide the culprit. The monthly AI-minutes bar per page shows exactly where the minutes go — the busy page, the chatty agent, the one that needs a lighter model. Paired with the frugal-model principle, it turns cost control from a policy into a dashboard: see the heavy page, give it the lightest model that does the job, watch the bar shrink.

Evidence, not risk

The counters are built with one hard rule: they can never break a call. They live in D1, they record on a path that fails silently into the counters and never into the call flow. A monitoring system that can take down what it monitors is not monitoring — this one observes from the side, and if its own write fails, the call carries on untouched.

Innovation by Vasily Starostenko.

How it works

1. One tab, every API: the Usage tab lists each provider’s consumption per day, week, and month, next to that provider’s real limits — the same honest limits shown on the provider’s card in the Infrastructure tab.
2. Status pills at a glance: every API carries a pill — OK, near, over, or unknown — so a budget running hot is visible long before it becomes a failed call. Near-limit APIs are the early warning the old system never had.
3. Per-page AI-minutes bars: each page gets a monthly AI-minutes quota bar, so owners can see which page is the heavy consumer and plan capacity instead of guessing.
4. Counters that can never break a call: usage is recorded in D1 on a fire-and-forget path — a failed counter write is a failed counter write, never a failed call. Monitoring observes the product; it does not participate in it.
5. Error and limit accounting: the counters track calls, errors, and quota-limited attempts separately, so ‘the API is erroring’ and ‘the API is exhausted’ read as the two different problems they are.

See it live: Mi Junta

Mi Junta — CallPage · Mi Junta — website

The Usage tab: per-API day/week/month counters against limits plus per-page monthly AI-minutes bars.

Related reading

• Every call record now carries the full AI picture — and a clear reason it ended
• Your catalog can’t be wiped by accident anymore
• Lossless catalog backup: export JSON, import into an empty store, identical state
• This guide in the Learning Center
• Release notes 2026.10.23

Open the Usage tab →

---

Innovation by Vasily Starostenko, founder of CallPages.