Dashboard
The Dashboard is the landing page of the portal. Its eyebrow reads Overview and its heading is Dashboard. It answers one question quickly: how much credit do I have, and what has my organization been doing lately?
The Dashboard has no date-range control. Every figure on the page covers a rolling 30 days ending now. To choose a period, open Usage, which has the full range picker.
Balance & spend
The first section carries a renews {date} chip showing when the current billing period rolls over.
| Element | What it shows |
|---|---|
| Credit wallet | The headline figure: your wallet's effective balance, with the word "effective" beside it. Effective means drawable credit minus usage that has been recorded but not yet settled. |
| Coupon credit | The part of the balance that came from redeemed coupons. |
| Top-up credit | The part that came from top-ups. |
| Pay-as-you-go this period | Hinted "List value of everything used this period." |
| Daily limit used | A percentage with a bar, captioned "{spend} of {limit} per day". |
Credit wallet shows - and Not available in two situations: the wallet call failed, or you are a plain member. Reading the wallet requires the tenant admin or owner role, so members see the placeholder rather than an error.
Daily limit used shows - and No daily limit set when you have no personal daily limit. You can set one from Profile. See Cost Limits.
The wallet card does not distinguish billing mode. On a postpaid account there is no prepaid balance to draw down, so Credit wallet, Coupon credit and Top-up credit all read $0.00. Read that as not applicable, not out of credit; nothing is blocked and no top-up is required. Once you redeem a coupon, real figures appear.
Billing periods currently stay open rather than closing on a schedule, so this figure keeps accruing from the moment the period opened instead of resetting. For the same reason the renews {date} chip may show a date in the past, or may not appear at all.
Usage over time
A 30-day chart with legend chips for Requests, Tokens, Input and Output. The totals beside each chip are real 30-day sums.
The chart has no y-axis, and each series is normalised against its own peak so all four fit one frame. Two lines sitting at the same height do not represent the same quantity. Use the numeric totals for magnitude, and the Usage charts when you need to read values off an axis.
Request activity
A grid of small squares, one per day of the window, shaded by how many requests that day carried. It answers one question: which days did this account call the API?
Open a day's detail by hovering a square, or click it to pin the panel open. On a keyboard, tab to a square and it opens on focus; Enter pins it. Escape, a click outside, or scrolling the page closes it.
| Row | What it shows |
|---|---|
| Requests | Requests made that day |
| Input tokens | Input tokens spent that day |
| Output tokens | Output tokens produced that day |
| Total tokens | The day's total |
Days with no traffic open too, and read 0 on every row. The point of the grid is telling a quiet day apart from a day with nothing at all, so an idle square is still a square you can ask about.
Total tokens is the figure the platform reports for that day, not the two other rows added together. It can account for tokens neither side of the split carries, and it is the same total the tiles below and Usage report -- adding the split yourself would quietly disagree with them.
Cost, latency and the success/failure split are deliberately not on this panel. This card answers when you called; Usage owns the rest.
Headline metrics
Five tiles summarising the same 30 days:
| Tile | Notes |
|---|---|
| Total requests | Links through to Usage. |
| Total tokens | Input plus output. Links through to Usage. |
| Success rate | Share of requests that completed successfully. |
| Avg latency | Captioned with a throughput figure in tokens per second. |
| Top model | Your most-used model. Links through to Models. |
Platform
A status line reads All jobs idle or {n} jobs active, followed by six tiles that each link somewhere useful:
- Models available → Models
- Active API keys → API Keys
- Jobs running, pending, completed and failed → Jobs
Token split & cache
A donut splitting Input against Output tokens, followed by three rows:
| Row | Meaning |
|---|---|
| I/O ratio | Input tokens per output token across the window. |
| Cached tokens | Input tokens served from the prefix cache. |
| Cache hit rate | Share of requests that hit the cache. |
Cached input is charged at a reduced rate, so a rising cache hit rate is worth watching. Keeping the shared or system portion of your prompts identical at the start of every request is what makes the cache work.
What the Dashboard deliberately omits
The Dashboard reports balance state, never priced usage. There is no cost total, no per-model spend and no charge breakdown here. Money lives on Billing, and priced exports live on Usage.
Where an amount does appear, values below one cent render as <$0.01 rather than rounding to $0.00, so a small non-zero balance is never mistaken for an empty one.