Spend
The wallet at the bottom of the sidebar answers “how much is left”. Clicking it opens the other half of the answer: what it went on. The screen has no menu item of its own — the wallet is the way in, because this is read exactly when someone is looking at the balance.
Where the quarks went
Last 30 days
6,240 left of 7,000
| Metric | Value |
|---|---|
| Spent | 760 quarks |
| An answer, on average | 0.78 quarks |
| Enough for about | 8,000 answers |
The window and the three numbers
The period switches between 30 days, 90 days and a year. Three figures sit on top: the balance against what the plan includes, what was spent over the window, and the average price of an answer.
That average is computed from your runs rather than from a price list, which makes it truer than any table: the length of your documents and the model you picked are already in it. Until there are runs in the period, the screen says so instead of inventing a number.
The per-day chart
One bar per day, split into slices. The Group switch decides what a slice is:
- by kind — the run, a model call, a search, an image, a voice-over;
- by scenario — which of your bots spent how much;
- by channel — public chat, widget, Telegram, VK, webhook, schedule.
The slices always add up to the number above the chart: the slice and the total come from one query, so they cannot disagree. Charges with no run behind them at all — talking to the editor’s assistant, for instance — land in the “Outside a run” slice.
The list of charges
Below the chart are the charges themselves, one row per node per run: when, what, in which scenario, over which channel, the volume and the quarks. A row that links to a run leads to the run breakdown, where the same story is told second by second.
Export to CSV builds a file for the whole window, not for the page you are looking at — otherwise the export would contradict the screen it came from. If there are more rows than the export limit, the file says so in a line of its own.