Usage & metrics

The metrics page reads your consumption back to you: the tokens spent, the Stories
and prompts opened, the storage taken, and — for your whole organization — the
Corpus created. Running totals first, then the same figures period by period, so you can
see not just how much you have used but when it moved. 📈

Opening the report 🚪

It is the same page in two places, and where you opened it from decides what it counts:

• 🏢 Your organization — from the dashboard, the metrics panel under the
  cards. Tap it anywhere, or use Show more. The page is titled Organization metrics.
• ⭐ One Corpus — from a Corpus page, the View Metrics button. The
  page is then titled Metrics of that Corpus.

A Corpus report counts what that Corpus was asked and what its documents take up. The
organization report counts everything, across every Corpus, plus the Corpus themselves.

🔗

The report carries its scope in its address, so a metrics page can be bookmarked or

sent to a colleague and reopens on the same figures. A link that has lost its scope — hand
edited, or truncated on the way — says so instead of quietly reporting on something else.

Choose the period ⏱️

The Day · Week · Month · Year selector rides at the bottom of the screen, floating
over the report rather than parked off the top of a page you have scrolled down — it stays
under your thumb while you are reading a chart. It picks how the report is cut up, and with
it how far back it reaches:

The line at the top of the report — From … to … — names the exact window being counted.

⏳

The period in progress is left out. The window always stops at the end of the last

complete day, week, month or year — with daily buckets, that means it ends yesterday.
A half-finished bar standing next to finished ones reads as a drop in consumption rather
than as a period that simply has not ended yet. It joins the report once it closes.

The headline numbers 🔢

Above the charts, one tile per metric. The big figure is the running total; the line
under it is what the period added.

• 📊 Tokens — what your prompts and your documents cost. This is the meter.
• 📚 Stories — conversations opened. See Stories.
• 💬 Prompts — questions asked inside them.
• 💾 Storage — what your files occupy. Bear in mind a document weighs more than the file
  you uploaded: the page images, the text conversion, the summary and the embeddings all
  count. See Documents.
• ⭐ Corpus — how many knowledge bases exist. Organization report only — a Corpus
  does not create Corpus. 🙂

The charts 📈

Under the totals, the same metrics bucket by bucket:

• Tokens consumed — a histogram, one bar per period. This is where a spike shows up.
• Activity — Stories and prompts side by side, so you can tell whether your team
  is starting new conversations or going deeper into the ones it has.
• Storage used — a line rather than bars, because storage is a level, not a flow: it
  climbs when you upload and falls when you delete.
• Corpus — Created and Removed over the period. Organization report only.

Charts drawn from more than one series carry a legend above them. Touch or hover any
bar or point to read the exact figure and the period it belongs to. 🎯

Refreshing ↻

Usage figures move about once a day, so they are cached rather than recomputed on every
visit. The ↻ button in the header asks for them again.

While that call is in flight the report you were reading stays on screen, with a thin
progress line at the top — then the charts animate from the old numbers into the new ones,
which is the part worth watching. You never lose the page to a spinner. ✨

When there is nothing to show 🫙

• A brand new organization has no history to chart, and the page says so — “Nothing has
  been consumed over this period yet” — instead of drawing you a flat line. Come back once
  you have started a Story or two.
• If the figures cannot be read, the page says that too and offers Refresh, rather
  than charting zeroes it cannot vouch for.

Keeping an eye on consumption 💡

• Tokens are the meter. They are spent both when you ask questions and when a document is
  ingested — reading a file costs tokens once, answering from it costs them every time.
• A spike is easier to explain the sooner you look. On Day buckets you can still
  remember what happened that afternoon; on Year buckets you cannot.
• Switch to Week or Month for a trend, back to Day to find the culprit.
• Storage only ever goes down when you delete. If the line keeps climbing, the question
  is which Corpus — open the report on each one and compare.