Settings
Engine, spend and failover
What the engine is running on, how much of the API budget is used, and what happens when a dependency is down.
Public demo: there is no engine behind this site. The figures below are from the bundled record of real runs. Live runs are by request during early access.
Problem: a checker you can't inspect is another black box. What this does: every model, limit and fallback is listed here. Why it matters: you know exactly what checked your numbers and what it cost. How to use it: check the status before a live run; theme and engine address are below.
Engine
…
Gemini key
—
live runs need it
Reader fallback
—
numbers only, badged
Stored demo runs
—
used if Gemini is down
API spend
Engine offline.
Models and fallbacks
| Render and repair | gemini-nano-banana-2.1, 2K, thinking on |
| Blind reader, strict gate | gemini-3.8-flash, thinking low; sampling at the model's defaults (Google: custom temperature has no effect since Gemini 3.6 Flash) |
| Gemini down at render | The stored run for that scenario, badged as a replay with when it ran |
| Reader down | tesseract, numbers only, badged; the strict gate can't run, so every number is unconfirmed and the run is flagged |
| Masked edit refused, or a missing figure can't be placed | One full re-render with the corrected list, kept only if better |
| Spend cap reached | Run stops; the receipt lists what is unverified |
| Storage | In memory + an append-only journal (Postgres seam reserved, ADR-003) |
| Billing and accounts | Not live |
Appearance
Editorial and dark are the defaults; each choice is remembered on this device.