DSC-BRAIN USER GUIDE · USING THE DASHBOARD
The model inspector
The brain keeps a model of your kit: every device, every place and how they connect. The model inspector shows how sure the brain is about each part, and asks you only for what it still needs.
● Running on our test kit: 9.10.0 · kit firmware 9.0.0.0 · last report 22 Oct 2026
Status: Live. On by default in brain 9.9.0. The inspector has no switch of its own. It appears wherever its screen is on.
On this page
Two places to look
- Kit model. In Settings, in the Advanced row, choose “Kit model”. The page is headed “What the model still needs from you”. It has two lists: Open questions and Gaps.
- How sure is the model. Every device page has a short block with this name. It is the per-device view. See Maintenance, calibration, health and snooze for the rest of the device page.
Open questions
Each open question is a card. It shows the question and the current value with how it is known, for example “Now: tent? (a guess from the old setup)”, or “Now: not known”. Below that is the input. The input is always a pick, apart from a measured number. It is one of:
- Pick one: a list of options.
- Measured value: a number, with its unit.
- Measured size in cm: width, depth and height.
- Pick the pots: a set of checkboxes.
A question with no choices says “This question offers no choices yet, so nothing can be confirmed.” That is honest. It does not offer a default.
Choose “Confirm”. The brain then reads its list of open questions again. The card closes only if the question is gone. The line under the card says one of three things:
| Line | What it means |
|---|---|
| Applied. The model now reads this as declared by you. | The brain no longer lists the question. The value is saved as stated by you. |
| Unconfirmed: Sent, but the model still shows this question open. Nothing has changed yet. | The brain took the answer but its list still has the question. Try again, or check the brain. |
| Not applied: ... | The brain refused it or did not answer. The reason follows. |
When there is nothing left, the page says “No open questions. Every value the model holds has been confirmed or observed.” A collapsed Details line on each card holds the technical reference.
Gaps
A gap is something the model cannot place or size by itself. Each line gives the brain’s own title. It ends with “answer it under Open questions” when you can close it, or “nothing you can answer here closes this yet” when you cannot. With no gaps the page says “No gaps: the model has nothing it cannot place or size.”
How the brain knows a value
A name, a role or a size is shown with its basis. These are the words:
| Words on screen | What they mean |
|---|---|
| you said so | You picked or entered it. |
| you confirmed it | You answered a question about it. |
| seen by the hub | The hub observed it. |
| carried over from the old setup | It came from an earlier version of the kit’s setup. |
| from the product library | It comes from the DSC-KitLib record for the product. |
| a generic estimate | A typical value for this class. It is for display only and never becomes a control number. |
| a guess from the old setup | An earlier setup guessed it. |
| not known | Nobody has said, and the brain has not seen it. |
A value that rests on a guess or an estimate carries a question mark, so “Tent?” means the name is a guess. Answer the open question and the question mark goes.
How sure is the model, per device
On a device page this block lists: Name (with its basis if it is a guess), Role, Stage, the number of open questions with each one named, and the number of gaps. If a list could not be read, the block says “not read” instead of showing zero.
When readings disagree
Some readings come from more than one observer, for example two sensors, or a sensor and a camera. The brain compares them and marks the result:
| Mark | What it means |
|---|---|
| ≠ disputed · 2 of 3 observers | The observers disagree. The reading has no value until they agree. |
| one observer only · 1 of 2 observers | Only one observer is working. One is not corroboration. |
| unobservable · 0 of 2 observers | No observer can read it. |
A disputed or unobservable reading is unknown to the planner and to the cards. “Dark held” on the Today overnight card needs two observers that agree against the stated light threshold. One observer, a dispute or none reads unknown. A blank or covered camera frame reads unknown. It is never read as dark.
On a device page a line can name the observer furthest from the others, but only when the brain links that observer to that device. We never guess it from a name.
How screens say “unknown”
One renderer draws every absence. It always gives a reason, how long it has held when known, and where to fix it. It never shows zero, off or idle in place of unknown.
| Label | Used when |
|---|---|
| Not here | The thing does not exist in the model. |
| Not reporting | It exists but is not sending data. |
| Not needed right now | It does not apply at this moment. |
| Not set | Nobody has set it. |
| Unavailable | The source did not answer. |
The reading drawer
Click a reading or a device tile on a desk and a drawer opens. It shows when the reading last changed, today’s run time and cycles for a switch, the current state, and a short note on what the reading is and what to do about it. Below is a history chart with the previous period faded behind it. Chips say when history is thin: “Not recorded by the brain”, “Not recorded yet”, “Limited history” and “Hourly past the raw window”. A switch has Turn on or Turn off with a confirm step, and says Applied, Not applied or Unconfirmed from the device’s own reading.
Old and v2, side by side
In Settings, in the Advanced row, choose “Old and v2, side by side”. For each surface you can pick Old or v2. The line under the pick reads “The brain reads old” or “The brain reads v2”. After a change it says Applied, Not applied or Unconfirmed from what the brain then reads. For device labels it lists the old names beside the v2 names and marks the ones that differ. Surfaces that have no v2 feed yet say so and stay on Old. Putting a surface back on Old hides its v2 screen. The v2 screens in these pages are on by default. A few data surfaces, such as device labels, fan rows, chart series, wiring and twin ducts, still read Old until you switch them.