DSC-BRAIN USER GUIDE · HOW IT WORKS

The planner: what it decides and why

The planner is the part of the brain that works out the cheapest safe set of actions across all your devices. Today it works out its plans alongside the hub’s own control and writes down what it would do and why. The hub’s own control still runs the room.

●  Running on our test kit: 9.10.0 · kit firmware 9.0.0.0 · last report 22 Oct 2026

Status: Not in control yet. The planner sends nothing to any device.

On this page
  1. What it does today
  2. How it decides
  3. What it puts first
  4. Side effects and costs
  5. The Why feed
  6. Comparing the planner with the live path
  7. What the planner does not do

What it does today

The planner works out its plans alongside, and the hub’s own control still runs the room. The planner reads the same readings, works out a plan for each output and logs it. It writes decisions to a log and nothing else. No device, no hub setting and no setting of yours is touched. When its plan differs from what the room did, the log says so.

For every device type, control by the planner is “Not in control yet”. Who is in charge and rolling back shows how a type would be handed over, and how to take it back. See also What runs today and what is coming.

How it decides

  1. State. It reads each quantity with its observers, how many there are, and how old each reading is. A reading the observers dispute, or that nobody can observe, is unknown. It is never filled in.
  2. Problems. It lists what is out of band, such as humidity above its target.
  3. Options. For each problem it lists the ways to fix it, ranked from practice. An option whose precondition is unknown is marked unavailable, and the plan says why.
  4. Prediction. For each option it predicts how heat and water would change in each space over a short horizon, ten minutes by default. Every predicted number is labelled as predicted.
  5. Choice. Each output gets one decision from one owner. The hub’s own demand for the same output is written beside it as the hub’s would-do. It is never a second master.
  6. Explanation. The reason line holds the observed numbers, where they came from and how old they are.
  7. Log. The decision goes into a log that the Why feed reads.

What it puts first

It handles problems safety first. Between options that fix a problem it prefers, in this order: safety, the target bands, side effects on other readings, energy, wear, and noise.

It also keeps rules it never trades against a band:

  • It never overrides an emergency in the hub. The whole plan stands down.
  • It never turns a lamp on in a dark window, or in a window whose state is unknown.
  • It never humidifies a flowering space in the dark, or during the humidity spike after lights off.
  • It never turns heat on by default. A heater or a mat runs only for a known, real shortfall. A mat answers a known root zone shortfall, never an air reading.
  • When a heater runs, outside exhaust stays at zero with its damper shut, and the room exhaust stays at or above its floor.
  • When lamps are on and the space is too hot, all air goes outside. Lamp heat is never recirculated into the room.
  • Devices that conflict never run together in one air volume, for example a humidifier and a dehumidifier.

Side effects and costs

Each device moves heat and water. The planner carries a size for each effect: a rated value, a value learned from earlier reactions, or unknown. An unknown size is never replaced with a typical one. The direction still counts, so the planner can check that fixing one reading does not push another out of band.

  • A heater adds its power as heat and lowers relative humidity without removing water.
  • A dehumidifier removes water and adds heat: its input power plus the heat released as the water condenses.
  • An ultrasonic humidifier adds water and cools the air. Its mist lands where its output is placed, which on this kit is inside the clone dome.
  • A lamp’s power ends up as heat where it hangs.
  • A fan moves air from a source to a destination and carries heat and water with it.
  • A single-hose air conditioner pulls room air out, so the room is under pressure, and the same volume comes back through gaps from the house.

The Why feed

The Why feed is the planner’s log in plain words. It is on Today, in a card headed “WHY · WHAT THE PLANNER DID, HELD OFF AND WOULD DO”. It shows the latest five lines with a “See all” button. Three pickers filter it: Space, Device and Show. Show offers everything, did, held off, not run and would do.

On each row What it means
DID, HELD OFF, NOT RUN, WOULD DO A verb tag. DID means the live path acted. WOULD DO is the planner’s own plan, and nothing was sent.
Role · space, and an age Which device role, in which space, and how long ago.
The plan line What it would do.
The reason The decision log’s own reason, with its numbers. If there is none, the row says “Reason unavailable: the decision log holds no reason for this row.”
Why unknown: ... is not reporting Shown only when the brain names the observer it is missing.
Live path did: ... What the live path did, with “agrees”, “differs” or “agreement unknown”.
Hub would do: ... What the hub’s own control would do for the same output, with agrees or differs.
unchanged for 3 h (12 checks) Identical repeat decisions are folded into one line.
Rejected options A collapsed list of options the planner set aside.

Health chips can sit above the list: WATCH, ACTING DODGY, FAILED and HEALTH UNKNOWN. Health is advice. See Maintenance, calibration, health and snooze.

The humidifier and dehumidifier tiles on the climate screen can show a one-line hint: “Would do:” with the plan and its reason, or “Would do: unavailable (no planner decision for this tile)”.

Comparing the planner with the live path

Each Why row can say whether the planner and the live path agree. The Control screen counts these comparisons as evidence, as described on Who is in charge and rolling back. A row that says differs is evidence, not an alarm.

What the planner does not do

  • It does not drive any device today.
  • It does not guess a missing input. It marks every option that needs it as unavailable.
  • It does not override the hub’s safety limits. See When the brain stops: the hub and the plugs.