Greenhouse control systems · Chatham-Kent, Ontario

The controller should ask what you are worried about.

13:20 sun up since 07:11 Demo · simulated house

Equipment every output, and why it is where it is

Advisor this week's report

What the house did. Air ran inside its band on six days; on three afternoons the venting setpoint reached its ceiling and the house sat 0.9 °C over target for about two hours. Humidity deficit stayed under 3 g/m³ on four nights with the crown-rot concern on for 12 days; the fans carried it.

Worth a sentence. The two feed EC cells agreed within 0.03 all week — good. The wind vane held one reading for five hours on Tuesday; check it before the next windy day.

Proposal · 1 of 1

Period 4 venting 24.0 → 24.5 °C

Three afternoons at the ceiling with the vents already at their wind clamp; half a degree of room stops the vents chasing a number they cannot reach.

source: settings sheet · band 22.0 – 26.0 · your number, your call

Every climate computer holds setpoints. None of them hold a grower's judgement. Rootline takes what you are actually worried about this week and moves every setpoint downstream of it, inside limits you set once — and every output on this screen can tell you why it is where it is.

Built on Reliable ControlsBACnet native · no proprietary lock-in

Twelve modules, three shelvesOne engine. Modules you do not have show greyed, with the reason

An AI advisor on every systemReads the week, proposes the change, you tick the box

Written by the installerThe man who built it answers the phone

The AI layer · Advisor

A better decision, faster. That is how yield goes up.

Every week the advisor reads what the house actually did — every alarm, every shot, every number that drifted — and writes it up in plain language against the targets you set. When something is worth changing it proposes the change, with the reason, and you accept it with one tick. It never touches the house on its own.

It finds it sooner

The week you would have lost

Yield goes in the days a wrong-but-in-range number sits unnoticed — a sensor holding last-good, a ceiling that reads fine and does nothing. The advisor's job is to turn that week into a day.

It does the arithmetic

The decision arrives worked out

Not "EC is high". The setting, the value now, the value proposed, the reason in one sentence, and where the number came from. You judge; you do not compute.

You stay in charge

Your numbers win

What you set on the controllers is the definition of correct. The advisor reports against it, never over it. Nothing changes until a person ticks the box.

Why this exists

Growers know things the setpoints do not.

A grower walks his rows every morning and comes back knowing something. Today the fruit is soft. Tonight will be wet. That knowledge then has to be translated, by hand, into eleven numbers on nine pages — and it usually is not, because there is not time.

01 · The translation

Intent becomes arithmetic

Tick what you are worried about. Every affected setpoint shifts by an amount you set once, within a ceiling you set once. Nothing hidden, nothing automatic you did not authorise.

02 · The ceiling

Nothing moves further than you allow

Concerns stack, and stacked concerns are how a system runs away. Every lever carries a magnitude limit, so five worries at once still land somewhere you would have chosen.

03 · The record

It writes down why

Every shift, every diversion, every alarm — logged with the reading that caused it and how many days it has been running. When something goes wrong, the system can tell you what it was thinking.

The platform

One engine, twelve modules.

The engine ships identical to every greenhouse. There is no tomato version and no cannabis version. What changes per site is a profile — which modules the crop uses, the ceilings, and the numbers nobody can guess from an office.

Water · the clean-test layer

Prove the water is clean. Do not hope.

You already own the disinfection. What you do not have is any way of knowing it worked today. Rootline measures everything entering the loop and everything leaving it, stage by stage, and tells you the day it changed.

Fit

Who this suits.

Substrate growersBerries, vegetables and ornamentals on recirculating fertigation, where drain water is an asset and a liability at once.
Anyone on a replacement cycleA climate computer at end of life, or an expansion that will not fit the licence you already bought.
Sites with real disinfectionIf there is UV, ozone or a sand filter in the room, the clean-test layer pays for itself in what it catches.
Growers who want the logic openEvery table, formula and influence curve is yours to see and change. Capability is not sold back to you in packages.

Advisor · the AI layer

A better decision, faster. That is how yield goes up.

A greenhouse produces more numbers in a week than anyone can read. The advisor reads all of them. Once a week it tells you what the house did against the targets you set, what drifted, what is due, and what it would change — worked out, with the reason attached. You accept with a tick or you don't. That is the whole relationship.

What it does

Three things, every week.

01 · The report

What the house did

What alarmed and when it cleared. What drifted outside its band. What the house did against your targets. What changed because a concern was ticked, and for how many days. What is due — a calibration, a filter, a sensor gone quiet. What the weather cost or bought. Plain language, no code, under a page.

02 · The proposal

The change, worked out

When something is worth moving: the setting, the value now, the value proposed, the reason in one sentence, and which authority it came from — your settings sheet, the crop tables, or what it saw. Checked against the band for this house before you ever see it. One tick to accept. Nothing moves until you do.

03 · The witness

What is wrong but in range

The failure that costs yield is a number that looks fine. A sensor that has read the same value for two days is dead, and the report says so. Two readings that agree perfectly, a flat line, a first-shot time that never moves with the weather — these are the sentences it is there to write.

The order of authority

The grower wins. Always.

An AI that has read every horticulture paper will still be wrong about your house. So the advisor is told, in writing, whose numbers count and in what order — and every report says which one it used.

1

Your live numbers

The setpoints, ceilings, periods and concerns on the controllers this week are the definition of correct. If you set a night temperature of 14 °C and the published table says 16, the target is 14. It may say "below the published range" once, as information. It does not call you wrong, and it does not keep raising it.

2

The settings sheet for this house

The minimum, recommended and maximum for every configurable value, set at commissioning with a confidence rating per row. A value outside its band is worth a sentence. A value inside it is not.

3

The crop tables

Published starting points by crop and stage, with their sources. Where a proposal begins — never where a verdict comes from.

4

General knowledge

For explaining why, and for nothing else. It never introduces a number from the internet as a target. If it is not on the controllers, the sheet or the tables, it is not a number this system uses.

Why it cannot hurt the house

It watches from outside. It cannot reach in.

Everything that runs the house lives on the Reliable Controls hardware. The advisor runs on a small fanless box in the panel that reads the controllers and keeps a copy. It has no write path into the controllers — not a policy, a construction. Its only output is a proposal into a box you tick.

Internet down

The house waters, heats, vents, doses and alarms exactly as before. The only thing that stops is a message leaving the building.

Box unplugged

Nothing in the house changes. When it comes back it catches up from the controllers' own logs.

Your data

Stays on your controllers and in your building. The box reads it; it does not own it.

Never a number it was not given

It does not present a value it was not handed, does not call your setting a fault, and does not describe the system as having "decided" something you set.

Watch · the same box

Alarms that reach a person.

The alarms themselves live on the controllers, where they belong. The box is what gets them to a phone: text message, Telegram, push, WhatsApp. Page the first person; no answer in ten minutes, call the next. Who is on tonight, not who is on the list. A chattering alarm becomes one message and a count. And the all-clear is sent — "the heat came back" at four in the morning is the message that lets someone go back to sleep.

It also watches for the faults the head end cannot report about itself: a controller gone quiet, the display unreachable, the internet down for more than an hour. Once a week, the health report. Watch comes with every system; the advisor sits on top of it.

The platform

Twelve modules on three shelves. Start with one.

One engine underneath, one crop profile on top. Every site starts with the base. Field modules are bought one at a time, each in its own controller, so a house can start with one and grow into the rest without anything being replaced. A module you do not have still shows on your screens — greyed, with the reason — so the picture of what the system is doing is always honest.

Catalogue

What each module does.

Stacking

What the second module buys.

Every module earns its keep alone. Put two together and something appears that neither had. This table is also a promise: with the second module absent, the first behaves exactly as its own card says.

A look

What you actually see.

The pages a grower works on. Every box with an outline is a number the grower can change; the row at the foot of each page shows how the live value is put together, so nothing is hidden. Tap any page to open it full size.

Climate setpoints page: six periods anchored to clock, first light or last light, one heating and one venting temperature each, six influences with their live worth, and a foot row showing how the current setpoint is built up — **Climate setpoints.** Six periods, anchored to the clock or to first and last light. One heating and one venting temperature per period, then the influences — radiation, radiation sum, humidity deficit — each with what it is worth right now. The foot row shows how the setpoint the house is holding this minute was arrived at.

Standing concerns page: twelve yes-or-no questions, each moving nine levers by a set offset, with a ceiling row at the bottom and a summary of what the house is doing right now — **Standing concerns.** Twelve yes-or-no questions about the crop. Tick one and it moves the levers — joules, shot size, dry-back, humidity, fans — by a set amount on every shot, every day, until it is unticked. The bottom row is the ceiling: tick everything and it still cannot run away with the crop.

Dosing page: eight periods with joules per shot, minimum and maximum gap and EC offset, a chart of joules banked against the shot threshold, shot-size influences, and three foot rows for when the next shot fires, how much it puts on and its EC target — **Dosing.** A period never asks for water. It sets how many joules must be banked before the next shot, the gaps either side, and the EC offset. The chart shows the joules climbing towards the line; the three rows at the foot show when the next shot fires, how much it puts on, and what is in it.

The hanging line page: live weight of a twenty-cell gutter through the day, channel A and B readings, tare and live weight, integrity checks, dry-back and transpiration figures, and how tonight's dry-back is worked out — **The hanging line.** Twenty load cells on one gutter. Each rise is a shot, each fall is the crop drinking. Two channels check each other; the tare turns gross weight into water plus crop; dry-back, transpiration and water-use efficiency come from there.

The water module configurator: crop, objectives, treatment train and authority level, producing a staged sample-point and instrument list — **The water module configurator.** Pick the crop, what you are trying to catch, and the treatment train; it lays out the sample points and the instruments each stage needs.

By crop

Where each crop starts.

The engine does not change. The profile does — ceilings, targets, and which lab test the system tells you to order.

How a system is built

Four tiers, and the order matters.

Every number in a control system belongs to one of four groups. Knowing which group a number is in tells you who decides it, when it gets set, and what happens if it is wrong.

1ships

Identical everywhere

The engine. Offset program, period resolvers, setpoint and influence logic, curve and trigger programs, the object skeleton, the graphics generators. Would the line be identical in a tomato house in Leamington? Then it belongs here.

2template

Filled per site, from a default

Concern grid and ceilings, stage setpoints, EC and shot grids, outside-temperature and wind tables, dry-back windows, crop profile. Ships with a starting value and a cited source against every figure, so you can see where a number came from and argue with it.

3measured

Measured on site — no default is safe

Stroke times, pulses per litre, sump volume, EC span, valve ranges, compass offset, tank depths. These cannot be guessed from an office, and commissioning refuses to proceed until they are proven.

This tier comes first. A 15% ceiling on shot size means nothing until pulses-per-litre is proven — until then it is 15% of an unknown.

4yours

The grower decides

Which concerns are on, day and night targets, photoperiod, dusk cutoff, whether the hanging line and substrate EC are used at all. These are opinions about the crop, and they are not ours to hold.

Water · the clean-test layer

Prove the water is clean. Do not hope.

It fits alongside whatever control system is already in the building. Everything entering the loop and everything leaving it, measured stage by stage — so you find out the disinfection stopped working on the day it stopped, not when a crop fails.

What it does

Four jobs, all of them checkable.

01

Verify every stage

Filter, ozone, UV — each verified the way that treatment actually works, not with one method borrowed from another. Continuously.

02

Measure uptake, not concentration

In minus out, times volume, per ion. What the plant consumed. No climate computer shows you this.

03

Dump on the ion, not the calendar

Salt cannot be treated out — it leaves by dilution or discharge. Everything else is a treatment problem, not a reason to empty the tank.

04

Aim the lab

This line, this house, today. A schedule of mostly negative tests becomes targeted ones, earlier.

The number that sells it

Your UV is probably sized for bacteria.

250

mJ / cm² · target dose

Viruses are the most UV-resistant thing in your water. Bacteria clear at under 20 mJ/cm² for a three-log kill; the drain-water standard for tobamoviruses is 250. So a unit giving you a perfect bacterial result can be nowhere near virus-competent — and nobody in that room knows the number. Rootline computes delivered dose continuously from transmittance, flow and lamp output, and tells you the hour it falls short.

Your UV unit almost certainly measures its own intensity and run hours already and reports them on Modbus. We read what is there before specifying anything new.

The decision

Where the drain goes.

Three destinations, not two. The third is the one that stops a grower dumping because he is unsure.

To the tank

Normal

Chemistry inside the ceiling, treatment verified. The water goes on to be reused.

To discharge

Chemistry

Sodium, chloride, boron or bicarbonate at the ceiling — nothing can treat those out. Logged with the reading and the volume.

To hold

Unsure

Something is wrong but not yet identified. It must not enter the clean tank, and it should not become regulated discharge either. Hold it until somebody looks.

Aoption

Advise only

No valve, no outputs, nothing it can break. It names the right destination and says why; a person turns the valve. It sits beside your existing control system without touching it.

Boption

Divert within limits you set

You decide what it may do on its own and where the ceiling is. Default is advise. Enable automatic diversion above a figure you chose, and it acts and writes down why.

Say this out loud before you buy anything

What no water system can do.

There is no live virus detection. Not here, not anywhere.

No instrument on the market continuously detects tobamoviruses in irrigation water. Flow cytometry counts bacterial cells and is blind to viruses; species-level answers come from a lab, in days. Anyone selling you live virus detection is selling something that does not exist, and the first outbreak it misses is your crop, not theirs.

What can be done is guarantee the dose that inactivates them, prove it every hour, and put the lab sample on the right line on the right day. That is prevention by proof rather than by promise — and it is a claim we will put in writing.

Trial

Find out whether it actually did anything.

Every season brings a new jug — silicon, kelp, humics, amino acids, microbials. None of them can be shown to be needed by any test, because there is no deficiency to be in. Trial settles it on your own crop, in your own house, with numbers.

Why no test answers it

Nobody official has checked that it works.

In Canada a product like this is a supplement under the Fertilizers Act, and one not on the exempt list has to be registered. But the assessment is safety and standards compliance — not efficacy. A registration number means it will not hurt your crop or the person handling it. It does not mean it does anything.

Which leaves exactly one way to find out: treat some rows and not others, and weigh the difference.

How it works

Six rules, and they are the product.

A badly run trial is worse than none, because it justifies a standing purchase for years. These are built in, not left to memory.

01

Uptake, not EC

Drain EC on its own moves with drain volume as much as with the plant. In minus out, times volume, per ion, is the number that means something.

02

Account for what you added

Most of these products are salts. The treated row's EC changes because you put it in, not because the plant responded.

03

Separate feed and separate drain

The treated row needs its own dosing and its own drain collection. A shared mixing tank or a common gutter means there is no trial, only a story. First thing we check.

04

Pair, alternate, and swap

One row against one row is position, not treatment — vents, heating pipe, gable end, light gradient. Several paired rows across the house, and the treatment swapped partway so position cancels out.

05

Baseline first

Two weeks with both rows instrumented and neither treated. What they differ by with nothing applied is your noise floor — any effect smaller than it cannot be seen, however long you run.

06

Threshold declared before you start

What improvement pays for the product across the whole house, at your price. Written down first, because once a pallet is bought a two percent difference starts looking like a result.

The point

The only feature that saves you money by telling you to stop.

Everyone else in your supply chain is selling you the next jug, and every one of them arrives with a glossy trial run somewhere warmer. With Trial you have numbers from your own rows, in your own house, on your own crop — which means you can say no, and say why.

How we work

Open hardware, ordinary protocols, and a system you could take off us.

A control system you cannot leave is not a purchase, it is a subscription with a wire attached. Everything here runs on hardware anybody can buy and protocols anybody can read.

Underneath

What it runs on.

Controllers

Reliable Controls, BACnet native. An open authorised dealer network across North America, not a single-source dependency on us.

Instruments

Modbus and 4–20 mA. Standard probes from whichever supplier you prefer. Nothing is keyed to us.

Existing kit

Your UV unit, your weather station, your dosing skid. We read what is already there before specifying anything new.

Your data

On your controllers, in your building, readable without us. If you replace us, the system keeps running.

Commissioning

Measured before it is set.

Most control problems are not logic problems. They are a number that is wrong but still in range — a stroke time nobody checked, a flow meter assumed rather than proven. Commissioning here runs in a fixed order with a proof gate at each step, and the system will not accept the next tier until the one below it is signed off.

1

Prove the physical numbers

Strokes, pulses per litre, spans, volumes, ranges — measured on site and recorded with the date and who checked them.

2

Fill the template against them

Ceilings, grids and windows set as real quantities rather than percentages of an unknown, each against a minimum, a recommended start and a maximum.

3

Hand the opinions to the grower

Targets, photoperiod, which concerns are live. Set with you, at the end, when everything underneath is known to be true.

Afterwards

Remote access and one report a week.

No system goes in without remote access — not for us to poke at, but so that a problem at three in the morning is a phone call rather than a drive. Once a week the system sends a health report: what alarmed, what drifted, what is due for calibration, and what it did on its own.

Support scope, response times and what happens if a crop suffers are agreed in writing before installation, per site. We would rather have that argument at the start than in the middle of a heat wave.

Talk to us

One conversation, in your greenhouse.

Bring the problem you cannot currently solve with the system you have. If Rootline is not the answer we will say so — there are jobs better served by a setpoint change than by a new controller.

info@rootline.ca
519 380 5104
Chatham-Kent, Ontario

Status · September 2026

Reference site running in southwestern Ontario. Early conversations welcome; overselling is not.