IoT-ready growing equipment: what a manufacturer should hand over
Six handover decisions that make growing equipment integrable: interface documentation, test access, control authority, failure states, change behaviour and identity.
Blog
IoT for growing operations, answered.
Practical articles on sensors, monitoring, alarms and data for controlled growing operations—and the IoT engineering behind them—from Manna Insect Oy.
-
-
Traceability in biological production: what a batch record should contain
Identity, time, context, actions and deviations: the five elements a batch record needs before a growing operation can trace, hand over and explain a batch.
-
The hard parts of agricultural IoT: context, mixed equipment and changeover
Context, mixed equipment and changeover: where agricultural IoT projects stall, and the three decisions to take before they do.
-
What agricultural IoT has to solve in a controlled growing room
Climate, equipment state, process records and batch identity: the four kinds of information agricultural IoT has to carry in a controlled growing room.
-
What knowledge qualifies an IoT integration for reuse
A finished IoT integration is reusable only when knowledge travels with the code — scope, registers, evidence, failure behaviour, rights and a named owner.
-
Does local AI keep the whole workflow local — or only the model?
A local model is not a local workflow. Trace retrieval, storage, logs and fallback across the data path before repeating any data-location claim.
-
Verify an agent's grow-room diagnosis against the record behind it
An agent's diagnosis of a grow room is a claim. Verify it against the data, timestamps, log entries and system state behind it before the team acts.
-
Before retrying an AI workflow, check what already happened in the rooms
A missing acknowledgement is not a missing effect. Read back the alarm, the batch entry and the report before retrying consequential agent work.
-
Do not make a growing-room AI task recurring before one result is acceptable
Accept one grow-room agent run before scheduling it. Fix the task definition, overlap and missed-slot policy, authority checks and visible failure states.
-
A read-only AI job still needs a real boundary around the grow room
Name what an analysis task must never change in a grow room, then classify each control by what actually enforces it before trusting the promise.
-
An email request is input to a workflow, not unrestricted authority
An operator's email asking for one room's climate summary defines a task, not a set of permissions. Keep the verified requester, untrusted quoted text and the allowed actions apart.
-
An agent's tool list is not its permission model
Reading a room's climate history, preparing a report and changing a setpoint are different powers. A tool list does not say which one an agent may use, on what, or who approved it.