Skip to content

6.9.2.7 Model

6.9.2.7.1 Purpose

A model other systems can call: its name, what it predicts and the entitlement feature an Organization needs to use it. Staff register and retire models in the Console; farm lists the deployed ones for Organizations, asks for predictions on their behalf and stores the outputs as computed observations or events.

6.9.2.7.2 Questions it answers

  • Which models can an Organization use, and which feature does each one need?
  • What does a prediction become in the canonical data: an observation of which type, or an event of which kind?
  • Which version answers predictions today?

6.9.2.7.3 Fields

Field Type Cardinality Description
id string (uuid) 1..1 Key.
name string 1..1 Unique.
e.g. mastitis-risk.
description string 1..1
feature string 1..1 The entitlement feature an Organization needs, e.g. farm.models.mastitis (see Entitlement).
output_kind enum 1..1 Values: observation, event.
What a prediction is written as in farm: a computed observation or a computed event, with the model and its version as origin (see ComputedSourceOrigin) and a derivation naming the input revisions (see CanonicalDerivation).
output_observation_type ObservationType 0..1 Only for output_kind = observation.
output_event_type string 0..1 The CanonicalEvent subtype, e.g. HeatEvent; only for output_kind = event.
deployed_version ModelVersion 0..1 The version that answers predictions; empty when none is deployed.
status enum 1..1 Values: active, retired. Default: active.
retired: the model refuses calls.
created_at datetime 1..1
updated_at datetime 1..1

6.9.2.7.4 Relationships

References

Referenced by

6.9.2.7.5 Rules & constraints

  • A model other systems can call, with the entitlement feature that gates it. Registered and retired by staff with ai-model-platform.models.manage (ai-model-platform.models.manage); farm lists the deployed models for Organizations and asks for predictions on their behalf, and stores the outputs as computed canonical records.