Fields

Built-in field kinds:

Kind	Value	Options
text	string	placeholder, minLength, maxLength, pattern
number	number | null	min, max, step, unit, placeholder
boolean	boolean	required requires true
category	string | null	options as strings or { label, value }
mapped-category	string | null	options as { label, value, mapping }. See Mapped Category.
onehot-category	string | null	options as { label, value, mappedTo }. Emits 0/1 one-hot columns. See OneHot Category.
date	Date | null	min, max, step
series	named row values	columns (one or more columns with unique id), minPoints, maxPoints, aggregations, showAggregationSummary (default false); see Series

Shared options:

{
  id: "email",
  kind: "text",
  label: "Email",
  description: "Used for notifications.",
  showDescriptionInline: true,
  required: true,
  defaultValue: "",
  includeInSubmission: true,
  mappedTo: "email",
  hiddenWhen: { kind: "field-value", field: "anonymous", equals: true },
  ui: { autocomplete: "email" }
}

Shared field options also include:

• showDescriptionInline: Shows description by default instead of waiting for the help button.
• inactiveFieldPolicy: Controls whether hidden/disabled fields are submitted.
• includeInSubmission: Set to false to keep a field out of inputs, displayValues, and modelValues.
• mappedTo: Writes the field to a backend feature name, numeric position, or backend-specific map.
• valuePath: Writes the field into a nested modelValues path.

Submission paths reject __proto__, prototype, and constructor segments to keep backend payload construction confined to its own object.

Incoherent built-in constraints are schema errors, not value errors. Text length ranges and regular expressions, numeric ranges, and date ranges are checked during schema normalization so invalid forms fail before a runtime is created.

Series fields

Each series point is an object keyed by the ids declared in columns. There is no fixed column count. Column configurations use the same constraints as the corresponding built-in kind: for example, text length and pattern constraints, number ranges, date ranges and steps, and category options. Invalid nested configuration fails during schema normalization. Nested values also use the built-in serialization contract, so dates are emitted as full ISO timestamps just like top-level date fields.

minPoints and maxPoints control validation and the built-in editor. The editor disables row removal at the minimum and row creation at the maximum. MLForm preserves the supplied point order and duplicate values; sort or de-duplicate them before assigning the series when the backend requires that contract.

Use aggregations to turn numeric observations into multiple backend features without adding editable derived fields. Each aggregate declares its numeric source, operation, label, and mappedTo. The runtime preserves the original series and computes the model values from it; the built-in editor shows only the observations by default. Set showAggregationSummary: true to show the calculated results. Each aggregate’s source identifies a declared numeric column. See Series for the statistical and missing-value contract.