YAML Param Schema Format

The parameter schema tells the OmniLoop dashboard what controls (sliders, toggles, text inputs) and readouts to render for your application. This means no manual UI code (like App.jsx edits) is needed.

When you use the TrainingLoop API, this schema generation is handled under the hood. However, you can also define it explicitly via a YAML file or programmatically.

When to use

Use explicit schemas when you are building a custom integration using the Raw SDK Primitives and need to define how the dashboard should render your state and mutations.

YAML Format

Load your YAML schema at startup with load_param_schema("params.yaml").

params:
  - name: learning_rate   # key used in get_mutations()/publish_telemetry_raw
    type: float           # float | int | bool | string
    kind: slider          # slider | toggle | readonly | text
    min: 0.00001
    max: 0.01
    step: 0.00001
    default: 0.0003
    label: "Learning Rate" # optional, defaults to `name`
    group: "Optimizer"     # optional section heading in the dashboard
  - name: reward_mean
    type: float
    kind: readonly         # display-only metric, no mutation control

Field Reference

Field	Required?	Description
`name`	Yes	Key used in mutations and telemetry payloads.
`type`	Yes	How the dashboard parses the value: `float`, `int`, `bool`, or `string`.
`kind`	Yes	Control type: `slider` (requires min/max/step), `toggle` (for bools), `readonly` (display-only), or `text` (free-form entry).
`min`, `max`, `step`	No	Required if `kind` is `slider`. Defines the range and granularity.
`default`	No	Initial value shown in the dashboard before the first telemetry frame arrives.
`label`	No	Display name in the UI. Defaults to the `name` field if omitted.
`group`	No	Section heading to group related parameters together visually.

Programmatic Alternative

If you prefer defining schemas in code rather than YAML, use the declare_param function:

from omniloop import declare_param

declare_param(
    "learning_rate",
    type="float",
    kind="slider",
    min=0.00001,
    max=0.01,
    step=0.00001,
    default=0.0003,
    label="Learning Rate",
    group="Optimizer"
)