Skip to main content

Workflow inputs

Workflow inputs allow you to parameterize workflows with typed values. Inputs are defined at the workflow level and can be provided when the workflow is triggered manually or by schedules.

Schema​

See Schema Reference for the complete #StringInput, #BoolInput, and #IntInput definitions.

Input Types​

Workflows support three input types, each with type-safe default values:

  • string: Text values.
  • bool: Boolean true/false values.
  • int: Integer numeric values.

Input Definition​

Inputs are defined as a map where keys are input names and values are Input objects.

Example with all types:

inputs:
  model_type:
    type: string
    required: true
    description: Type of ML model to train

  debug_mode:
    type: bool
    default: false
    description: Enable debug logging

  epochs:
    type: int
    default: 100
    description: Number of training epochs

Input Properties​

Common Properties​

These properties are available for all input types:

type (required)​

The data type of the input.

Type: "string" | "bool" | "int"

Example:

type: string  # or bool, or int

required (optional)​

Whether this input must have a value.

Type: bool

Default: false

Example:

required: true

Behavior:

  • If default is provided, the input always has a value (satisfied by the default).
  • If no default, the input must be provided by the trigger.

description (optional)​

Human-readable description of the input.

Type: string

Example:

description: Model architecture (xgboost, random-forest, etc.)

String Input Specific​

String inputs support an additional secret property.

secret (optional)​

Whether this input contains sensitive data that should be masked in logs and UI.

Type: bool

Default: false

Only available for: type: string

Example:

inputs:
  api_key:
    type: string
    secret: true

Type-Specific Defaults​

Default values must match the input type:

String defaults:

environment:
  type: string
  default: "production"  # String value

Boolean defaults:

dry_run:
  type: bool
  default: false  # Boolean value (true or false)

Integer defaults:

max_retries:
  type: int
  default: 3  # Integer value (no quotes)

Using Inputs​

Inputs are referenced in expressions as ${{ .inputs.<name> }} - see Expressions.

Type Coercion in Expressions​

When inputs are used in expressions (${{ .inputs.<name> }}), they are automatically converted to strings:

See Expressions for more details on type coercion.

In Environment Variables​

inputs:
  model_type:
    type: string
    required: true

env:
  MODEL_TYPE: "${{ .inputs.model_type }}"

In Shell Commands​

steps:
  - name: Train model
    run: python train.py --model ${{ .inputs.model_type }}

In Drive Paths​

upload:
  path: models/
  destination: "drive://bucket/${{ .inputs.model_type }}/models/"

Providing Input Values​

Schedule Triggers​

Schedules provide input values through the inputs field:

inputs:
  dataset:
    type: string
    required: true

trigger:
  schedule:
    - cron: "0 0 * * *"
      inputs:
        dataset: full-training-set
    - cron: "0 12 * * *"
      inputs:
        dataset: validation-set

Manual Triggers​

Input values are provided through the trigger mechanism (UI, API, CLI) when manually starting the workflow.

Feedback

Related content

Expressions
Trigger configuration
Secrets