Skip to main content

Schema Files

Schema files (schema.json) define the user interface for Drona Workflows through JSON configurations that generate dynamic forms. These files control what users see, how they interact with the workflow, and what data gets collected for job execution.

Schema Structure

Schema File Organization​

Every Drona Workflow must contain a file called schema.json in its root directory. This file defines the complete user interface for the workflow and serves as the entry point for form generation. Schema can be described via multiple files as described in the Schema Decomposition section.

Standard Field Structure​

Each field in a schema follows a consistent structure with required and optional properties:

{
  "fieldName": {
    "type": "text",
    "name": "formFieldName", 
    "label": "Display Label",
    "help": "Descriptive help text for users"
  }
}

The type property determines which form component will be rendered, while name specifies the field identifier used in form submission. The label provides user-facing text, and help offers additional guidance through tooltips or inline descriptions. Elements often have optional properties that differ from component to component and can be found in the Form Components.

Component Types​

Schema files can utilize all available Frontend Form Components, including text inputs, selectors, file pickers, and specialized components like module selectors and time inputs. Each component type has specific properties and behaviors documented in the frontend components section.

Components can be organized using containers such as rowContainer and collapsibleRowContainer that group elements together for conditional display or stylistic purposes to control layout. Containers enable better organization of related fields and create intuitive user interfaces that guide users through workflow configuration.

Dynamic Content Integration​

Schema files integrate dynamic content through retriever scripts that populate field options or display real-time information. The script provided will be executed when the field loads and whenever user input triggers an update. For detailed information about creating and configuring retriever scripts, see Retriever Scripts.

{
  "dynamicField": {
    "type": "autocompleteSelect",
    "retriever": "scripts/retrievers/search_data.py",
    "placeholder": "Search for options..."
  }
}

Conditional Logic​

Schema files support conditional field display based on user selections, allowing workflows to show relevant options while hiding unnecessary complexity. Conditional logic is evaluated when the form loads and whenever referenced field values change. For more detailed information about conditional expressions and advanced patterns, see Conditional Logic.

{
  "operationType": {
    "type": "select",
    "name": "operation",
    "options": [
      {"value": "basic", "label": "Basic Mode"},
      {"value": "advanced", "label": "Advanced Mode"}
    ]
  },
  "advancedOptions": {
    "type": "text",
    "name": "advancedConfig",
    "label": "Advanced Configuration",
    "condition": "operation.advanced"
  }
}

This creates adaptive interfaces that respond to user choices and provide contextual configuration options through sophisticated workflow branching.

Schema Decomposition​

For complex workflows, schema files can be decomposed into multiple files using JSON references. This approach allows for better organization and reusability of form components across different workflow sections.

{
  "workflowSelector": {
    "type": "select",
    "name": "operationType",
    "options": [
      {"value": "inference", "label": "Inference"},
      {"value": "training", "label": "Training"}
    ]
  },

  "inferenceContainer: {
    "elements": {
      "inferenceSection": {
        "$ref": "/absolute/path/schema_components/inference.json",
      }
   },
   "condition": "operationType.inference"

  },

  "inferenceContainer: {
    "elements": {
      "trainingSection": {
        "$ref": "schema_components/training.json#/modelContainer", 
      } 
    },
    "condition": "operationType.training"
  }
}

A reference element can either reference an entire json file such as inference.json, or a specific element as in the modelContainer following the standard conventions of jsonref. The path used for references is relative to the environment directory, but can also be an absolute path. Current limitation involves that if an element is a reference element, it cannot include additonal fields such as condition.

This modular approach enables component reuse and maintainable schema organization for workflows with multiple execution paths or complex configuration requirements.

Texas A&M University High Performance Research Computing