Skip to main content

Transforms

Transforms use JOLT (JSON to JSON transformation language) to reshape data as it flows through pipelines. Transform files are stored in the directory specified by transforms_path in config.toml.

Transform Files​

Transform files are JSON specifications stored in the transforms/ directory:

config.toml
transforms/
  ├── patient_to_fhir.json
  ├── observation_mapping.json
  └── dicom_metadata.json

Using Transforms in Pipelines​

Reference transform files in middleware configuration:

# pipelines/api.toml

[middleware.patient_transform]
type = "transform"

[middleware.patient_transform.options]
spec_path = "patient_to_fhir.json"  # Relative to transforms/ directory
apply = "left"                       # Apply to request
fail_on_error = true

Transform Options​

spec_path​

Path to the JOLT specification file (relative to transforms_path):

[middleware.my_transform.options]
spec_path = "patient_to_fhir.json"

apply​

When to apply the transform:

  • "left" - Apply to incoming requests (default)
  • "right" - Apply to outgoing responses
  • "both" - Apply to both requests and responses
[middleware.my_transform.options]
spec_path = "normalize.json"
apply = "both"

fail_on_error​

Whether to fail the request if transformation fails:

  • true - Fail the request (default, recommended)
  • false - Log error and continue
[middleware.my_transform.options]
spec_path = "optional_enrichment.json"
fail_on_error = false

JOLT Specification Format​

Transform files contain JOLT specifications in JSON format:

// transforms/patient_to_fhir.json
[
  {
    "operation": "shift",
    "spec": {
      "id": "identifier[0].value",
      "firstName": "name[0].given[0]",
      "lastName": "name[0].family",
      "birthDate": "birthDate"
    }
  }
]

Common Transform Patterns​

Field Mapping​

Map fields from one structure to another:

[
  {
    "operation": "shift",
    "spec": {
      "patient_id": "id",
      "patient_name": "name",
      "dob": "birthDate"
    }
  }
]

Nested Data​

Extract or restructure nested data:

[
  {
    "operation": "shift",
    "spec": {
      "patient": {
        "demographics": {
          "firstName": "name.given[0]",
          "lastName": "name.family"
        }
      }
    }
  }
]

Adding Static Values​

Add constant values to the output:

[
  {
    "operation": "shift",
    "spec": {
      "*": "&",
      "#fhir": "resourceType"
    }
  },
  {
    "operation": "default",
    "spec": {
      "resourceType": "Patient"
    }
  }
]

Array Handling​

Transform arrays of data:

[
  {
    "operation": "shift",
    "spec": {
      "observations": {
        "*": {
          "code": "entry[&1].resource.code",
          "value": "entry[&1].resource.value"
        }
      }
    }
  }
]

Transform Middleware Types​

Transform (JOLT)​

Standard JSON-to-JSON transformations:

[middleware.data_transform]
type = "transform"

[middleware.data_transform.options]
spec_path = "mapping.json"
apply = "left"
fail_on_error = true

Metadata Transform​

Transform request metadata (key-value pairs used for routing):

[middleware.metadata_transform]
type = "metadata_transform"

[middleware.metadata_transform.options]
spec_path = "metadata_mapping.json"
apply = "left"
fail_on_error = true

Multiple Transforms​

Chain multiple transforms in a pipeline:

[pipelines.multi_transform]
middleware = [
    "normalize_input",    # 1. Normalize incoming data
    "enrich_data",        # 2. Add additional fields
    "format_output"       # 3. Format for target system
]

[middleware.normalize_input]
type = "transform"
[middleware.normalize_input.options]
spec_path = "normalize.json"

[middleware.enrich_data]
type = "transform"
[middleware.enrich_data.options]
spec_path = "enrich.json"

[middleware.format_output]
type = "transform"
[middleware.format_output.options]
spec_path = "format.json"

Testing Transforms​

Test transforms using the echo backend:

[pipelines.transform_test]
endpoints = ["test_endpoint"]
middleware = ["test_transform"]
backends = ["echo"]

[middleware.test_transform]
type = "transform"
[middleware.test_transform.options]
spec_path = "my_transform.json"
apply = "left"

[backends.echo]
service = "echo"

Send test requests:

curl -X POST http://localhost:8080/test \
  -H "Content-Type: application/json" \
  -d '{"id": "123", "firstName": "John", "lastName": "Doe"}'

Automatic Download from Runbeam Cloud​

When using Runbeam Cloud integration, transform files are automatically downloaded:

  1. Configure middleware with transform ID:

    [middleware.patient_transform.options]
    spec_path = "01k81xczrw551e1qj9rgrf0319.json"

  2. Harmony downloads the specification from Runbeam Cloud

  3. File is saved to transforms/01k81xczrw551e1qj9rgrf0319.json

  4. Transform is ready to use

Hot Reload​

Transform files are automatically reloaded when changed:

  1. Edit a transform file in transforms/
  2. Save the file
  3. Harmony detects the change
  4. New transform applies to subsequent requests

No service restart required.

Validation​

Transform specifications are validated when loaded:

  • JSON syntax must be valid
  • JOLT operations must be recognized
  • Spec structure must be correct

Invalid transforms are rejected with error details in logs.

Examples​

See the harmony-proxy repository for complete examples:

  • transforms/patient_to_fhir.json - Patient data transformation
  • transforms/observation_mapping.json - Observation mapping
  • transforms/metadata_set_dimse_op.json - Metadata transformation

JOLT Resources​

Learn more about JOLT transformations:

Next Steps​