Development beta · test data

Spona

Data mapping you can trace.

A Windows development beta for visually mapping complex HL7 and JSON flows: repeated groups, separate rules for each direction and an optional LLM step with selected input.

WindowsHL7 ↔ JSONVisual mapping

Discuss an application

An interactive example illustrating data mapping in Spona.

SponaHL7 ↔ JSON

Complex formats · synthetic example

1patient2orders4results

Repeated segments become nested arrays. Each result retains its order and source.

Explore the data flow

A complex message. Clear rules.

Choose a transformation direction, trace connected fields or see how an invalid LLM output stops mapping.

SponaInteractive demonstration · synthetic data

Repeated segments into a nested structure.

HL7 2.5.1
Versioned profile
LLMnot needed
Local validation
Nested JSON

Demo profile ORU-JSON · rev. 21 patient · 2 orders · 4 results

Source HL7 2.5.1

Destination Nested JSON

R03Transformation trace

OBR[1] / OBX[1..3]patients[0].orders[0].results[0..2]

The each(OBR) / each(OBX) rule keeps all three results within the first order. Numeric values remain numbers.

Inspect the complete input and output

HL7 2.5.1

MSH|^~\&|TEST-LAB|TEST-SITE|TEST-APP|TEST-SITE|20261024104500+0200||ORU^R01|TEST-M001|P|2.5.1
PID|1||TEST-P001
ZSP|TEST-JIL|TEST-B03
OBR|1||TEST-O001|TEST-PANEL-A|||20261024103000+0200
OBX|1|NM|TEST-A||1.20|TEST-UNIT|||||F
OBX|2|NM|TEST-B||2.50|TEST-UNIT|||||F
OBX|3|NM|TEST-C||0|TEST-UNIT|||||F
OBR|2||TEST-O002|TEST-PANEL-B|||20261024104000+0200
OBX|1|NM|TEST-D||3.40|TEST-UNIT|||||F

Nested JSON

{
  "message": {
    "type": "ORU^R01",
    "version": "2.5.1",
    "controlId": "TEST-M001"
  },
  "patients": [
    {
      "id": "TEST-P001",
      "location": {
        "ward": "TEST-JIL",
        "bed": "TEST-B03"
      },
      "orders": [
        {
          "id": "TEST-O001",
          "observedAt": "20261024103000+0200",
          "results": [
            {
              "code": "TEST-A",
              "value": 1.2,
              "unit": "TEST-UNIT"
            },
            {
              "code": "TEST-B",
              "value": 2.5,
              "unit": "TEST-UNIT"
            },
            {
              "code": "TEST-C",
              "value": 0,
              "unit": "TEST-UNIT"
            }
          ]
        },
        {
          "id": "TEST-O002",
          "observedAt": "20261024104000+0200",
          "results": [
            {
              "code": "TEST-D",
              "value": 3.4,
              "unit": "TEST-UNIT"
            }
          ]
        }
      ]
    }
  ]
}

The reverse direction with separate rules.

JSON
Versioned profile
LLMnot needed
Local validation
HL7 2.5.1

Demo profile JSON-ORU · rev. 11 patient · 2 orders · 4 results

Source JSON

Destination HL7 2.5.1

BACK-R03Transformation trace

patients[0].orders[0].results[0..2]OBR[1] / OBX[1..3]

The example uses separate reverse rules to construct the HL7 message.

Inspect the complete input and output

JSON

{
  "message": {
    "type": "ORU^R01",
    "version": "2.5.1",
    "controlId": "TEST-M001"
  },
  "patients": [
    {
      "id": "TEST-P001",
      "location": {
        "ward": "TEST-JIL",
        "bed": "TEST-B03"
      },
      "orders": [
        {
          "id": "TEST-O001",
          "observedAt": "20261024103000+0200",
          "results": [
            {
              "code": "TEST-A",
              "value": 1.2,
              "unit": "TEST-UNIT"
            },
            {
              "code": "TEST-B",
              "value": 2.5,
              "unit": "TEST-UNIT"
            },
            {
              "code": "TEST-C",
              "value": 0,
              "unit": "TEST-UNIT"
            }
          ]
        },
        {
          "id": "TEST-O002",
          "observedAt": "20261024104000+0200",
          "results": [
            {
              "code": "TEST-D",
              "value": 3.4,
              "unit": "TEST-UNIT"
            }
          ]
        }
      ]
    }
  ]
}

HL7 2.5.1

MSH|^~\&|TEST-LAB|TEST-SITE|TEST-APP|TEST-SITE|20261024104500+0200||ORU^R01|TEST-M001|P|2.5.1
PID|1||TEST-P001
ZSP|TEST-JIL|TEST-B03
OBR|1||TEST-O001|TEST-PANEL-A|||20261024103000+0200
OBX|1|NM|TEST-A||1.2|TEST-UNIT|||||F
OBX|2|NM|TEST-B||2.5|TEST-UNIT|||||F
OBX|3|NM|TEST-C||0|TEST-UNIT|||||F
OBR|2||TEST-O002|TEST-PANEL-B|||20261024104000+0200
OBX|1|NM|TEST-D||3.4|TEST-UNIT|||||F

Selected input, structured output and type validation.

Selected fields
Versioned profile
LLMselected input
Local validation
Allowed fields

Demo profile SELECTED-JSON · rev. 1One optional LLM step per case

01 · Explicitly selected input

The model receives only selected fields.

{
  "selectedResults": [
    {
      "code": "TEST-A",
      "value": 1.2
    },
    {
      "code": "TEST-B",
      "value": 2.5
    }
  ],
  "instruction": "Summarise only the supplied test values."
}

PID identities, the MSH header and message routing remain in the local flow.

02 · Example structured output

The result first passes local validation.

{
  "summary": "Two test values were supplied.",
  "items": [
    {
      "code": "TEST-A",
      "value": 1.2
    },
    {
      "code": "TEST-B",
      "value": 2.5
    }
  ]
}

The example passes structure validation.

summary is text; items is a bounded array; each value is numeric. The result may be mapped only to configured output fields.

03 · Allowed mapping

model.summary → review.summary
model.items[*] → review.items[]

Ready to map the example

The model response is prepared demonstration data. This page sends no model requests; a local browser example checks the data types.

The examples explain development-beta capabilities. This is not a running Spona editor or a full HL7 conformance check.

What the development beta includes

Mapping, validation and control of AI steps.

HL7 and JSON in both directions

HL7→JSON, JSON→HL7 and same-format transformations use separate rules. Nested loops retain patient and order scope, while configuration records exact profile revisions and digests.

A visual graph of processing steps
Versioned mapping profiles

From the result back to the source

The source and output views help verify which field produced a value. Output validation and provenance highlighting support review of a specific profile and test message.

Source-content highlighting
Result and validation-message review

AI only in the selected step

Each case allows one optional model call. It receives only selected values, and text or structured JSON output is validated locally before mapping into allowed fields. Identities and routing remain controlled by rules.

Selected data passed to the model
Validation of the expected output shape

Example test flow

A laboratory message into a structured record.

Using a synthetic HL7 message, fields are selected for sample identifiers, observations, units and time. A profile transforms them into agreed JSON, with the origin of each value visible in the review.

Current stage

Spona is a functional development beta for local testing, not a deployed hospital integration. Broader FHIR support in the main editor and full HL7 conformance checking are part of further development.

How we approach integrations

Our own development

Related projects.

Available online

Working with HL7 messages or JSON records?

Describe the input format, desired output and rules the transformation needs to follow.

Let’s talk

Spona

A complex message. Clear rules.

Source HL7 2.5.1

Destination Nested JSON

HL7 2.5.1

Nested JSON

Source JSON

Destination HL7 2.5.1

JSON

HL7 2.5.1

The model receives only selected fields.

The result first passes local validation.

Mapping, validation and control of AI steps.

A laboratory message into a structured record.

Current stage

Related projects.

Zdravko.ai

Sonima

DTS Assistant

Working with HL7 messages or JSON records?

Spona

A complex message. Clear rules.

Source HL7 2.5.1

Destination Nested JSON

HL7 2.5.1

Nested JSON

Source JSON

Destination HL7 2.5.1

JSON

HL7 2.5.1

The model receives only selected fields.

The result first passes local validation.

Mapping, validation and control of AI steps.

A laboratory message into a structured record.

Current stage

Related projects.

Zdravko.ai ↗

Sonima ↗

DTS Assistant ↗

Working with HL7 messages or JSON records?

Zdravko.ai

Sonima

DTS Assistant