content-model-test-data

Content Model Test Data Generation

Generate realistic, client-friendly synthetic test content that strictly follows a provided generic JSON content model and is safe for programmatic ingestion by downstream agents.

When to Use

• You have a content model and need valid sample data quickly.
• You need data that respects field constraints (type, min/max, required, enums, formats).
• You want domain-relevant data based on project context (for example, healthcare, ecommerce, finance).
• You need output as clean JSON for automated import pipelines.

Inputs

Provide:

• Content model definition as generic JSON field specifications
• Desired number of records
• Optional project context (README snippets, app purpose, domain hints)
• Optional locale, tone, and deterministic seed
• If a seed is provided, generation must be repeatable for the same inputs.
• If no seed is provided, generation may vary across runs.

Bundled Resources

• Example JSON output: example-output.json
• Digital badge block model example: digital-badge-block-model.json
• Digital badge block output example: digital-badge-block-output.json
• Output envelope validator: validate-output-envelope.ps1

CMS Table Normalization

If input comes from CMS property tables, normalize property types to generic JSON types before generation.

• String -> string
• Basic Rich Text -> richText
• Checkbox -> boolean

Keep original property names, required flags, translatable flags, defaults, and helper text as field metadata.

Output Contract (Required)

Return JSON only (no prose, no markdown fences) using this envelope:

{
  "meta": {
    "industry": "string",
    "industrySource": "inferred|provided|generic-default",
    "recordCount": 0,
    "modelName": "string",
    "generatedAt": "ISO-8601",
    "notes": []
  },
  "records": []
}

Rules:

• records must validate against the provided model.
• Keep field names exactly as defined by the model.
• Use values within min/max bounds and accepted formats.
• If context is missing, set industry to generic-business.
• Optional fields should be represented with a mixed strategy: omitted in some records and null in others.
• Do not include explanatory text outside JSON.

Procedure

1. Normalize the model.

   • Parse each field from generic JSON specs: type, required, enum/options, min/max, pattern, format, nested objects/arrays.
   • Build a generation plan per field with validation constraints.

2. Infer industry.

   • Prioritize explicit user-provided industry hints when present.
   • Else infer from project context (repository docs, naming, feature language).
   • Else default to generic-business.

3. Build value strategies by field type.

   • String: honor minLength/maxLength/pattern/format.
   • Number/Integer: honor minimum/maximum and precision.
   • Boolean: include realistic true/false variation.
   • Enum: sample valid options only.
   • Date/DateTime: produce valid ISO values and realistic timelines.
   • Object/Array: recursively satisfy child constraints and item counts.

4. Apply realism layer.

   • Map generated values to the inferred industry vocabulary.
   • Keep internal consistency (for example, status and timestamps align).
   • Use synthetic mock values only; keep them sensible and client friendly.

5. Generate records.

   • Produce the requested count.
   • Ensure every required field is present.
   • Apply mixed optional-field handling with configurable sparsity percentages:
     • Omit some optional fields.
     • Set some optional fields to null.
     • Populate remaining optional fields with valid values.

6. Validate before returning.

   • Check each record against all constraints.
   • Regenerate invalid fields/records until valid.
   • Return JSON envelope only.

7. Optionally validate the envelope with the bundled script.

   • Run: powershell -ExecutionPolicy Bypass -File ./scripts/validate-output-envelope.ps1 -InputPath ./output.json
   • Ensure meta.recordCount matches records length.

Decision Logic

• If model constraints conflict (for example min > max):
  • Return JSON with empty records and a note in meta.notes describing the conflict.
• If industry inference confidence is low:
  • Use generic-business and set industrySource to generic-default.
• If both user hint and inferred project industry exist:
  • Use the user hint and set industrySource to provided.
• If requested count is omitted:
  • Default to 10.
• If seed is provided:
  • Use deterministic generation so outputs are reproducible.

Quality Gates

A result is complete only if all are true:

• 100% of generated records conform to declared types and bounds.
• Required fields are never missing.
• Enum and pattern-constrained fields contain only valid values.
• Output is strict JSON, parseable without cleanup.
• meta.industry and meta.industrySource are populated.
• Optional fields follow the mixed omit/null strategy according to configured sparsity.

Example Prompt

Generate 25 records from this content model. Infer industry from the project context. Return JSON only in the required envelope.