Adaline Datasets

Concepts

Datasets are structured test cases for evaluations. Rows supply prompt inputs and optional expected values. Columns define the cell modality or dynamic generation source.

Key terms:

• Dataset — table of evaluation cases in a project
• Column — named field; types are static, prompt, or api
• Row — one test case; values are keyed by column ID or column name
• Dynamic column — prompt or api column whose values are generated on demand

Configuration

Set these environment variables when credentials are available:

• ADALINE_API_KEY — workspace API key from Admin > API Keys
• ADALINE_PROJECT_ID — project ID

Base URL: https://api.adaline.ai/v2

Key Rule: Use Current Batch Shapes

Column creation takes { "columns": [...] }. Row creation takes { "rows": [...] }. Do not send a single bare column object to /columns.

Quick Triage

Creating a Dataset

curl -X POST "https://api.adaline.ai/v2/datasets" \
  -H "Authorization: Bearer $ADALINE_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "projectId": "project_abc123",
    "title": "Support eval set",
    "icon": { "type": "emoji", "value": "📚" },
    "description": "Support questions and expected answers"
  }'

Adding Columns

curl -X POST "https://api.adaline.ai/v2/datasets/dataset_abc123/columns" \
  -H "Authorization: Bearer $ADALINE_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "columns": [
      { "name": "question", "type": "static" },
      { "name": "expected_answer", "type": "static" },
      {
        "name": "draft_answer",
        "type": "prompt",
        "settings": { "promptId": "prompt_abc123" }
      }
    ]
  }'

API dynamic column:

{
  "name": "retrieved_context",
  "type": "api",
  "settings": {
    "method": "POST",
    "url": "https://example.com/retrieve",
    "headers": { "Authorization": "Bearer token" },
    "bodyTemplate": "{ \"query\": \"{{question}}\" }"
  }
}

Adding Rows

curl -X POST "https://api.adaline.ai/v2/datasets/dataset_abc123/rows?valuesBy=columnName" \
  -H "Authorization: Bearer $ADALINE_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "rows": [
      {
        "values": {
          "question": { "value": "How do I reset my password?" },
          "expected_answer": { "value": "Send the reset link." }
        }
      }
    ]
  }'

Dynamic Columns

curl -X POST "https://api.adaline.ai/v2/datasets/dataset_abc123/dynamic-columns/fetch" \
  -H "Authorization: Bearer $ADALINE_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "columnIds": ["column_abc123"],
    "datasetRowIds": ["row_abc123"],
    "runMode": "all"
  }'

runMode can be all, failed, or first.

SDK Usage

await adaline.datasets.list({ projectId, limit: 20 });
await adaline.datasets.create({ dataset });
await adaline.datasets.columns.create({ datasetId, columns });
await adaline.datasets.rows.create({ datasetId, valuesBy: 'columnName', rows });
await adaline.datasets.columns.fetchDynamic({ datasetId, query });

await adaline.datasets.list(project_id=project_id, limit=20)
await adaline.datasets.create(dataset=dataset)
await adaline.datasets.columns.create(dataset_id=dataset_id, columns=columns)
await adaline.datasets.rows.create(dataset_id=dataset_id, values_by="columnName", rows=rows)
await adaline.datasets.columns.fetch_dynamic(dataset_id=dataset_id, query=query)

Best Practices

1. Prefer valuesBy=columnName for authoring fixtures; use column IDs for immutable machine integrations.
2. Batch rows and columns when possible.
3. Add static input columns first, then rows, then dynamic columns.
4. Use datasetRowIds to rerun dynamic generation for a focused subset.
5. Keep dataset column names aligned with prompt variables where rows feed prompt evaluations.

References

See references/api.md for the full REST contract.