Developer API

ZeroTrain Data Inputs

ZeroTrain evaluates structured business data using deterministic rule-based logic. Both InputSource (runtime data) and ModelSource (rules and configuration) accept the same structured formats.

Important: ZeroTrain is format-agnostic at the boundary. All supported input formats are internally normalized into a consistent tabular representation (rows and columns) before deterministic evaluation.

Supported Data Formats

Format Description
JSON Object Standard structured key-value payload
JSON Array (Table) Array of objects representing tabular rows
Tabular Result Set Row/column records produced by databases, APIs, or exports (not SQL query text)
Dataverse Dataverse table records (e.g., "List rows") normalized into tabular format
ADO.NET / DataTable .NET structured tabular records
XML Structured hierarchical document converted to tabular representation
CSV Comma-separated tabular data
Base64 Encoded Encoded representation of any supported structured format

Example: JSON Object

{
  "InputSource": {
    "CustomerTenureMonths": 48,
    "AccountStatus": "Active",
    "SentimentScore": -0.62
  }
}

Example: JSON Array (Tabular)

{
  "InputSource": [
    { "CustomerId": 1, "Spend": 200, "Status": "Active" },
    { "CustomerId": 2, "Spend": 450, "Status": "Active" }
  ]
}

Example: Tabular Result Set (Rows & Columns)

Columns: CustomerId | Spend | Status
Rows:
  1 | 200 | Active
  2 | 450 | Active

Note: ZeroTrain does not ingest SQL query statements. It ingests the resulting structured records (rows and columns).

Example: Dataverse Records

Dataverse Table: Accounts

Columns: AccountId | Revenue | Status
Rows:
  A001 | 150000 | Active
  A002 |  85000 | Inactive

Example: CSV

CustomerId,Spend,Status
1,200,Active
2,450,Active

Example: XML

<Customers>
    <Customer>
        <CustomerId>1</CustomerId>
        <Spend>200</Spend>
        <Status>Active</Status>
    </Customer>
    <Customer>
        <CustomerId>2</CustomerId>
        <Spend>450</Spend>
        <Status>Active</Status>
    </Customer>
</Customers>

Example: Base64 Encoded

Q3VzdG9tZXJJZCxTcGVuZCxTdGF0dXM=

(Base64 representation of CSV or JSON payload)

Core Design Principle

Regardless of origin or format, all supported data types are normalized into a deterministic internal tabular model prior to rule evaluation.

This ensures:

  • ✔️ Consistent inference behavior across formats
  • ✔️ Deterministic execution
  • ✔️ No format-dependent logic paths
  • ✔️ Portable execution (InMemory / ONNX)
  • ✔️ Enterprise-ready integration across systems