Field Definitions

Payload Structure

The Supervised AI testing API utilizes a structured JSON payload for all requests and responses. This ensures consistency across model evaluations and automated testing workflows.

Request Fields

The following fields define the configuration and input data for a testing request.

| Field | Type | Required | Description | | :--- | :--- | :--- | :--- | | model_id | string | Yes | The unique identifier for the model being tested within the Supervised AI ecosystem. | | version | string | No | Specifies the semantic version of the model. If omitted, the API defaults to the latest tagged version. | | input_data | object | array | Yes | The core data payload for the model. This can be a single JSON object for individual inference or an array of objects for batch processing. | | parameters | object | No | A key-value map of model-specific hyper-parameters (e.g., temperature, top_p, max_tokens). | | metadata | object | No | Custom key-value pairs used for tracking, experiment versioning, or internal tagging. | | callback_url | string | No | A valid HTTPS URL where the platform will send an HTTP POST request once the testing process is complete (useful for asynchronous tasks). |

Example Request Payload

{
  "model_id": "llm-classifier-v4",
  "version": "1.2.0",
  "input_data": {
    "text": "The model performance on edge cases was exceptional.",
    "context": "Customer Feedback"
  },
  "parameters": {
    "confidence_threshold": 0.85,
    "debug_mode": false
  },
  "metadata": {
    "environment": "staging",
    "test_suite": "regression-q3"
  }
}

Response Fields

The API returns the following fields upon successful processing of a testing task.

| Field | Type | Description | | :--- | :--- | :--- | | request_id | uuid | A unique identifier for the specific request, used for logging and debugging. | | status | string | The current state of the request. Possible values: queued, processing, completed, failed. | | results | array | An array containing the model's predictions or evaluation scores corresponding to the input_data. | | metrics | object | Technical performance metrics, including latency_ms and memory_usage_mb. | | timestamp | string | The ISO 8601 formatted timestamp of when the request was completed. |

Example Response Payload

{
  "request_id": "550e8400-e29b-41d4-a716-446655440000",
  "status": "completed",
  "results": [
    {
      "label": "positive",
      "confidence": 0.982
    }
  ],
  "metrics": {
    "latency_ms": 142.5,
    "compute_region": "us-east-1"
  },
  "timestamp": "2023-10-27T10:00:00Z"
}

Data Types and Constraints

To ensure compatibility with the Supervised AI platform, adhere to the following constraints:

• Strings: All string fields must be UTF-8 encoded.
• Model IDs: Must match the regex pattern ^[a-zA-Z0-9-_]+$.
• Input Size: The total size of the input_data object must not exceed 10MB per request for synchronous calls. For larger datasets, utilize the asynchronous callback_url pattern.
• Enums: The status field is restricted to the values defined in the response table above. Any internal state transitions are abstracted from the public API.