MODEL DISCOVERY

List models

GET/v1/models

Return the deployed DarkSeek model registry. Use a returned identifier in generation requests. Authentication, rate and concurrency policy apply. The gateway constructs this registry instead of exposing internal backend paths or aliases.

Request

No body or query parameters. Send a Bearer key in the Authorization header.

curl "$DARKSEEK_BASE_URL/v1/models" -H "Authorization: Bearer $DARKSEEK_API_KEY"

Response · 200

{
  "object": "list",
  "data": [
    {
      "id": "darkseek-v4.1",
      "object": "model",
      "owned_by": "darkseek",
      "name": "DarkSeek V4.1",
      "max_output_tokens": 16384,
      "output_modalities": [
        "text"
      ]
    }
  ]
}

Illustrative configured deployment. The output ceiling is deployment policy; query your actual endpoint for its current value. This is not a live production inventory result. owned_by identifies the serving namespace, not ownership of the underlying model weights.

Field	Type	Meaning
object	string	Always list.
data	model[]	Available model records.
data[].id	string	Identifier accepted as model.
data[].object	string	Always model.
data[].owned_by	string	Serving namespace.
data[].name	string	Display label.
data[].max_output_tokens	integer	Maximum accepted output-token request limit.
data[].output_modalities	string[]	Configured output modalities.

Extended capability metadata

Context windows, input modalities, reasoning levels and protocol-specific behavior must come from verified DarkSeek capabilities. Omitted metadata means unspecified, not automatically supported. Full extended model discovery remains an acceptance gate.

Errors

401: invalid authentication. 429: rate or capacity limit. A disconnected web-to-Go binding returns 503. Model enumeration alone does not prove that inference is healthy.