LLMModelTestRequest

External LLM endpoint connection test request. Supports both pre-save (form base_url + api_key) and post-save (model_id with api_key omitted → server loads the stored key). A body api_key always takes precedence over the stored one.

providerProvider (string)필수

Provider identifier to test.

base_urlBase Url (string)필수

Base URL of the external model endpoint to test.

model object

Model identifier whose availability should be checked.

하나 이상(anyOf)

• string
• null

string

api_key object

Write-only API key used for this test. Omit it with model_id to use that model's stored key.

하나 이상(anyOf)

• string
• null

string

model_id object

Existing model identifier whose stored API key should be used for a post-save test.

하나 이상(anyOf)

• string
• null

string

default_headers object

Write-only custom header rows used for this test. Send them to test a model before it is saved. When the field is omitted, a test that uses a stored API key probes with the saved headers; when it is sent, these rows replace the saved headers so the test verifies the state about to be stored. An explicit empty list probes without custom headers.

하나 이상(anyOf)

• object[]
• null

Array [

nameName (string)필수

wire header name

valueValue (string)필수

header value

secretSecret (boolean)

Whether to store the header value as a write-only secret. When it is false, the value is stored as a plain-text constant and is returned by read operations. The default is false.

기본값: false

]

collection_id object

Identifier of the containing collection.

하나 이상(anyOf)

• string
• null

string

verify_inferenceVerify Inference (boolean)

Whether to run one minimal inference request after the model list is retrieved successfully. Each request consumes upstream usage, so send it only for an explicit user action. The result is reported separately in inference. The default is false.

기본값: false

message object

User message sent when verify_inference is true. It defaults to ping when omitted. Generation parameters cannot be overridden here, so the stored model settings apply.

하나 이상(anyOf)

• string
• null

string

Possible values: <= 2000 characters

measure_ttftMeasure Ttft (boolean)

Whether to stream the verify_inference round trip and report the time to the first token in inference.ttft_ms. A non-streaming call reports a single timestamp and cannot separate waiting from decoding. The default is false.

기본값: false

LLMModelTestRequest

{
  "provider": "string",
  "base_url": "string",
  "model": "string",
  "api_key": "string",
  "model_id": "string",
  "default_headers": [
    {
      "name": "string",
      "value": "string",
      "secret": false
    }
  ],
  "collection_id": "string",
  "verify_inference": false,
  "message": "string",
  "measure_ttft": false
}