API docs by Redocly

mf-model-manager large model APIs (0.1.0)

Download OpenAPI specification:

Large model management APIs maintained by mf-model-manager.

This document currently records the API contracts touched by the model management bug fixes in fix/model-management-bugs.

ModelManager

Large model management APIs

Test large model connection

Tests whether the submitted large model configuration can reach the upstream model service.

For model_series=openai, the service sends a real Azure OpenAI chat completions request to the configured deployment. Invalid api_url, api_model or api_key values return 400 instead of being reported as success.

Successful responses are normalized to { "status": "ok", "id": "..." }.

Authorizations:
BearerAuth
header Parameters
Accept-Language
string
Example: en-US

Preferred response language. Supports zh-CN and en-US.

Request Body schema: application/json
required
model_id
string

Existing large model id. Used when testing a saved model.

change
boolean
Default: false

Whether to test the submitted config instead of the saved config.

model_series
string

Model provider or protocol series.

model_type
string
Enum: "llm" "rlm" "vu"

Large model type.

object

Upstream model service configuration.

Responses

Request samples

Content type
application/json
Example
{
  • "model_id": "1234567890123456789"
}

Response samples

Content type
application/json
{
  • "status": "ok",
  • "id": "1234567890123456789"
}

Set or clear default large model

Sets or clears the default large model.

default=true marks the model as the default model and clears the previous default. default=false clears the default flag from the specified model. The response includes the resulting default state.

Authorizations:
BearerAuth
header Parameters
Accept-Language
string
Example: en-US

Preferred response language. Supports zh-CN and en-US.

Request Body schema: application/json
required
model_id
required
string

Large model id.

default
required
boolean

true sets the model as default; false clears its default flag.

Responses

Request samples

Content type
application/json
Example
{
  • "model_id": "1234567890123456789",
  • "default": true
}

Response samples

Content type
application/json
{
  • "status": "ok",
  • "id": "1234567890123456789",
  • "default": false
}

Get large model monitor overview

Returns large model usage metrics in the requested date range.

start_time and end_time must use YYYY-MM-DD format. start_time must be earlier than or equal to end_time; otherwise the service returns 400.

The aggregate result only includes records whose f_model_id still exists in t_llm_model, so non-current or stale model operation records are excluded from the overview.

Authorizations:
BearerAuth
query Parameters
start_time
required
string <date>
Example: start_time=2026-07-01
end_time
required
string <date>
Example: end_time=2026-07-20
model_id
string

Optional large model id. Empty means all visible large models.

header Parameters
Accept-Language
string
Example: en-US

Preferred response language. Supports zh-CN and en-US.

Responses

Response samples

Content type
application/json
{
  • "summary": {
    },
  • "trends": [
    ],
  • "qps_data": [
    ]
}