[BETA] Create Model

View as markdown
POST /ai/model

Model files will be downloaded from Huggingface.

Name must be the exact name of the model on huggingface (ex: openai/gpt-oss-120b or ggml-org/gpt-oss-120b-GGUF).

If the model is under a license then you must provide a Huggingface access token for an account that signed the license agreement

application/json

Body Required

  • huggingface-token string

    Huggingface Token

  • name string

    Model name

    Minimum length is 1.

Responses

  • 200 application/json

    200

    Hide response attributes Show response attributes object
    • id string(uuid)

      Operation ID

    • reason string

      Operation failure reason

      Values are incorrect, unknown, unavailable, forbidden, busy, fault, partial, not-found, interrupted, unsupported, or conflict.

    • reference object

      Related resource reference

      Hide reference attributes Show reference attributes object
      • id string(uuid)

        Reference ID

      • link string

        Link to the referenced resource

      • command string

        Command name

    • message string

      Operation message

    • state string

      Operation status

      Values are failure, pending, success, or timeout.
POST /ai/model 
curl \
 --request POST 'https://api-ch-gva-2.exoscale.com/v2/ai/model' \
 --header "Content-Type: application/json" \
 --data '{"huggingface-token":"string","name":"string"}'
Request examples 
{
  "huggingface-token": "string",
  "name": "string"
}
Response examples (200) 
{
  "id": "string",
  "reason": "incorrect",
  "reference": {
    "id": "string",
    "link": "string",
    "command": "string"
  },
  "message": "string",
  "state": "failure"
}