Skip to content

Listing Models

You can use the models endpoints to find which models and Assistants you can use, and the values to put in a request's model field.

There are two endpoints:

  • GET /onveris/v1/models — OnVeris's own list, with details such as context size and image support.
  • GET /v1/models — the same models and Assistants in OpenAI's format, for libraries that expect it.

Both endpoints require the models.read scope. See Authentication.

GET /onveris/v1/models

The response has three lists: language models, embedding models and Assistants.

Example response:

{
  "object": "list",
  "chat_models": [
    {
      "id": "fp:9189ab12cd34ef56",
      "name": "Qwen3.5-9B",
      "fingerprint": "fp:9189ab12cd34ef56",
      "object": "chat_model",
      "owned_by": "optima",
      "ctx_max": 32768,
      "supports_thinking": true,
      "input_modalities": ["text", "image"],
      "description": null
    }
  ],
  "embed_models": [
    {
      "id": "fp:4be01c7d9a2f3e10",
      "name": "nomic-embed-text-v1.5",
      "fingerprint": "fp:4be01c7d9a2f3e10",
      "object": "embedding_model",
      "owned_by": "optima",
      "dimensions": 768,
      "description": null
    }
  ],
  "assistants": [
    {
      "id": "assistant:hr-helper",
      "assistant_id": "0b6c…",
      "name": "HR Helper",
      "description": "Answers questions about leave, benefits and policy.",
      "icon": "…",
      "colour": "…",
      "object": "assistant",
      "owned_by": "optima",
      "available": true,
      "enabled_for_chat": true,
      "enabled_for_api": true,
      "model": "Qwen3.5-9B",
      "uses_fallback": false,
      "ctx_max": 32768,
      "supports_thinking": true,
      "input_modalities": ["text", "image"],
      "attachments_allowed": true,
      "agents": []
    }
  ]
}

Chat Model Fields

Field Description
id The value to use as model in requests. A fingerprint of the exact model build, starting fp:
name The model's name. Also accepted as model
fingerprint The same fingerprint as id
object Always "chat_model"
owned_by Always "optima", the product's name before it was renamed. It is kept so that existing integrations keep working
ctx_max The context window in tokens — the most one request can hold
supports_thinking Whether the model reasons before answering, so reasoning_effort has an effect. null if unknown
input_modalities What the model reads, such as ["text", "image"]. null if unknown
description Optional description set by your administrator

Embedding Model Fields

Field Description
id The value to use as model in embedding requests
name The model's name
fingerprint The same fingerprint as id
object Always "embedding_model"
owned_by Always "optima", the product's name before it was renamed. It is kept so that existing integrations keep working
dimensions The length of the vectors the model produces
description Optional description

Assistant Fields

Field Description
id The value to use as model, such as "assistant:hr-helper"
assistant_id The Assistant's internal ID
name, description, icon, colour How the Assistant is presented
object Always "assistant"
available Whether one of its models is running, so it can answer now
enabled_for_chat, enabled_for_api Where it can be used. Only Assistants enabled for the API can be called through it
model The model answering for it now
uses_fallback true when a fallback model is answering because its main model is not running
ctx_max, supports_thinking, input_modalities As for chat models, for the model answering now
attachments_allowed Whether it accepts attached documents and images
agents The other Assistants it can hand tasks to

GET /v1/models

This is the OpenAI-format list. Each model and API-enabled Assistant will appear once. For example:

{
  "object": "list",
  "data": [
    { "id": "fp:9189ab12cd34ef56", "object": "model", "created": 0, "owned_by": "optima" },
    { "id": "assistant:hr-helper", "object": "model", "created": 1760000000, "owned_by": "optima" }
  ]
}

Which Models Appear

Not every model and Assistant on the Gateway will be listed for you:

  • Only models running on a connected Node are listed. If a model you expect is missing, ask your administrator to check that it is started.
  • Only Assistants shared with you are listed.
  • If your administrator has turned on Only allow assistants, language models are not listed at all, unless you are an Assistant administrator. Use an Assistant instead.

Model IDs And Names

A model's id identifies one exact build of the model. If two Nodes have different files under the same name, the id still picks the right one, so it is the safest value to use. A model's name also works, as long as only one build has that name.