Skip to Content
Referenceoo CLI

oo models

Models agents can use.

Models agents can use. These commands call the /api/v1/models routes; oo api routes --path /api/v1/models shows their inputs and permissions.

oo models list

List models

oo models list [options]
ArgumentRequiredDefaultChoicesDescription
--query NAME=VALUEno

It also takes the connection options.

oo models create

Create one from a JSON body

oo models create [options]
ArgumentRequiredDefaultChoicesDescription
--json @FILE|@-yes
--output OUTPUTnowrite response to a protected file (required for a new credential)
--overwriteno

It also takes the connection options.

oo models update

Change one by ID from a JSON body

oo models update [options] id
ArgumentRequiredDefaultChoicesDescription
idyesID from oo models list
--json @FILE|@-yes
--output OUTPUTnowrite response to a protected file (required for a new credential)
--overwriteno

It also takes the connection options.

oo models delete

Delete one by ID

oo models delete [options] id
ArgumentRequiredDefaultChoicesDescription
idyesID from oo models list
--output OUTPUTnowrite response to a protected file (required for a new credential)
--overwriteno

It also takes the connection options.

oo models providers

The provider credentials this Sunport's models use, as Models › Providers shows them: each a provider's API key, or the fields its provider takes, under a name that models reference (litellm_credential_name in oo models create). The model proxy keeps them encrypted and never shows them again: lists show a masked hint, and whether the proxy has each (synced). The values come from a file or standard input, never from arguments, as a JSON object of the provider's fields: api_key, and api_base for one at an address of its own; for Amazon Bedrock, aws_access_key_id, aws_secret_access_key and aws_region_name, and aws_session_token if it has one. Listing needs credentials.read, and the rest credentials.manage.

oo models providers [COMMAND]

It also takes the connection options.

oo models providers list

The provider credentials, with masked hints

oo models providers list

It also takes the connection options.

oo models providers add

Add a provider credential under NAME: 2 to 63 lowercase letters, digits, dashes and underscores. A model then uses it by that name.

oo models providers add [options] NAME
ArgumentRequiredDefaultChoicesDescription
NAMEyesthe credential's name
--provider PROVIDERyesthe provider, such as openai, anthropic or bedrock
--values @FILE|@-yesa JSON object of the provider's fields, from a file or standard input
--description DESCRIPTIONnowhat it is for, up to 500 characters

It also takes the connection options.

oo models providers rotate

Replace a provider credential's values, such as a rotated API key; the models that use it keep it by name.

oo models providers rotate [options] NAME
ArgumentRequiredDefaultChoicesDescription
NAMEyesthe credential's name
--values @FILE|@-yesa JSON object of the provider's fields, from a file or standard input

It also takes the connection options.

oo models providers remove

Remove a provider credential. It is refused while a model uses it (credential_in_use names those models).

oo models providers remove NAME
ArgumentRequiredDefaultChoicesDescription
NAMEyesthe credential's name

It also takes the connection options.

oo models catalog

Search the model catalog, as Add model does: the provider models a Sunport model can be made from, each with its mode, token limits and per-token prices, and the catalog's providers. --query takes q (text in the model's name), provider, mode, limit and offset; oo api route GET /api/v1/model-catalog lists them. Needs models.read.

oo models catalog [options]
ArgumentRequiredDefaultChoicesDescription
--query NAME=VALUEnoan input of the read; repeatable

It also takes the connection options.

oo models health

Check one model through the model proxy, as Models › Providers does, and print how many of its endpoints answered (healthy) and how many didn't (unhealthy), with their errors, never their settings. The check makes a real call to the provider, which can incur usage. Needs models.read.

oo models health MODEL
ArgumentRequiredDefaultChoicesDescription
MODELyesthe model's alias, from oo models list

It also takes the connection options.

oo models usage

Print the Models page's ledger as JSON: each model with its requests, tokens, cache use, cost and when it was last used, in the period, and what OpenOrange charged for it where it bills this Sunport (billedOnPlatform). The period is this month unless --query gives range (today, current_week or current_month), or from and to (its first and last day), and timeZone; provider, filter, search, sort and dir choose and order the models, as oo api route GET /api/v1/models/usage lists them. Needs billing.read. With --mine, the same ledger over your own requests only, made by you or for you through an agent: your allowed models with their prices and your use and spend, as Models shows a member. Previously used models remain visible with access_allowed=false after access is revoked. It needs only models.read, and a person's credential.

oo models usage [options]
ArgumentRequiredDefaultChoicesDescription
--minenoonly your own requests and spend
--query NAME=VALUEnoan input of the read; repeatable

It also takes the connection options.

oo models key

Get and keep your own key for the Sunport's models (OpenAI-compatible). A connected agent with model access has one; oo connect gets it already. Signed in as yourself, it makes a model key of your own, named --name (this computer by default), under your own allowed models and budget: administrators may, and members given models.key. The key is kept where your sign-in is and is not printed; --output FILE also writes it to a protected file once. Each has its own name, and you may have as many as you need.

oo models key [options]
ArgumentRequiredDefaultChoicesDescription
--replacenoreplace the current model key; the old one stops working first
--name NAMEnoa person's key: its name, by default 'oo on <this computer>'
--models MODEL,MODELnoa person's new key: only these of the models you may use
--output FILEnoalso write the key to this protected file, once
--overwritenowith --output: replace that file
--listnoa person's own model keys, never their secrets
--revoke NAMEnorevoke one of your own model keys, by name or ID

It also takes the connection options.

oo models env

Print shell export lines for the model key, for eval "$(oo models env)". They contain the key: never paste them into a chat.

oo models env

It also takes the connection options.

oo models exec

Run a command with OPENAI_BASE_URL and OPENAI_API_KEY set to this agent's model key, without printing it.

oo models exec ...
ArgumentRequiredDefaultChoicesDescription
commandnothe command and its arguments, after --

It also takes the connection options.

On this page