organization-pricing

Create Organization Pricing

Changed on

Set the organization's rate for a model over a period.

Refused with a 409 when the period overlaps one already stored for that model, naming the period it collides with, rather than shadowing it. Refused with a 403 when the deployment, not the caller's organization, holds the credential that serves the model, whether through one of its own provider instances or a hosted credential the bound port supplies because no usable BYO credential of the organization's own covers every one of its workspaces: either way the deployment settles the upstream bill, so its rate is the deployment price list's rather than a tenant's.

The key is normalized to its canonical instance:model form first, the same call POST /api/v1/pricing makes, and that is what makes one model one row rather than one per spelling. Stored verbatim, openai:gpt-4o and openai/gpt-4o are two keys: the overlap rule would not see them as colliding, and both would resolve, with the canonical one preferred, leaving the other dormant until the first is deleted. Normalizing on the way in is what stops that pair existing at all.

post/api/v1/organizations/me/pricing

Request

  • The document declares no server URL.
  • Auth: one of:
    • API key in header Otari-Key
    • API key in header x-api-key

Request body

cache_read_price_per_millionnumber nullable

Price per 1M cached-input tokens

cache_write_1h_price_per_millionnumber nullable

Price per 1M Anthropic 1-hour cache-write tokens

cache_write_price_per_millionnumber nullable

Price per 1M cache-write (creation) tokens

effective_fromstring date-time nullable

ISO 8601 datetime from which this rate applies, inclusive. Defaults to now.

effective_tostring date-time nullable

ISO 8601 datetime at which this rate stops applying, exclusive. Null leaves it open ended. Because the end is exclusive, the next period may begin at exactly this instant without overlapping.

input_price_per_millionnumber required

Price per 1M of the rate's unit: input tokens, requests or images

model_keystring required

Model identifier in 'provider:model' form, matching the key the deployment price list uses. A provider instance name is valid here ('home_lab:llama-3'), because pricing keys on the instance a request resolves to.

output_price_per_millionnumber required

Price per 1M output tokens; unused by a rate in requests or images

unit'tokens' | 'requests' | 'images'

What the rates are per: tokens for a model, requests or images for a non-token endpoint, each priced per million. A completion model priced per requests is charged one request per successful call.

Response

Successful Response

cache_read_price_per_millionnumber nullable required
cache_write_1h_price_per_millionnumber nullable required
cache_write_price_per_millionnumber nullable required
created_atstring date-time required
effective_fromstring date-time required
effective_tostring date-time nullable required
idstring uuid required
input_price_per_millionnumber required
model_keystring required
organization_idstring uuid required
output_price_per_millionnumber required
unitstring required
updated_atstring date-time required

Changes