Skip to main content
POST
/
v1
/
embeddings
/
estimate
Estimate embedding credits
curl --request POST \
  --url https://api.aurous-labs.com/v1/embeddings/estimate \
  --header 'Content-Type: application/json' \
  --header 'X-Api-Key: <api-key>' \
  --data '
{
  "model": "aurous-embed-vision",
  "input": "<string>",
  "dimensions": 1024
}
'
{
  "estimated": true,
  "tokens": {
    "text": 5000,
    "image": 2000,
    "video": 0,
    "total": 7000
  },
  "credits_estimated": 0.19125,
  "breakdown": {
    "input": {
      "text": 0.09375,
      "visual": 0.0975,
      "video": 0
    },
    "model": "aurous-embed-vision"
  }
}

Authorizations

​
X-Api-Key
string
header
required

Your team API key (starts with al_live_).

Headers

​
Aurous-Version
string

Optional API version pin (YYYY-MM-DD). Defaults to your team's pinned version, or the system default 2026-05-15 for unauthenticated requests.

Pattern: ^\d{4}-\d{2}-\d{2}$
Example:

"2026-05-15"

Body

application/json
​
model
string
required

Public model slug (e.g. "aurous-embed-vision"). Pass exactly as listed by GET /v1/models.

Example:

"aurous-embed-vision"

​
input
required

Input — accepts a string OR an array of content parts ({type: "text"|"image_url"|"video_url"}) for multimodal. String-array (string[]) batch input is NOT accepted on v1 (same rules as POST /v1/embeddings).

​
dimensions
integer

Output vector dimensions. Most models return a fixed dimension and reject this parameter. If the model does not support dimensions, the estimate returns 400 embeddings_unsupported_dimensions — identical to the dispatch path, so an estimate and a real request fail at the same gate.

Example:

1024

Response

Estimate produced.

​
estimated
boolean
required

Always true — distinguishes this from a real /v1/embeddings response (which uses object: 'list').

Example:

true

​
tokens
object
required

Estimated token counts per modality + total.

​
credits_estimated
number
required

Estimated credits for this request. Real credits_charged from POST /v1/embeddings may differ slightly for URL-fetched media (image / video tokenization variance), but the estimate is generally a tight upper bound.

Example:

0.19125

​
breakdown
object
required

Per-modality credit decomposition + model echo.