Embeddings

Turn text into vectors with the cruxy-embed model.

POST /api/v1/embeddings takes one string or an array and returns a vector per input, ready to store and compare for search, clustering, or retrieval. It is OpenAI-compatible. Here is a single-string embedding:

curl https://api.cruxy.ai/api/v1/embeddings \
  -H "Authorization: Bearer $CRUXY_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "cruxy-embed",
    "input": "The quick brown fox"
  }'

const res = await fetch('https://api.cruxy.ai/api/v1/embeddings', {
  method: 'POST',
  headers: {
    Authorization: `Bearer ${process.env.CRUXY_API_KEY}`,
    'Content-Type': 'application/json',
  },
  body: JSON.stringify({
    model: 'cruxy-embed',
    input: 'The quick brown fox',
  }),
})
 
const { data } = await res.json()
const vector = data[0].embedding // number[]

The model

model must be "cruxy-embed". It is the only embedding model on the gateway.

Single string or array

input accepts either a single string or an array of strings. Pass an array to embed a batch in one call:

curl https://api.cruxy.ai/api/v1/embeddings \
  -H "Authorization: Bearer $CRUXY_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "cruxy-embed",
    "input": ["The quick brown fox", "jumps over the lazy dog"]
  }'

Each input maps to one entry in data, and the entry's index matches the position of its input.

Response shape

{
  "object": "list",
  "model": "cruxy-embed",
  "data": [
    {
      "object": "embedding",
      "index": 0,
      "embedding": [0.0023, -0.0117]
    }
  ],
  "usage": {
    "prompt_tokens": 8,
    "total_tokens": 8
  },
  "cruxy": {
    "request_id": "a1b2c3d4-e5f6-7890-abcd-ef1234567890",
    "cost": { "tier": "cruxy-embed" }
  }
}

The embedding array is shown truncated; a real vector holds many more floats. The cost object carries the same fields documented under Chat.

The embedding array holds the vector as floating-point numbers. Read a result at data[i].embedding, matching data[i].index to your input's position.

Errors

Embeddings returns 400 for a malformed request, 401 for a bad key, and 429 for rate_limited or budget_exhausted. See Errors.

Next steps

• Retrieval-augmented generation: index a corpus and answer from it, end to end.
• Rerank: order retrieved candidates against a query.
• Chat: completions across the tiers.
• Errors: the error shape.