Embeddings

POST /v1/embeddings turns one or more texts into vectors of numbers. Two texts with similar meanings produce similar vectors: that is the basis of semantic search, document clustering and duplicate detection. The endpoint follows the OpenAI format, so the SDKs work unchanged.

POST https://api.realmoon.ai/v1/embeddings

• Authentication: an sk-moon-… key with the inference scope (see Authentication).
• Body: JSON, 8 MB at most.
• Response: a list object of vectors.

Parameters#Link to the Parameters section

Any other parameter is rejected with 400 unknown_parameter.

Example#Link to the Example section

curl https://api.realmoon.ai/v1/embeddings \
  -H "Authorization: Bearer $MOONAI_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "mistralai/mistral-embed-2312",
    "input": ["How do I cancel my contract?", "Store opening hours"]
  }'

import os
from openai import OpenAI

client = OpenAI(base_url="https://api.realmoon.ai/v1", api_key=os.environ["MOONAI_API_KEY"])

result = client.embeddings.create(
    model="mistralai/mistral-embed-2312",
    input=["How do I cancel my contract?", "Store opening hours"],
)
for item in result.data:
    print(item.index, len(item.embedding))

import OpenAI from 'openai';

const client = new OpenAI({ baseURL: 'https://api.realmoon.ai/v1', apiKey: process.env.MOONAI_API_KEY });

const result = await client.embeddings.create({
  model: 'mistralai/mistral-embed-2312',
  input: ['How do I cancel my contract?', 'Store opening hours'],
});
for (const item of result.data) console.log(item.index, item.embedding.length);

Response#Link to the Response section

{
  "object": "list",
  "model": "mistralai/mistral-embed-2312",
  "data": [
    { "object": "embedding", "index": 0, "embedding": [0.0123, -0.0456, 0.0789] },
    { "object": "embedding", "index": 1, "embedding": [-0.0311, 0.0042, 0.0517] }
  ],
  "usage": { "prompt_tokens": 18, "total_tokens": 18, "credits": 1 }
}

The vectors are shortened in the example: a real vector from this model has 1,024 numbers.

Headers: X-Moon-Request-Id, X-Moon-Prompt-Storage: none, X-Moon-Blur and, when Moon Blur replaced values, X-Moon-Blur-Redactions.

Moon Blur#Link to the Moon Blur section

When Moon Blur is on for the key (the default), the personal data it detects in the texts is replaced before sending to the model. Only the vectors of the affected passages change. The X-Moon-Blur header is on or off, and X-Moon-Blur-Redactions gives the number of values replaced. Text given as tokens is not read by Moon Blur. See Privacy.

Models#Link to the Models section

List as of November 5, 2026; the current list, with prices and vector sizes, comes from GET /v1/models?type=embedding.

Use the same model for your documents and for your queries: vectors from two different models cannot be compared.

Costs#Link to the Costs section

Before the call, the request holds the most the texts can count (one token per byte of text, plus a few tokens per text). It is then billed on the tokens actually counted, or on the actual cost reported by the provider. A request that fails costs nothing.

Limits#Link to the Limits section

Beyond the response size, the API tells you how many texts this model accepts in one request: split them across several calls.

Errors#Link to the Errors section

Key, balance and spend cap errors are in Errors.

See also#Link to the See also section

• Models: the dimensions, custom_dimensions and context_length fields.
• Privacy: what Moon Blur detects and does not detect.