Embeddings API

The endpoint and authentication

Embeddings live on the OpenAI-compatible entry point and use the same key as chat: only the path changes.

The key goes in Authorization: Bearer sk-your-tokenator-key. There is no separate key for embeddings.

Your first request

curl https://api.tokenator.top/v1/embeddings \
  -H "Authorization: Bearer sk-your-tokenator-key" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "gemini-embedding-2",
    "input": "how to connect my own domain"
  }'

{
  "object": "list",
  "data": [
    {
      "object": "embedding",
      "index": 0,
      "embedding": [0.0023, -0.0147, 0.0091]
    }
  ],
  "model": "gemini-embedding-2",
  "usage": { "prompt_tokens": 6, "total_tokens": 6 }
}

The answer always has this shape: data in the order of the input strings, the Tokenator model name and a usage block with the token count. Whatever the upstream added of its own never reaches the client.

Which models compute vectors

Request parameters

A batch of strings in one request

The input field accepts an array. That is one request, one response and one round trip — noticeably faster for indexing a corpus than one string at a time. The vectors come back in the order of the strings, and each element's index field confirms the match.

curl https://api.tokenator.top/v1/embeddings \
  -H "Authorization: Bearer sk-your-tokenator-key" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "gemini-embedding-2",
    "input": ["the first paragraph", "the second paragraph", "the third paragraph"]
  }'

SDKs and ready-made examples

The endpoint is compatible with the OpenAI SDKs: only the base URL and the key change.

from openai import OpenAI

client = OpenAI(
    api_key="sk-your-tokenator-key",
    base_url="https://api.tokenator.top/v1",
)

result = client.embeddings.create(
    model="gemini-embedding-2",
    input=["the first paragraph", "the second paragraph"],
)

print(len(result.data[0].embedding))

import OpenAI from "openai"

const client = new OpenAI({
  apiKey: "sk-your-tokenator-key",
  baseURL: "https://api.tokenator.top/v1",
})

const result = await client.embeddings.create({
  model: "gemini-embedding-2",
  input: "how to connect my own domain",
})

console.log(result.data[0].embedding.length)

The base URL and the key are the same as for chat, so one client can hold both chat and embeddings.

How embeddings are billed

Exactly like ordinary tokens. The prompt_tokens figure from the provider's answer is multiplied by the model's multiplier and drawn from the key's limit — there is no separate bundle for embeddings. A request has no output, so you pay for the input text only.

The spending shows up in the key dashboard next to chat, and the remainder in /v1/tokens.

Error codes

FAQ

Read next