Embeddings

POST /v1/embeddings — vector representations of text for search, RAG and clustering. Six models plus a legacy alias, 768 to 4096 dimensions. OpenAI-compatible; you pay for input tokens only.

Embeddings turn text into a numeric vector for semantic search, RAG, clustering and recommendations. The endpoint is fully OpenAI-compatible — only the base address and key change.

POSThttps://getvibeapi.com/v1/embeddings

Body parameters

Parameter	Type	Description
`model`req.	string	The embedding model (see below).
`input`req.	string \| string[]	A text or an array of texts. An array = a batch in one request.
`encoding_format`	"float" \| "base64"	The vector format in the response. Defaults to float.
`dimensions`	integer	Optional output width. For text-embedding-3-small, only 1536 is supported.

Models and price

Parameter	Type	Description
`text-embedding-3-small`		POST /v1/embeddings — vector representations of text for search, RAG and clustering. Model qwen3-embedding-8b. OpenAI-compatible; you pay for input tokens only.
`qwen3-embedding-8b`		POST /v1/embeddings — vector representations of text for search, RAG and clustering. Model qwen3-embedding-8b. OpenAI-compatible; you pay for input tokens only.
`bge-m3`		POST /v1/embeddings — vector representations of text for search, RAG and clustering. Model qwen3-embedding-8b. OpenAI-compatible; you pay for input tokens only.
`multilingual-e5-large`		POST /v1/embeddings — vector representations of text for search, RAG and clustering. Model qwen3-embedding-8b. OpenAI-compatible; you pay for input tokens only.
`embeddinggemma-300m`		POST /v1/embeddings — vector representations of text for search, RAG and clustering. Model qwen3-embedding-8b. OpenAI-compatible; you pay for input tokens only.
`nemotron-embed-vl`		POST /v1/embeddings — vector representations of text for search, RAG and clustering. Model qwen3-embedding-8b. OpenAI-compatible; you pay for input tokens only.
`qwen3-embedding-0.6b`		POST /v1/embeddings — vector representations of text for search, RAG and clustering. Model qwen3-embedding-8b. OpenAI-compatible; you pay for input tokens only.

Vector width differs per model, and you cannot swap models after the fact without rebuilding the index: the database column is created for one specific number of dimensions. Choose the model before you load the first million vectors.

Example request

curl https://getvibeapi.com/v1/embeddings \
  -H "Authorization: Bearer $GETVIBE_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "qwen3-embedding-8b",
    "input": "Text to vectorize"
  }'

Response

{
  "object": "list",
  "data": [
    { "object": "embedding", "index": 0, "embedding": [0.0021, -0.014, ...] }
  ],
  "model": "qwen3-embedding-8b",
  "usage": { "prompt_tokens": 8, "total_tokens": 8 }
}

You are charged only for input tokens (prompt_tokens); embeddings have no output tokens. Pass an array of strings in input to vectorize a batch in one request.

← PreviousFiles API Next →Speech-to-text