Skip to main content
The /v1/embeddings endpoint converts one or more text strings into dense vector representations (embeddings). You can use these vectors to find semantically similar content, build retrieval-augmented generation (RAG) pipelines, cluster documents, or train classifiers. OpenOpen8 is compatible with the OpenAI Embeddings API format, so any OpenAI-compatible embedding client works without modification.

​
POST /v1/embeddings

​
Request body

​
model
string
required
The embedding model to use. For example, text-embedding-3-small, text-embedding-3-large, or text-embedding-ada-002. The available models depend on your configured channels.
​
input
string | string[]
required
The text to embed. Can be a single string or an array of strings. Each string is embedded independently. Arrays are useful for batch embedding multiple documents in one request.
​
encoding_format
string
The format of the returned embedding vectors. float returns an array of floating-point numbers; base64 returns a base64-encoded binary string. Defaults to float.
​
dimensions
integer
The number of dimensions for the output embedding vector. Supported only by certain models (e.g., text-embedding-3-small and text-embedding-3-large). Truncates the embedding to the specified length.
​
user
string
An optional identifier for the end user making the request. Used for monitoring and abuse detection on the provider side.

​
Response

​
object
string
Always "list".
​
model
string
The model that generated the embeddings.
​
data
object[]
An array of embedding objects, one per input string.
​
usage
object
Token usage for the request.

​
Examples

curl https://openopen8.ai/v1/embeddings \
  -H "Authorization: Bearer YOUR_TOKEN" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "text-embedding-3-small",
    "input": [
      "OpenOpen8 is a unified AI gateway.",
      "You can embed multiple strings in one request."
    ],
    "encoding_format": "float"
  }'

​
Common use cases

  • Semantic search — embed your document corpus and a user query, then rank documents by cosine similarity to the query vector.
  • Retrieval-augmented generation (RAG) — retrieve the most relevant chunks from a knowledge base before passing them to a language model.
  • Clustering — group semantically related documents without labeled training data.
  • Classification — use embedding vectors as features for downstream classifiers.