Embeddings
Convert text into dense vector representations for semantic search, RAG, clustering and similarity. Compatible with the OpenAI Embeddings API.
POST
/v1/embeddingsSingle input
📋Request
1{
2 "model": "your-embedding-model",
3 "input": "The quick brown fox jumps over the lazy dog",
4 "encoding_format": "float"
5}Batched input
Send an array of strings to embed many inputs in one round-trip:
📋Batched request
1{
2 "model": "your-embedding-model",
3 "input": [
4 "First document",
5 "Second document",
6 "Third document"
7 ]
8}Parameters
modelstringrequired
Embedding model id (type: embedding).
inputstring | string[]required
Text to embed. Send an array to embed multiple inputs in a single call.
dimensionsinteger
Truncate output vectors to this many dimensions. Requires a matryoshka model.
encoding_formatstring
"float" (default) or "base64".
userstring
End-user identifier for analytics.
Response
📋200 OK
1{
2 "object": "list",
3 "data": [
4 {
5 "object": "embedding",
6 "index": 0,
7 "embedding": [0.0023, -0.018, 0.041, ...]
8 }
9 ],
10 "model": "your-embedding-model",
11 "usage": { "prompt_tokens": 10, "total_tokens": 10 }
12}✓
Matryoshka models
Models with the
matryoshka capability accept a dimensions parameter to truncate the output vector while preserving semantic quality — handy for reducing storage and search latency.