Embeddings & Reranking
Vector representations and relevance scoring
Embeddings and reranking APIs are on the Token Hub roadmap. They will follow the same principles as the rest of the platform: one OpenAI-compatible endpoint, transparent per-token pricing, and metering with micro-dollar precision.
Planned surface
| Endpoint | Purpose |
|---|---|
POST /v1/embeddings | Vectorize text for search, clustering, classification, and RAG. |
POST /v1/rerank | Score documents against a query to reorder retrieval results. |
The embeddings API will accept the familiar input (string or array),model, and optional dimensions parameters, and return vectors in OpenAI's response shape so existing vector-store integrations keep working.
Until then
Coming soonThe /v1/embeddings endpoint is not yet live. Chat models remain available in the meantime — see Text Models. We recommend pausing embedding-dependent features rather than shimming around the gap, so you can switch over with one base-URL change when it ships.