Embeddings
Available
Embedding generation alongside inference, in the same region, so retrieval does not become the slow part of a live conversation.
What you get
- Batch and streaming generation
- Co-located with inference to keep retrieval inside the turn budget