Choose a model for the work you need to doMe
← Use cases
Embeddings & retrieval: how to choose a model
Embedding models represent content as numerical vectors for semantic retrieval. Embedding, retrieval, and reranking are distinct stages.
Before integrating
- 01Check input languages, batch size, and text length.
- 02Confirm embedding dimensions and whether they are configurable.
- 03Use compatible models and versions for indexing and querying.
- 04Design retrieval, reranking, and answer generation separately.
Embedding models do not directly generate answers. A formal integration contract for this scenario is not yet verified.
Example request: “Generate embeddings for a knowledge base to support semantic search”Find models for this scenario →
Related integrations
Browse the catalog →No complete integration is available for this scenario yet. Explore the provider and model directories and check official API documentation.