Upstage Embeddings Model icon

Upstage Embeddings Model

Embedding Model for Vector DB - Upstage Solar Embeddings. Supports up to 100 strings per request with max 204,800 total tokens. Each text should be under 4000 tokens (optimal: under 512 tokens).

Upstage Embeddings Model

AI-generated

Summary

Generate vector embeddings for input texts using the Upstage embedding model optimized either for search queries or for documents and passages.

Inputs

  • model — Select the Upstage embedding model to use, either 'embedding-query' for queries and questions or 'embedding-passage' for documents and passages.

Output shape

an array of embedding vectors corresponding to the input texts, emitted as output named 'Embeddings' with MIME type 'ai_embedding'.

Supports up to 100 input strings per request with a total token limit of 204,800; each input text should be under 4000 tokens (optimal under 512 tokens). Outputs the embeddings in the same order as inputs.

Examples

Example 1: Embed multiple short passages to vector form for semantic search.

Set 'Model' to 'Embedding Passage' to optimize embeddings for document content.

Discussion