Upstage Embeddings icon

Upstage Embeddings

Generate embeddings using Upstage Solar embedding models. 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

AI-generated

Summary

Generates vector embeddings for text using Upstage Solar models.

Inputs

  • Model (required) — Selects the Solar embedding model: 'embedding-query' for search queries/questions or 'embedding-passage' for documents/passages.
  • Input Type (required) — Determines how to process input: 'Single Text' for a direct string or a field from input data, or 'Array of Texts' for multiple strings.
  • Text — The specific text string to embed. Used when Input Type is 'Single Text' and Text Field is empty.
  • Texts — Multiple texts to embed (one per line). Used when Input Type is 'Array of Texts'. Max 100 texts (204,800 tokens), each max 4,000 tokens.
  • Text Field — Field name from incoming item data containing the text to embed. Used when Input Type is 'Single Text'. Defaults to the 'Text' parameter if empty.

Output shape

For single input: a single record with text, embedding array, model, usage, and dimension. For array input: a list of items, each containing a record per text with text, embedding, index, model, and usage.

Embeddings are returned as arrays of floating-point numbers. The node continues on error if configured.

Examples

Example 1: Embed a single search query for semantic search.

Model: embedding-query, Input Type: Single Text, Text: 'What is quantum computing?'

Example 2: Embed a list of documents for a retrieval system.

Model: embedding-passage, Input Type: Array of Texts, Texts: [doc1, doc2, doc3]

Example 3: Embed text found in an incoming item's 'content' field.

Model: embedding-query, Input Type: Single Text, Text Field: content

Discussion