Interface IndexConfig

Configuration for indexing documents for semantic search in the store.

This configures how documents are embedded and indexed for vector similarity search.

interface IndexConfig {
    dims: number;
    embeddings: Embeddings;
    fields?: string[];
}

Properties

dims: number

Number of dimensions in the embedding vectors.

Common embedding model dimensions:

embeddings: Embeddings

The embeddings model to use for generating vectors. This should be a LangChain Embeddings implementation.

fields?: string[]

Fields to extract text from for embedding generation.

Path syntax supports:

Simple field access: "field"
Nested fields: "metadata.title"
Array indexing:
- All elements: "chapters[*].content"
- Specific index: "authors[0].name"
- Last element: "array[-1]"

["$"] Embeds the entire document as one vector