Embeddings Overview

Embeddings power the semantic (vector) search component of Lucerna’s hybrid search. When embeddingFunction is not set, the indexer auto-selects:

1. CloudflareEmbeddings — if CLOUDFLARE_ACCOUNT_ID and CLOUDFLARE_API_TOKEN are set in the environment
2. NomicCodeEmbeddings — otherwise, runs a local code-optimised model via ONNX (no API key required)

Options

Option	Description
Local models	NomicCodeEmbeddings, BGESmallEmbeddings, HFEmbeddings — run locally via ONNX
Cloudflare	CloudflareEmbeddings — remote, uses Workers AI
Custom	Implement EmbeddingFunction to use any provider
false	Disable semantic search entirely — BM25 only, no model download

Disable semantic search

const indexer = new CodeIndexer({
  projectRoot: '.',
  embeddingFunction: false,
});