embedding

Local SQLite-backed document store with keyword similarity search. Built-in kdeps component.

Supports index, search, upsert, and delete operations. Useful for RAG pipelines without an external vector database.

Usage

# Index a document
run:
  component:
    name: embedding
    with:
      operation: index
      text: "The quick brown fox jumps over the lazy dog"
      collection: docs

# Search
run:
  component:
    name: embedding
    with:
      operation: search
      text: "fox"
      collection: docs
      limit: 5

Inputs

Name	Required	Default	Description
operation	yes	—	index, search, upsert, or delete
text	no	—	Text to index or search for
collection	no	default	Namespace for documents
dbPath	no	/tmp/kdeps-embedding.db	SQLite database path
limit	no	10	Max results for search