Skip to main content
Ragrails is six stages. The first four build your index; the last two query it.
StagePurposePage
IngestionPull content from URLs, files, and APIsIngestion
ChunkingSplit content into searchable passagesChunking
EmbeddingTurn passages into vectorsEmbedding
StoringSave vectors to a databaseStoring
RetrievalFind the chunks relevant to a queryRetrieval
ChatAnswer questions grounded in your dataChat
New to RAG? Read How RAG works first.

​
Run stages individually or together

  • Individually - call each stage when you need to inspect or control intermediate output.
  • Together - use ingest() and query() to run the whole pipeline in one call.

​
Keeping your index current

Stored data isn’t frozen. As sources change, update or remove chunks by ID with edit() and delete() so answers stay accurate. See Keep your knowledge base current.