Settings Input / Output Advanced History

Column handling

*required fields
Retrieval-Augmented LLM {{retrievalColumnAlert.label}} is using {{retrievalColumnAlert.retrievalColumn}} as retrieval column, but it isn't available in the metadata. Embedding column will be used instead.

Embedding column: Contains the text to embed and save in the vector store.

Metadata columns: Columns containing extra information, such as dates, sources, or broader context. For augmented models later defined in the Knowledge Bank, metadata can be used to enrich the LLM generated response or for retrieval in place of the embedding column.

Splitting settings

Splitting can be done by character count in this recipe. For more options and an interactive preview, use the "Split into Chunks" processor in an upstream prepare recipe. Read the documentation to learn more

Update settings

The Knowledge Bank update strategy to use for this run and subsequent runs of the recipe.
Used to determine if a document already exists in the Knowledge Bank.
Some parameters have changed, which require the knowledge bank to be rebuilt.
Knowledge bank will be cleared on the next run of this recipe

Records settings

Leave empty for unlimited
Leave empty for default value of 10.000 records when loading a dataset