Shelves for AI visual
Overview of shelves for Cloudera Data Visualization AI visual.
The AI visual supports the following shelves:
- Embeddings
-
The Embeddings shelf is designed for the vector database comparisons. The user input is vectorized and compared to the defined embeddings to retrieve rows containing the most similar embeddings. The first field that contains results is returned.
Add a Dimension or Measures vector field that contains embeddings.
This is a mandatory shelf that accepts multiple fields.
- Embedding Context
-
The Embedding Context shelf allows you to add fields that you want to send as part of the chat prompt in a formatted manner, incorporating all embedding context data.
This is a mandatory and sensitive shelf that accepts multiple fields. For information on how to mark a dataset field as 'Sensitive' to restrict its usage in specific visual shelves, see Editing dataset fields.
- Context Aggregates
-
The Context Aggregates shelf allows you to include aggregated fields that summarize data for similar records in the AI visual. You can use this shelf to enhance data insights by providing more accurate responses to questions involving quantities, distributions, or comparisons.
Aggregation is performed on rows that share the same Dimension, Tooltip, and Vector fields. By including Context Aggregates, you ensure that numeric values (such as counts, sums, or averages) are calculated across similar records, rather than returning individual row-level results that may not provide meaningful insights.
- Tooltip
-
The Tooltip shelf facilitates adding a field for source information to be shown along the displayed results.
- Limit
-
This setting defines the maximum number of data rows that are processed by the visual.