Skip to main content

Documentation Index

Fetch the complete documentation index at: https://docs.summand.com/llms.txt

Use this file to discover all available pages before exploring further.

A dataset is the unit of analysis in Summand. Every CSV upload or connector table becomes a dataset, and every other feature — Ask Summand, Surprise Finder, reports, the semantic layer — hangs off of it.

​
Creating a dataset

  • From a CSV. Upload a file directly; Summand infers the schema and target column.
  • From a connector. Pick a table from a Snowflake or Delta Sharing source. See Connectors.
When a dataset is created, Summand kicks off a semantic layer run, followed by Surprise Finder.

​
Visibility and sharing

Each dataset has one of three visibility modes:
ModeWho can access
PrivateOnly the owner
SharedOwner plus specific users invited by email
PublicAnyone with the link (data is redacted by default in the explorer view)
Sharing is per-dataset. Connectors and semantic layers inherit the dataset’s permissions.

​
Exploring a dataset

The dataset view has several tabs:
  • Overview — dataset metadata, target column, feature importance.
  • Data Explorer — paginated raw rows for spot-checking.
  • Ask Summand — natural-language Q&A. See Ask Summand.
  • Surprise Finder — ML-generated findings. See Surprise Finder.
  • Report — block editor and PDF export. See Reports.

​
Lifecycle

Datasets can be renamed, re-shared, or deleted at any time. Deletion is cascading — semantic layer artifacts, surprise findings, report templates, and chat history all clean up alongside the dataset record.