Skip to content
DRAFTEnglish end user manual · web beta f7b7740e7371 · Android beta ed5d9494be9eDownload PDF

File Areas

A file area is a workspace for related uploads. Use it for recurring CSV or delimited text, XLS/XLSX spreadsheet, or JSON files that belong together. Parquet is not decoded by the current File Areas parser.

When you upload a file, Rokks samples the data and proposes a transform plan. The plan describes display names, data types, mappings, units, and constraints. In a regular AREA, each file can load to its own physical table. Use a Virtual File when several files must combine into one table, or a Constellation when the source represents several related tables. After a successful load, the resulting table or tables become available in the catalog and can be selected in the Widget Builder.

A Constellation is a leaf container inside a regular File Area. It accepts one JSON source, owns one multi-table plan, and displays each governed result under Output tables with its own Inspect action. The same tables are available through Analytic Warehouse → Data Source Tables and Relationship Map. They are File Areas outputs, so Local Tables does not list them.

BETA PRODUCTVerified product capture · build f7b7740e
ACME File Areas with one loaded airline sample and another awaiting a plan
ACME File Areas shows the airline guide loaded with 144 rows. The other airline file and Pagila constellation still need a plan; this is not a completed Pagila import.Captured 12 Sept 2026 UTCOpen documented view ↗

The first capture contrasts a loaded airline sample with files still awaiting a plan. The second shows the real six-table Pagila rental projection after loading: all output tables are ready, with 51,805 rental rows and 51,056 rental-level payment aggregates.

BETA PRODUCTVerified product capture · build f7b7740e
ACME Pagila constellation with all six output tables loaded
Real beta File Areas output: six loaded tables, including 51,805 rentals and 51,056 rental-level payment aggregates.Captured 12 Sept 2026 UTCOpen documented view ↗
  1. Open Data Fabric.
  2. Create a file area with a clear business name.
  3. Upload one or more related files. For one-table output from several files, create a Virtual File instead of relying on ordinary area grouping.
  4. Review each proposed transform plan, or configure the Constellation when the source contains related tables.
  5. Adjust types, labels, units, and mappings.
  6. Run the load and monitor the job.
  7. Build dashboards from the resulting catalog entries.

Sidekick can perform the same attachment-to-plan sequence in Agent mode. Use the File Area and constellation cookbook when you want a chat-driven path with an explicit plan-review stop before loading.

Do not assume that putting several files in one regular area merges them. Ordinary files retain separate table outcomes; use a Virtual File for N-to-one output or a Constellation for a related multi-table model.

Do not rename fields only inside a dashboard. Fix display names in the source metadata so every widget and assistant workflow sees the same names.