Skip to content
Saiku Cloud

Files

Files is the second tab of Data sources, and it exists for the very common case where the data you want to analyse isn’t in a warehouse. Upload a file, and Saiku spins up a database from it and queries that. From the cube’s point of view there’s no difference — same authoring flow, same MDX, same API.

It’s the right surface for prototyping against an export, for one-off analysis of something that never made it into the warehouse, and for small reference tables that ride alongside your main data. For a production workload over hundreds of gigabytes you want a warehouse connection instead.

Uploading

Pick a workspace, choose a file, click Upload. One file per upload for now.

Supported formats:

  • Parquet (.parquet) — columnar, and by a distance the most efficient. Only the columns your query touches get read, so a multi-gigabyte Parquet file stays comfortable.
  • CSV (.csv) — with a header row.
  • JSON (.json) — newline-delimited records or a single array.
  • Excel (.xlsx, .xls).

Anything else — images, archives, PDFs — isn’t readable by the engine and will be rejected.

Building a cube from a file

  1. Upload the file.

  2. Click Create a cube on its row. You land in the cube author with the file already profiled — no re-upload, no picking it out of a dropdown.

  3. Describe what you want and let AI draft the cube, or build it by hand. It’s the same flow as authoring from a warehouse.

Each row also has a download button, so the file you uploaded is never trapped in the product.

Quotas

The page header shows storage used as both a percentage and a figure against your cap:

TierStorage
Starter5 GB
Team50 GB
Business500 GB
Enterprise5 TB (custom beyond that)

At 80% of your cap we start warning you; past 100% new uploads are rejected until you delete something or upgrade. Queries against files already uploaded keep working either way — hitting the cap never breaks a live cube.

There’s a per-file ceiling as well as a total — 100 MB for a single upload. If one file is bigger than that, split it or load it into a warehouse and connect that instead.

Deleting

Delete a file from its row. It’s purged from storage immediately, and any cube reading from it stops resolving until you re-upload or re-point it. Deleting a workspace deletes its files with it.

Uploaded files are encrypted at rest, the same as warehouse credentials — see Tenant isolation.

When to prefer a warehouse instead

  • The data moves. An upload is a snapshot. If the underlying numbers change daily or hourly, point at a warehouse so cubes read fresh.
  • You’re over the cap. 5 GB on Starter goes quickly. Upgrade, or move the data somewhere it belongs.
  • Several cubes need to join the same dataset. A warehouse table joins from many cubes cleanly; a file is read per cube.