Manage datasets
On Desktop, the local owner manages datasets. On Web, administrators also control which users may access them.
Add and load data
- Open Admin > Data and add a prepared dataset or choose one from the catalogue. Follow Install datasets.
- Wait while COHORTome validates the package and builds its query store.
- Check the patient and assay totals.
- Open Datasets, select the datasets you want and apply the workspace.
A query store is reused. Rebuild query store is a maintenance action for source or mapping changes and failed builds. It is not an installation step or something users run each time they open the app. Bundled Desktop demos build their query stores automatically on first use.
For your own inputs, see the package format and server registration reference.
Grant access
Web only: in Admin > Data, choose dataset visibility and allowed groups or users. Prefer groups for shared team access. Users with access previews the resulting grants. Manage the accounts themselves in Admin > Users.
Visibility determines who can use the dataset in COHORTome. It does not decide whether data may be sent to an LLM.
Set data and AI permissions
| Setting | What it controls |
|---|---|
| Data classification | Whether the dataset is private or public |
| In-app Query detail | Whether Query can return detailed records or summaries only |
| AI providers | Whether models receive detailed context, safe summaries or nothing |
New registrations default to private. Keep restricted data private and AI access blocked until the permitted use has been reviewed.
Private data never goes to an external provider. A local provider still needs the dataset's AI permission. When several datasets are selected, the strictest policy applies. Safe summary means structural, non-clinical context—not aggregate clinical findings.
These AI controls also apply on Desktop. See Connect an LLM.
Change the workspace
Checking a box changes the pending selection. The active workspace changes only after Apply workspace succeeds. If a selected store is unavailable, the previous workspace stays active.
To unload it, clear the selected datasets and choose Unload workspace. This does not delete the datasets or their source files.