-
Notifications
You must be signed in to change notification settings - Fork 0
Gen3 Datasets
- Open Datasets from the tenant sidebar.
- Create a dataset for each logical corpus (project, unit, assessment cycle).
- Upload or import documents (Uploading and Importing Documents).
- Attach datasets in chat or agent defaults when you are ready to query them.
Datasets are the durable knowledge layer for chat retrieval, workflow review, and GT API upload automation.
Datasets are the active Gen 3 retrieval boundary for tenant documents. The current Datasets page owns dataset creation, document upload/import, ZIP import/export, document review, sharing posture, and deletion. Gen 3 no longer relies on a separate tenant Documents route for normal document operations.
- create new datasets
- edit dataset metadata and retrieval defaults
- upload files directly into a dataset
- import datasets from ZIP bundles or legacy document bundles
- export one or more datasets
- open the per-dataset document view
- share datasets privately, to groups, or more broadly when allowed by the deployment
The page includes GT2-style search and filter controls so you can narrow the list by ownership, category, tags, access posture, and sort order before opening a specific dataset.
Datasets supports bulk export, bulk delete, and bulk upload entry points so you do not have to open one dataset at a time for every operation.
Opening a dataset's document modal lets you inspect the current document list, upload more files, import additional documents, and remove outdated content.
- Open
Datasets. - Select Create Dataset.
- Enter the dataset name and optional description.
- Choose the access posture and any required group shares.
- Save the dataset.
Use direct upload when you already know the destination dataset. Use import when you are bringing in a packaged dataset archive or a larger prepared file set.
Open the dataset's documents view to confirm the files present, the ingestion results, and whether outdated material should be removed or replaced.
Datasets can be used in two places:
Use agent defaults for repeatable workflows and conversation-level attachments for one-off context.
Dataset visibility depends on the access option selected for the dataset and, when group-shared, on the relevant group. If another user cannot see a dataset, check its sharing posture before assuming the upload failed.
- Use clear dataset names that match a team, project, or content purpose.
- Remove outdated documents instead of leaving conflicting versions in the same retrieval set.
- Prefer group sharing over broad visibility when the content is team-specific.
- Revisit retrieval defaults after import so the dataset uses the intended embedding and chunking posture.