Organize with labels
Group documents and notes into flat, overlapping collections, and use a label as the dataset for a chat or workflow.
Labels are how you keep a growing library navigable. Pick a few documents that belong together (a folder of contracts, a project's reading list, a quarter's research papers) apply a label, and you can act on the group as one.
What a label is (and isn't)
A label is a flat, overlapping grouping. A document or note can belong to many labels at once, and labels don't nest. Think of them more like tags than folders.
| Labels are… | Labels are not… |
|---|---|
| Flat (no parent/child) | Folders |
| Overlapping (one source, many labels) | Mutually exclusive |
| The persistent organizing primitive | The same as a dataset |
Labels are durable: they live in the workspace and persist across sessions. A dataset is the live composition a chat or workflow runs against; you'll often build a dataset by picking one or more labels.
Apply a label
From any row on the documents or notes page, click the label icon to open the label picker. Pick existing labels or create a new one inline. Apply lands optimistically: the chip shows up immediately while the change persists in the background.
To remove a label, click the small × on the label chip in the row, or use the picker again.
Bulk-apply across many items
Multi-select rows on the documents or notes page, then use the bulk-action menu to apply (or remove) a label across the whole selection. This is how a one-off pile of uploads becomes a named, reusable group in one move.
Use a label as a dataset
The point of labels isn't tagging; it's using the group. From the labels page you can:
- Open a Focus on the label. The dataset expands at retrieval time to every source the label currently holds.
- Spawn a chatbot bound to the label. Visitors ask questions across the label's documents.
- Run a workflow (Summarizer, Data Extractor, Blueprint, etc.) over every source in the label.
Adding or removing a member document updates the live group; the next Focus or workflow run picks up the change.
Limits
Each plan has a maximum number of labels per workspace. The cap is shown in your plan settings; if you hit it, you can delete or combine labels from the labels page.
What's next
- Combine sources into a dataset: what happens when you take labels into a chat.
- Pick documents with natural language: building a selection from a description, then labelling it.
- Embed a chatbot: turn a label into a public Q&A widget.