Datasets
A dataset is like the AI's working notebook — a tidy, organized copy of your data that the chat puts together while it works, so it can chart it, search it, or build an app around it. You don't create datasets by hand; the chat makes them for you whenever a task involves real data.
Why it matters
When you upload a spreadsheet, paste a table, or ask the chat to pull together some information, that data needs to live somewhere organized — not just buried in the conversation. A dataset is that organized home. It lets the chat:
- Work with large amounts of data without cluttering the conversation
- Reuse the same data across several messages in the same chat
- Power an interactive app — a chart, dashboard, or table — built on top of it
How it works
| Step | What happens |
|---|---|
| You provide data | You upload a file, paste a table, or ask the chat to gather information |
| The chat organizes it | It builds a dataset — a structured collection it can read and re-read reliably |
| It stays with your chat | The dataset is tied to that conversation, so later messages can reuse it |
| Apps build on it | When the chat creates an app, the app reads from the dataset to show your data live |
Datasets work quietly in the background. You generally won't manage them directly — you'll just notice that the chat remembers and reuses your data within a conversation, and that the apps it builds stay in sync with it.
Example
You upload a list of 2,000 customer orders and ask for "a breakdown by region with a chart." The chat organizes the orders into a dataset, then builds an app that reads from it — so you can filter and explore all 2,000 rows without re-sending the file each time.
What's next
- Apps — the interactive views built on top of datasets
- Capabilities — everything else the chat can do