Use the dataset panel to see and act on a dataset

Open, resize and close the dataset panel

1. In the Data Library, click a dataset row. The panel opens on the right, next to the list.
2. To change its width, drag the panel's left edge, or focus the edge and use the arrow keys. Double-click the edge to reset it to half the page.
3. To close it, click the × in the panel header or press Esc.

On a phone the panel replaces the list, and a back arrow takes you back to your datasets.

Read the panel header

The header shows the dataset's icon, its name, and a line with when it was last updated. If the dataset has more than one View, a View switcher sits on the title line so you can move between its Views (All 4 views lists them).

The header's main button changes with the dataset's state:

Rename a dataset

1. Click the dataset name in the header. It turns into a text field.
2. Type the new name.
3. Press Enter, or click outside the field, to save. Press Esc to cancel.

You can also choose Rename from the panel's icon strip or the row's ⋯ menu. With view-only access to the project, the header shows View only and the name can't be edited.

Understand the notices at the top of the panel

Some datasets show a card above the Overview. Each one says what happened and offers the fix.

Views that need attention

When a View on the dataset is failing, out of date, or not syncing, a banner explains it:

On a dataset with several Views, the banner reads, for example, 2 of 4 views need attention, with counts such as 1 failing and 1 sync off. When more than one View needs you, Show me lists them. In the list, the same states show as chips on each View: Failing, Out of date, Sync off or Updating.

Files that did not fit the table

When files you add can't go in as they are, a card says so, for example "2 files arrived and did not fit the table." It also tells you how many rows the dataset still holds from before. The card's button names the fix:

Added the same way as last time

A new workbook can go in automatically using the saved tab and reading (see Choose how new files are added). A card then reports Added the same way as last time, with the number of rows and the tab they came from.

• See how opens Settings & history.
• Undo asks, for example, "Take out 120 rows?". Click Take them out to remove that workbook's rows from the dataset. This can't be reversed.
• Undo is not offered when the dataset replaces all data.

Share it through datasets

This card can appear if you manage the workspace's AI Agent settings and people outside the project ask about its folders. It lists each shared folder with Has it or Not yet.

• Add it for 3 folders hands the job to the AI Agent, which makes a filtered copy of the dataset for each folder that lacks it.
• Not now hides the card.

The AI Agent is available on Enterprise plans.

See what a dataset holds in the Overview

The Overview section opens first. It shows the dataset's shape (for example, 1,200 rows · 8 columns) and these cards:

• Summary (or AI summary): what the dataset is, what its pipeline tasks did, and where its data goes, such as dashboards or exports that it feeds.
• In the workflow: a small map of where this dataset sits, with the sources that feed it and the Views, exports and dashboards it feeds.
• Preview: the dataset's rows. Switch between Rows and Columns to see either the data or its column list. If the preview can't load, it says "That did not load. The dataset is not necessarily empty."

If a View returns unexpected results, check the Preview first. If the source data looks wrong here, the issue is upstream of the pipeline.

Check the recipe for data read out of a PDF or document

If a recipe read the dataset's rows out of a PDF or another document, the preview says the rows were "read out of" that file "by a recipe", and offers View recipe. The Recipe dialog (with the source file name underneath) shows:

When one file produces several datasets, the dialog adds, for example: "These 3 datasets come from one file, so they always update together." If a new file changes in a way Mammoth can't match, all of them keep their last good data until you look.

Use the doors: Explore, Export and publish, Add data, Automate

Below the Overview, a row of large buttons (doors) holds the main things you can do. Each one says what it does underneath.

Pull the latest data with Retrieve or Get the latest

When you click Retrieve or Get the latest (or Refresh on a webhook dataset), the door shows the progress in place: Getting the latest…, then Putting the new rows in… with a count, and finally Up to date. If it takes a while, it says "This is taking longer than usual. You can leave", and a message tells you when it finishes.

The finish message gives the outcome in rows:

If the pull fails, the door reads Couldn't get the latest with the reason and "Your data is as it was.", plus a Try again button. If refreshing is switched off for the dataset, you see "Refreshing is switched off for this dataset, so there is nothing to pull."

Find other actions in the icon strip and More options

Next to the main buttons is an icon strip with the dataset's other actions, and a More options (⋮) menu with whatever doesn't fit. Hover an icon to see its name. Depending on the dataset, these include:

The same actions are available from the ⋯ menu on the dataset's row in the list, and by right-clicking the row.

Download a dataset as CSV or Excel

Download opens a short dialog before the file is made. Choose a format:

• On a dataset with several Views, the dialog says which View it downloads: the one you were last in.
• When some of the selected datasets have the same columns, Stack the ones with the same columns into one file appears, with how many files you'll get.
• To download several at once, tick them in the list and click Download in the selection bar. A folder's Download folder gets every dataset inside it.
• Raw files that have not become datasets are left out ("Raw files have no rows to download yet").

While the file is prepared, the dialog shows, for example, "3 of 5 ready. The file saves on its own when it's done." If you leave the page, a notification brings you the file when it's ready.

For other ways to get data out, see Download data.

Move a dataset to Trash from the panel

Move to trash opens Move to trash?. The dataset leaves the library at once, and you can restore it from Trash for 30 days, which the dialog states. Its Views go with it. See Get things back from Trash.

Some items can't go to Trash. A raw file that never became a dataset shows Delete permanently instead, and its dialog is titled Delete for good?. If you select several items and some can't go to Trash, the selection bar warns, for example, "1 item cannot go to Trash and would be deleted for good."

Resolve a file that needs your input

A file that is waiting on you (a password, a tab to pick, or a reading to choose) opens a smaller panel: "This file needs your input before it can be processed." Its actions are:

• Resolve: opens the fix;
• Explore: look at the raw rows;
• Replace the file and Rename: once it is already a dataset;
• Move to;
• Move to trash or Delete permanently.

See Upload files.

Act on a View from the panel

Views appear under their dataset in the list. Click a View row and the panel shows that View, labelled, for example, View of Daily Sales. A View has a shorter set of actions, because batches, update settings and files belong to the dataset:

A View can't be moved on its own, because it lives with its dataset. Move the dataset instead.

How datasets, batches and Views fit together

Every dataset in Mammoth follows this structure:

Dataset
  └── Batch(es)       ← individual data loads (each file upload = one batch)
  └── View(s)         ← transformation pipeline outputs built on the batches
       └── Pipeline tasks  ← filters, formulas, joins applied to produce the View

• Batches are the raw inputs: the individual data loads you can suspend, delete, or retire with a retention policy. See Manage batches.
• Views are the processed outputs, shaped by the pipeline tasks you've built. Each View has its own pipeline, and a dataset can have several.
• The Preview in the Overview shows the dataset's rows. When you open a View with Explore, you see the pipeline's output.

Limits

Troubleshooting

Related

• Dataset settings and history: update method, column rules, batches and history.
• Find your way around the Data Library: the list the panel opens from.
• Views: create, rename and delete: work with a dataset's Views.
• Copy, move and delete datasets: clone and move a dataset.
• Add or replace data: append or replace data in a dataset.
• Get things back from Trash: restore a dataset you deleted.