Run Extraction

Once you have a schema, run extraction on documents to produce structured results.

Steps

1. Navigate to Extract in the sidebar
2. Click New Run in the top-right corner
3. Select a schema from the dropdown
4. Select documents (see options below)
5. Click Start Extraction

You'll be redirected to the run detail page where results appear as processing completes.

Selecting documents

Option A: Individual documents

1. Click the documents dropdown
2. Check boxes next to documents to include
3. Select as many as needed

Option B: Select a folder

1. Click the folder dropdown
2. Select a folder (shows document count)
3. Confirm in the dialog

Note: You can only use one option. To switch, clear your current selection.

Folder limits: Folders must have 1-50 documents. For larger collections, split across multiple runs.

Shortcut from Files

1. Go to Files
2. Select documents using checkboxes
3. Click Extract in the toolbar
4. You'll be taken to New Extraction with documents pre-selected

Monitoring progress

The run detail page shows:

The page auto-refreshes every 5 seconds while processing.

Canceling a run

1. Click the actions menu (⋮)
2. Select Cancel

The run stops after the current document finishes. Partial results are preserved.

Viewing results

When complete, you'll see a table with:

• Document — Source document name
• Schema fields — One column per field
• Source Quote — The text passage used for extraction
• Chunk — Which section it came from (hidden by default)

Column visibility

Click Columns to show/hide columns.

Understanding values

Exporting results

1. Click Export CSV (appears when run completes)
2. Download opens automatically

The CSV includes all fields plus provenance columns.

Opening exports

Compatible with Excel, Google Sheets, Numbers, or any data tool.

Processing time

Extraction processes roughly 10 pages per minute.

Large runs can be left running — check back later.

Related

• Create an Extraction Schema
• Review Extraction Results
• Extraction Reference