Skip to main content
Discovery results table showing extracted data from multiple documents This guide walks you through creating a discovery from start to finish, including how to work with, verify, and export your results. To create a Discovery:
  1. Create a new discovery or start from a template
  2. Add columns to define what to extract
  3. Add documents and run the discovery

Creating a Discovery

1

Start a new discovery

Navigate to the Discovery section and click + New Discovery in the top right corner.Discovery section with New Discovery button highlighted
2

Name your discovery

Enter a descriptive name for your discovery. You’ll see an empty table ready for columns and documents.New discovery view with empty table and Add Column button
Libra mascot with clipboard
You can also start a Discovery from a pre-built template. Templates come with columns already configured for common document types.
https://mintcdn.com/libra-4206ec93/iRY_yL_HtBG1LhMP/images/icons/discovery.svg?fit=max&auto=format&n=iRY_yL_HtBG1LhMP&q=85&s=7bb098b5839452d200355b66fc3b77b3

Work with a Template

Learn how to use and create reusable templates with pre-configured columns.

Adding Columns

Columns define what information Libra extracts from each document. Each column becomes a question that Libra answers for every document.
1

Click Add Column

Click + Add Column in the table header or toolbar to open the column configuration dialog.Add New Column dialog with Column Label, Prompt, and Answer Type fields
2

Configure the column

Fill in the column details:
  • Column Label: A short name that appears as the column header (e.g., “Document Type”)
  • Prompt: Instructions telling Libra what to look for (e.g., “Determine and identify the category of this document, specifying the type of contract or any other relevant classification”)
  • Answer Type: The format of the extracted data
Answer Type dropdown showing Text, Yes/No, Number, Date, Bullet List, and Currency optionsAvailable answer types:
  • Text: General information and descriptions
  • Yes/No: Presence or absence checks
  • Number: Quantities and amounts
  • Date: Dates and deadlines
  • Bullet List: Multiple items
  • Currency: Monetary values
Start with your most important columns. You can always add more later after seeing initial results.

Adding Documents

1

Open the document selector

Click Add Documents in the toolbar to open the document browser.Discovery view with Add Documents button highlighted
2

Browse or upload files

Select documents from your project or upload new files using Upload Files or Add Folder.Add Documents dialog showing file browser with Upload Files and Add Folder buttons
3

Select documents

Check the documents you want to include. Choose whether to add them as separate rows or combine them.Document selection with checkboxes and Add as rows options
4

Wait for processing

Libra processes your documents. You’ll see a progress indicator as files are uploaded and prepared.Document processing progress showing files being uploaded
5

Start the discovery

Once documents are added, click Start discovery to begin extraction. Libra will analyze each document and fill in your columns.Discovery table with documents added and Start discovery button ready
6

Continue working while processing

Discovery runs in the background, so you can continue working in Libra while extraction completes. You’ll see a progress indicator and be notified when results are ready.Discovery processing in the background with progress indicator

Working with Results

After a discovery completes, results display as a spreadsheet-like table:
  • Rows: One row per document analyzed
  • Columns: The data points you defined in your template
  • Cells: The extracted information for each document/column combination

Sorting and Searching

Sorting: Click any column header to sort the table by that column. Click again to reverse the sort order. This helps you find the highest or lowest values, group similar items together, and identify outliers. Discovery table sorted by a column Searching: Use the global search field to find specific content across all documents and columns. Type your search term and the table updates to show only matching rows. Discovery table with search results

Verifying Results

Discovery results table with sidebar showing source text and reasoning for an extraction Click on any cell to see the extracted value, Libra’s reasoning, and the exact source text from the document.
Always spot-check extracted data, especially for important fields. AI extraction is highly accurate but not perfect.
To verify an extraction:
  1. Click on the cell you want to check
  2. Review the Source text that Libra identified
  3. Click through to view the text in the original document
  4. Confirm the extraction is correct
Libra mascot with clipboard
Need to correct an extraction? Click on any cell, then select the pen icon in the sidebar to edit the value. Your changes are saved to the discovery but don’t affect the original documents.

Exporting Results

Export dialog showing format options for discovery results Export your discovery data for use in other applications. To export:
  1. Click Export in the results view
  2. Choose your format
  3. Download the file

Continuing Analysis in Chat

Discovery results are a great starting point for deeper analysis. Once you’ve identified interesting findings in your table, you can continue the conversation in Chat.
Chat with Discovery interface showing follow-up conversation about extracted data After running a Discovery, click Chat with Discovery to ask follow-up questions about specific documents or compare provisions you’ve extracted in more detail.

Tips for Effective Discoveries

Discovery works best when documents are similar in structure. Running discovery on very different document types may produce inconsistent results.
Begin with the most important columns. Once those work well, add more. It’s easier to debug a template with fewer columns.
For high-stakes decisions, always click through to the source text to confirm extractions are correct.
After identifying interesting findings in your table, use Chat with Discovery to ask deeper questions about specific documents.

Next Steps