ParseSphereParseSphere

Parses

The Parses feature lets you extract text, tables, and metadata from documents without creating a workspace. Perfect for one-off document processing.

When to Use Parses

Parses

Quick document processing

  • Extract text from a document quickly
  • Get tables from a PDF
  • Process a document without storing it long-term

Workspaces

Ongoing data analysis

  • Chat with your data
  • Query multiple files together
  • Keep files for ongoing analysis

How to Parse a Document

1

Go to Parses

Click Parses in the navigation
2

Upload a file

Drag and drop a file onto the upload area, or click to browse
3

Configure options

Optional: Enable OCR, table extraction, or chunking
4

Click Upload & Extract

Start the document processing
5

Wait for processing

Processing time depends on file size and options selected

Parsing Options

OCR (Optical Character Recognition)

Enable OCR to extract text from:

  • Scanned PDFs
  • Images embedded in documents
  • Photos of documents

Information

OCR increases processing time but is essential for scanned documents.

Table Extraction

Enable Extract Tables to get structured table data from your documents. Tables are returned as JSON with headers and rows.

Text Chunking

Enable Chunking to split extracted text into smaller pieces. Useful for:

  • Building RAG (Retrieval Augmented Generation) systems
  • Storing in vector databases
  • Processing with AI tools

Configure:

  • Chunk Size — Characters per chunk (default: 600)
  • Overlap — Characters shared between chunks (default: 200)

Session Duration

Choose how long to keep results available:

30 minutes
OPTIONALduration
Default: true

Quick temporary extractions

1 hour
OPTIONALduration

Short-term analysis

2 hours
OPTIONALduration

Extended working session

6 hours
OPTIONALduration

Day-long project work

24 hours
OPTIONALduration

Maximum retention period

Warning

After the session expires, you'll need to re-upload to access results.

Viewing Results

After processing completes, expand a parse card to see:

  • Metadata — Processing time, page count, character count, token count, table count, and expiration date (if a session was created)
  • Extracted Text — A compact preview of the first 800 characters. Click Open Reader to view the full text in a dedicated reader with theme options (Light, Dark, Sepia), adjustable font size, line numbers, search with highlighting, and a reading time estimate
  • Tables — Each extracted table shown inline with row/column dimensions and a preview of the first 10 rows

Export Options

  • Copy All — Copy the full extracted text to clipboard
  • Download — Save the extracted text as a .txt file

Both options are available in the compact preview and the full reader modal.

Recent Parses

Your recent parses appear below the upload area as expandable cards. Each card shows:

  • File name with extension badge
  • Status badge (Queued, Processing, Completed, or Failed)
  • File size and relative time (e.g., "5m ago")
  • Expiration countdown (if a session was created)

Click a card to expand it and view the full results. The list auto-refreshes while any parse is still processing.

What's Next?

  • Workspaces — Create a workspace for ongoing analysis
  • API Keys — Get an API key to parse programmatically