Skip to main content
The parse node extracts raw text from documents using optical character recognition (OCR). It does not interpret the content. It converts document pages into plain text that downstream nodes can process. Use parse before extract when you need more control over the text extraction step, or when you want to filter or split a document based on its content.

​
When to use parse

  • You want to filter, split, or classify documents based on raw text before deciding whether to extract them.
  • You need the parsed text as a separate output for search, archiving, or audit.
  • You’re running the same document through several extract schemas and want to OCR once.
  • Skip parse when extract is the next step. Extract already runs OCR internally; chaining them runs OCR twice and bills both.
For deeper guidance on engines, precision, and cost, see parse best practices.

​
Configuration

​
Engines

  • Standard: Traditional OCR. Best for clean, typed documents. 1 credit per page.
  • Vision: AI-powered vision model. Handles handwriting, complex layouts, and low-quality scans. 2 credits per page.
  • KeyValuePair: Extracts text as key-value pairs using layout analysis. 1 credit per page + LLM processing cost.
  • Layout: Layout-aware OCR that preserves the document’s structure (tables, columns, and reading order) in the parsed text. Use it for structured documents where layout matters. See credits for its cost.

​
Inputs and outputs

Allowed inputs: upload trigger, email trigger, webhook trigger, forward trigger, filter action, split action, and classify action. Output: Raw text content extracted from the document.

Parse best practices

Pick the right engine, control cost, and avoid pitfalls

Extract action

Extract structured data using AI after parsing

Filter action

Keep or drop pages based on parsed text content

Credits

Per-engine credit costs