Overview

​

Default

• Choose this strategy by adding the query parameter “default”, i.e. ?strategy=default at the end of the URL (or, by simply leaving it blank).
• The fastest and cheapest strategy.
• Extracts text from files, and makes no attempt to extract anything more.
• This means it’s great for long contracts full of words, for example, but not a great option for documents which are full of tables (and where you need the information contained in the tables).
• Outputs text (i.e. you should expect to receive a string back).
• Supports PDF and Word documents.

​

Vision

• Choose this strategy by adding the query parameter “vision”, i.e. ?strategy=vision at the end of the URL.
• Can read everything from files, i.e. tables, images, watermarks, text, etc.
• Outputs text (i.e. you should expect to receive a string back).
• This is a great option for documents which contain tables, and documents where the structure of the information on the page is important for human-level understanding (for example: invoices, receipts, and bills-of-lading).
• Supports PDF, Word documents, and JPEG images.

​

SOTA

• Choose this strategy by adding the query parameter “sota”, i.e. ?strategy=sota at the end of the URL.
• Can read everything from files, i.e. tables, images, watermarks, text, etc.
• Outputs text, HTML, Markdown, JSON, and Chunks. All formats are computed at runtime, and you can choose which one you access later.
• This is a great option for documents which contain tables, and documents where the structure of the information on the page is important for human-level understanding (for example: invoices, receipts, and bills-of-lading).
• Supports PDF, Word documents, and JPEG images.