Unstract review: no-code document extraction tutorial

Unstract is an open source document automation platform that uses language models to extract structured data. This sponsored review demonstrates Unstract Cloud and explains that users can also install the software on their own servers. It does not demonstrate a self-hosted installation or establish offline operation.

The walkthrough creates a project, uploads documents, and selects extractor and agent models plus a Whisperer connector. The presenter extracts raw text from an air waybill, including handwritten entries, then generates a JSON summary. Agentic Prompt Studio produces an editable schema and an extraction prompt with field checks and output examples.

Although the demo uploads several document types, the presenter recommends building a schema around variations of the same document category. Clean scans, skewed pages and handwritten versions of one form are the suggested inputs for a reusable workflow.

The review also covers checking extracted fields against the original PDF and promoting reviewed results to verified data. The dashboard reports 100% project accuracy in this example; that result does not establish accuracy across other documents. Settings show GPT-4o and mention Gemini and Claude as connector options. The tutorial focuses on configuring extraction and reviewing its output, with no hardware requirements or local model setup provided.