Unstract - The platform purpose-built for LLM-powered unstructured data extraction

Unstract The platform purpose-built for LLM-powered unstructured data extraction

What is Unstract?

Unstract is an open-source, no-code platform designed to automate document processing workflows. Leveraging the power of Large Language Models (LLMs), it efficiently handles a wide variety of document formats without requiring manual annotations. It surpasses the current capabilities of IDP (Intelligent Document Processing) and RPA (Robotic Process Automation).

The platform focuses on providing structured data from unstructured sources and provides flexibility in the choice of LLM, Vector DB, Embedding Model, and Text Extraction service. Its LLMChallenge feature uses two separate LLMs to ensure accuracy by returning either the correct value or no value at all, eliminating hallucinations.

Features

LLMChallenge: Uses two separate LLMs to extract and challenge data, ensuring accuracy and eliminating hallucinations.
SinglePass Extraction: Constructs a single, large prompt from multiple field extraction prompts to reduce token usage.
Summarized Extraction: Creates a compact version of the input document to optimize processing efficiency.
Multi-LLM Support: Enables users to view and compare responses and costs from multiple LLMs.
Prompt Studio: A prompt engineering environment to build generic prompts quickly, with versioning and cost comparison features.
Layout-Preserving Mode: Preserves document layout, including multi-column layouts, forms, and tables.
Handwritten Text Detection: Offers state-of-the-art detection for processing challenging documents with ease.
Checkbox and Radio Button Detection: Facilitates easy processing of forms.