March 2026 Summaries

Unstructured's workflows enhance document processing by taking documents through a series of nodes, including a Partitioner, Enrichments, a Chunker, and an Embedder, ultimately preparing data for RAG, agentic AI, and model fine-tuning. A new addition, the Extract node, allows users to transform documents into structured, application-ready JSON records using a schema that follows the OpenAI Structured Outputs format. This node supports both LLM-based and regex-based extraction methods, catering to different data extraction needs, from understanding complex content to recognizing predictable patterns. The Extract node integrates seamlessly with existing workflows, providing intelligent document processing capabilities without additional infrastructure, and supports tasks like finance invoice processing, healthcare record extraction, and legal contract analysis. Additionally, the Extract node is easily incorporated into workflows via the Unstructured Python SDK and API, and users can test and refine their extraction processes using the interactive workflow builder before deploying at scale.