Hyper-Extract

Why we included this project

Most extraction tools hand you back plain text and call it a day. Hyper-Extract instead returns typed structures your code can actually use: simple lists and pydantic models when that is enough, knowledge graphs, hypergraphs, and spatio-temporal graphs when you need richer relationships. You point the CLI at a PDF or a folder of notes, pick one of the over eighty YAML templates covering finance, legal, medical, and general domains, and get a searchable knowledge base you can visualize or export as an Obsidian vault. Provider support includes OpenAI, Anthropic, DeepSeek, Alibaba's Bailian, and local vLLM servers, so teams with data-privacy constraints can run extraction on their own hardware. For anyone assembling a RAG pipeline, that removes a lot of the glue code that normally sits between unstructured sources and a system that can reason over them.