Unstructured is an open-source Python library that converts PDFs, HTML, Word files and images into structured elements for LLM workflows. partition() detects the file type automatically, and the separate unstructured-ingest project handles batch ingestion. Docker images exist for x86_64 and Apple silicon.
Install with pip install "unstructured[all-docs]" and call partition(filename=...). Some formats need system tools such as tesseract or libreoffice.
Apache-2.0 licensed. Telemetry is on by default; set DO_NOT_TRACK to opt out. The company also sells a hosted platform for heavier workloads.