Getting started

Three pages, in order:

1. Installation — Docker Compose for a real deployment, or a local dev setup with uv and npm.
2. Quickstart walkthrough — your first document, from paste to reviewed result.
3. Core concepts — spans, detectors, policy, validation: the five ideas everything else builds on.

What you need to decide first

The app runs out of the box with no external services at all — rule-based detection on pasted text and native PDFs. That is enough to see how it works, but it is not enough for real use: the rule detector alone will miss names.

Two decisions turn it into something usable:

Decision	Where	Why it matters
Which LLM endpoint detects PII?	OPENAI_API_BASE, LLM_MODEL	The prompted LLM is the primary detector for German clinical text. Without it, only structured identifiers (dates, phone numbers, labelled IDs) are found.
What happens to scanned PDFs?	OCR_ENGINE	With none (the default) scanned PDFs are rejected with a clear message rather than silently returning an empty result.

Both point at base URLs you control. See LLM endpoints and OCR engines.

Document content flows to whatever you configure

Every detection request sends document text to OPENAI_API_BASE, and every OCR request sends page images to the OCR endpoint. Point them at services inside your own network. The app shows a banner in the header when a configured endpoint is not local — treat that banner as a blocker, not a hint.