Demosourced from the handbookPOST /api/ask

Ask the handbook.

A live RAG chat over a fictional company handbook we call Northlight — 9 chapters on employment, leave, conduct, and GDPR. Answers stream in real time from a locally-hosted LLM; citations link back to the source chapter.

Try

The handbook is indexed in pgvector.

How it works
  1. 01Retrieve — pgvector cosine search over 768-dim embeddings.
  2. 02Augment — top passages are numbered and pinned to the prompt.
  3. 03Generate — the LLM answers in your language, citing passages by number.
Live index
—
chunks in pgvector
—
public documents
Source

Every answer is grounded in the fictional Northlight handbook — nine chapters covering employment, leave, conduct, and GDPR.

Open the full handbook →