The Atlas doc.haus documentation, bound to its code
108 documents
demo/README.md

The wholly fictional demo matter: a letter of engagement between invented parties (Aldgate & Crane LLP / Aldgate Mills Limited), generated by services/ingest/src/seed.ts and loaded idempotently via ./start.sh --demo. Lists five things to try that exercise every core surface: cited Q&A on the liability cap (§ 9) and billing terms, tabular review columns, the /review multi-agent pipeline, and CourtListener-backed research on limitation-of-liability case law. First run, demos to others, or safe testing of new agents and tools.

Demo matter

A wholly fictional letter of engagement so you can try doc.haus in two minutes without any document of your own. Aldgate & Crane LLP and Aldgate Mills Limited do not exist; the figures, clauses, and people are invented.

Letter-of-Engagement-Aldgate-Mills.docx is generated by services/ingest/src/seed.ts — the same script that seeds the demo matter.

Load it

The simplest path is to launch the whole stack with the letter pre-seeded:

./start.sh --demo

The web app opens with "Aldgate Mills — Engagement (Demo)" already selectable. --demo is idempotent — it only ingests on the first run, so it is safe to leave on.

To seed without launching (it creates a matter and ingests the letter through the real pipeline — local MiniLM embeddings into a per-matter SQLite index):

cd services/ingest && bun run seed

Then open the web app and select "Aldgate Mills — Engagement (Demo)". Or, to do it by hand, create a new matter in the UI and upload the .docx in this folder.

Things to try

  • Cited Q&A (qa agent): "What is the cap on the firm's liability?" → answers with [Letter of Engagement — Aldgate Mills § 9] and the verbatim clause.
  • "When are bills payable and what is the late-payment interest?" → clause 5.
  • Tabular review (extract agent): add columns like "Liability cap", "Payment terms", "Governing law" and review the document as a grid.
  • Legal review (/review): runs the reviewer → assumption-challenger → summarizer pipeline over the letter.
  • Research (research agent): "How do courts treat limitation-of-liability clauses for consequential loss?" → pulls real U.S. opinions from CourtListener alongside the matter's own clause.

Linked from