Short answer: run RagLeap Core, upload your menu, rooms, opening hours, policies and FAQs, and use the built-in hospitality_agent role. Related roles exist for event_planner, travel_agent, salon_spa_agent and fitness_studio_agent. It is open-source (MIT) software you run yourself. The steps are in the general setup guide; this page covers what is specific to restaurants, hotels and similar businesses.
Which roles apply?
Five of the 46 built-in roles: hospitality_agent, event_planner, travel_agent, salon_spa_agent and fitness_studio_agent. You can also combine support, sales and secretary. Names are the identifiers in defaults.py at v0.11.0.
What would you upload?
The documents your staff answer from today: menus and allergen sheets if you have them, room descriptions, price lists, opening hours, cancellation rules, event packages. Upload PDF, Word, Excel and other supported files in the Documents tab, or give a public web page by address. The assistant cites the source document and knows only what you load.
How do guests reach it?
Through WhatsApp, Telegram, Discord or a voice line (Twilio). A website chat widget is planned, not built. The voice channel is a Twilio integration; what exists and what is research is described separately.
What about bookings?
The README connector list (MySQL, PostgreSQL, MongoDB, REST API, Salesforce, HubSpot, Shopify, Google Sheets, Stripe, CSV upload) contains no reservation system. Its known limitations say sync runs on demand only. Treat the assistant as an answering and enquiry-capturing tool, and test any booking workflow yourself before relying on it. The optional n8n webhook can fire after a reply on WhatsApp, Telegram or Discord, which lets you connect your own automation.
What should you check first?
- Autonomy is
offby default;semiasks before acting;fullis opt-in. - Allergen and safety answers must come from documents you have checked. The assistant is not a substitute for staff for those questions, and no accuracy claims are made.
- Your documents and each conversation go to the AI provider you configure.
- Single-tenant: one deployment, one business. No named users or roles yet, and no independent security review yet.
Sources
The v0.11.0 README, defaults.py and the channels documentation.