Every AI feature in a travel product falls apart on the same question: "which one?" A chatbot that cannot see the stay you already opened forces you to re-describe it, and re-describing a listing is slower than reading it yourself.
GGNomad solves that at the shell level rather than the feature level. A floating assistant is mounted once in the app shell, so it is present on Hospitality, Tourism, Entertainment, Bookings, Trips, Provider, and Analytics alike — you never navigate to an "AI page" to use it. Every prompt carries a page context resolved from the route itself, so opening a property and asking "what's the cancellation policy on this one?" is unambiguous: the assistant is told the entity type and id of the thing on your screen.
Because the context is derived from the route rather than typed by the user, it stays correct as you move around. Open a tour, then an activity, then a trip plan, and the same conversation follows you — each turn scoped to whatever is currently in front of you, including the active tab and the search query you typed.
Ask by typing or by voice: the widget takes browser speech input, so you can talk to it while planning on a phone. Responses stream in as they are produced rather than appearing all at once, and anything that looks like a token, JWT, or API key is redacted out of the stream before it reaches the screen. If the underlying runtime goes stale mid-conversation, the transport quietly rebuilds it and retries the turn once; closing the widget cancels the in-flight request instead of leaving it running.
Every action the assistant takes runs through the same governed GraphQL surface a human uses, so it can never see or change anything your own role does not already allow. What is still being built is depth rather than presence: a GGNomad-specific domain guide so the assistant knows travel vocabulary rather than inferring it from the schema, rich property and tour cards in place of plain text, and per-screen personas.
The assistant, its page-context capture, voice input, and redaction are wired today; the richer rendering and domain grounding are on the roadmap.
Ready to make this your story?
