Skip to content

Advanced Topics

These topics build on the foundations from the earlier pages. Skim them once you’re comfortable deploying.

There are two audiences, and they see different knowledge:

  • Ragtime users — signed-in members of your organisation. In preview and in access-controlled contexts they can reach all active libraries linked to the project.
  • Public users — everyone reaching a published link without an org login. They can only reach libraries that are linked and marked public.

This is why an assistant can feel “fuller” to your team in preview than to the public: a library you linked but left non-public is invisible to public users.

Manage library links and their public flag under Dashboard → Access Control. Library links are frozen into the snapshot at publish time, so changes take effect on the next publish.

Devices of type kiosk with latitude/longitude and a location label appear on the Dashboard → Deployment → Map view — a geographic picture of where your physical deployments are.

This is how you manage a fleet of on-site assistants (lobbies, stores, event booths) alongside your web embeds, all from one project.

Pair kiosks with:

  • IP allow-lists for on-network-only access.
  • Closed captioning and avatar settings tuned for public displays.
  • Location-aware RAG if the assistant’s answers should vary by physical location (see Location-aware RAG).

Data out: webhooks, transcripts & analytics

Section titled “Data out: webhooks, transcripts & analytics”

Webhooks deliver events to your endpoints: message, session_started, session_ended, action_executed, knowledge_used, source_ingested. Every payload carries an isLive flag so you can tell preview/test traffic from real published traffic.

Transcript events forward conversation text to the embedding page in real time (opt-in under Additional Permissions). Ragtime does not store transcripts — you own the privacy notice.

Analytics breaks usage down per device and separates public vs. logged-in usage. This is the payoff for deploying each surface as its own device: you can see exactly how each touchpoint performs.