Selects which user is asking. Their group membership rides every chat turn into the RAG retrieval layer, where knowledge-base chunks are filtered by acl_groups before the LLM sees them.
- [email protected] — member of
sg-stormrunner-engineering. Sees Project Stormrunner; replies are grounded with citations. - [email protected] — no group. Same queries return "No relevant information found."
Switches the prompt-suggestion deck so each team only sees the tools they're authorised for under 3AIgent's tenant access controls. Demonstrates how the same chat shell adapts per-tenant.
- 🔧 Engineering — EPNM · Splunk · RAG
- 📡 Operations — ThousandEyes · Accedian · EPNM · RAG
- 👔 Executive — full toolset (all of the above)
Switches the entire UI and chat replies into the chosen language at runtime — no static translation files. Each chunk of visible text is sent to the 3Point LLM (llm.3point.ai) and swapped in place.
- 55 languages — covers every major world language; pinned "most spoken" group (EN · ZH · HI) at the top.
- Dynamic chat replies — bot responses, citations, and prompt cards are translated as they stream in.
- Bisect-retry — large translation chunks auto-split on failure so the UI never gets stuck.
Toggles between the default dark presenter theme (best on stage and projector) and a light theme tuned for daylight rooms and screen-share. Choice is persisted in localStorage so it survives reload.
- Dark — Nord-inspired surfaces, cyan accent rails, cinema-quality contrast for live demos.
- Light — soft #fafafa surfaces, deepened cyan (#0891b2) for ≥4.5:1 contrast, slate borders.
