Skip to main content

How Widget Evidence Reaches the Dashboard

Host page
→ bootstrap authentication
→ browser widget runtime
→ external assistant request
→ consent-aware message logging
→ conversation record and index
→ Captured Questions
→ Friction Insights

Runtime and trust boundaries​

The host supplies window.ChatbotConfig; the bootstrap exchanges its embed values for authenticated settings before loading the runtime. The browser runtime renders the widget, makes assistant requests, validates bounded host context, observes contextual launchers, and tracks SPA path revisions.

Assistant responses come from the external assistant service. Logging is a separate, consent- and plan-aware path. Eligible persisted messages can carry a context snapshot and navigation revision. Conversation detail derives a bounded context journey; Captured Questions exposes the original evidence.

Insights applies server-enforced date, assistant, surface, host, path, campaign, and widget-version scope before aggregating heuristic themes, representative questions, context rollups, and structured failure reasons.

What does not automatically happen​

  • Runtime context is not sent to the assistant request in the current release.
  • Not every telemetry event becomes conversation or Insights evidence.
  • A feedback reason does not establish the root cause.
  • A period comparison does not prove an intervention worked.
  • Sol Helps does not automatically detect every documentation conflict or stale statement.

For installation, see Embed and Deployment. For data interpretation, see Feedback and Evidence.