Skip to main content

Documentation Index

Fetch the complete documentation index at: https://developers.autoplay.ai/llms.txt

Use this file to discover all available pages before exploring further.

Coming soon. This page will show how to extend the bridge you built in Step 1 so that Autoplay can proactively message the user — not only reactively answer their questions. You’ll add:
  • agent_state_v2.SessionState — a 3-state FSM (THINKING / PROACTIVE / REACTIVE) per session that gates when an unsolicited nudge is allowed to fire. The session you’re already creating in Step 1’s bridge stays the source of truth here.
  • ProactiveTriggerRegistry — a registry of triggers (built-in or custom) that watch the action stream and decide whether to surface a message. The SDK ships ready-made triggers like canonical_url_ping_pong (user bouncing between two pages = stuck).
  • A push channel back to the chat widget — Rasa exposes POST /conversations/{sender}/tracker/events, which lets the bridge inject a bot message into an active chat session.
  • (Optional) popup triggers — surface guidance overlays in the app itself when the FSM transitions to PROACTIVE, so the user sees a hint without having to open the chat first.

​
Sketch of the flow

ActionsPayload arrives in bridge
   ↓
ProactiveTriggerRegistry.evaluate_first(context)
   ↓ returns trigger_id if a trigger fires
SessionState.transition_to_proactive(trigger_id)
   ↓ returns False if cooldown blocks the offer
bridge → Rasa POST /conversations/{user_id}/tracker/events
   ↓ injects bot message
Widget shows the proactive nudge in the open chat
   (or: bridge → app frontend → popup overlay)

​
What you’ll need

  • A working Step 1 bridge (this page builds on top of copilot_server.py).
  • Familiarity with autoplay_sdk.proactive_triggers — see the proactive triggers reference and the authoring guide.
  • A handler in your web app for popup events if you want on-screen guidance in addition to chat nudges.

​
Planned sections

  1. Wire the trigger registry into on_actions — evaluate against ProactiveTriggerContext.from_actions_payloads.
  2. Gate proactive offers with SessionState — respect the FSM’s cooldown rules so users aren’t spammed.
  3. Push a bot message into Rasa — bridge → POST /conversations/{user_id}/tracker/events with the trigger’s message body.
  4. (Optional) trigger popups in the app — server-sent event from the bridge to the frontend with a popup payload; React handler renders the overlay.
  5. Author a custom trigger — beyond the built-in registry, e.g. “user clicked Upgrade Plan twice in 60 seconds.”
Until this page lands, the proactive triggers SDK reference and the Intercom recipe’s Step 2 cover the underlying primitives in detail. The Rasa-specific glue is the only piece this page will add.