Call Control Asterisk 22.9 / ARI
Simulated, schema-accurateStasis app events as the call moves through the bridge. Tap any event for its JSON.
Live Transcript + AI
Scripted sample call, simulated audioEvery turn is sent to the AI endpoint. Chips under caller turns show intent, sentiment and entities.
Rolling summary
Waiting for the first analyzed turn.
Agent Assist + CRM
What the agent sees next, and the call state your CRM receives.
Suggested next reply
Appears after the first analyzed turn.
Routing decision
No decision yet.
Compliance
CRM call state
Simulated, schema-accurate- Lead
- Not yet
- Intent
- Not yet
- Plan interest
- Not stated
- Disclaimer
- Pending
- Recording
- Off
Latency, last reading
Round trips measured in your browser for each /api/analyze call.
This demo runs on a shared hosted model. On your own vLLM, speed follows your GPUs and model size.
How a live call flows through the AI layer
From the caller to your CRM, with the OpenAI-compatible model in the middle. Dashed boxes are simulated in this demo.
One request shape, two base URLs
The console sends the same OpenAI-style chat completions request. Moving from the demo to your vLLM changes the base URL and model name, not the request.
- Base URL
- Model
In this demo a server-side function reaches the model through its Workers AI binding with these same messages. Instructions and keys never reach the browser.
What each stage delivers
Mapped to the outcomes your platform is built to deliver.
Warm transfers with contextRouting rules and AI decide when a licensed agent must take over. ARI adds the agent to the bridge and the context packet arrives with the call.
Call insights in real timeIntent, sentiment, entities, rolling summary and a call score, per turn, from your own model.
Compliance while it mattersTPMO disclaimer before benefits, PHI mentions flagged, recording retained per 42 CFR 422.2274.
Outcome workflowsAMI events keep CRM call state current and trigger booking, follow-up SMS and team alerts.