Skip to content

Test the real surface

Testzilla drives the agent you actually shipped, on the surface your users actually touch. The results reflect what your users get, not a simulation.

What “real surface” means per connection type

Section titled “What “real surface” means per connection type”
  • Phone (Inbound) — Testzilla dials in to your agent’s phone number and runs the AI tester on the call. The full voice pipeline — telephony, ASR, LLM, TTS — is exercised exactly as a real caller would experience it, including call routing, IVR behaviour, and transfer flows.
  • Web Chat — Testzilla navigates to your page in a headless browser, detects the chat widget, opens it, and talks to it through the real DOM the same way a visitor would. It tests widget open/close behaviour, iframe rendering, and message display — not just the LLM underneath.
  • Web Voice — Testzilla launches a real browser, loads the provider’s web-call SDK, establishes a WebRTC call, and conducts the conversation with synthesized tester audio. It validates the web embed: WebRTC media negotiation, ASR, LLM, and TTS in a browser context.
  • LLM Chat — talks to the agent’s text LLM directly via the provider’s chat API, turn by turn. The fast, cheap path for verifying conversational logic.

A test that only hits an API can pass while the real widget is broken. Testzilla exercises the path your users take, so a regression in the browser embed, the telephony stack, or the widget UI shows up in a test rather than in a support ticket.