A telephony test should exercise the real phone path without creating real customer work. Use controlled caller numbers, fictional business records, isolated notifications, and an explicit cleanup plan.
LiveKit's telephony-testing guide recommends inspecting the call path through rooms, SIP participants, and logs. Add business-action verification to that transport evidence so the test proves more than audio connectivity.
Define fixture identity and side effects
Create a test register with called number, authorized inbound configuration, intended agent, client context, fictional customer, expected outcome, and allowed remote destinations. Keep live appointment creation, staff alerts, media retention, and usage billing under the approved test policy.
Use real controlled numbers for connectivity while keeping the scripted caller identity fictional. Do not dial unrelated numbers or send confirmation texts to invented contact details that might belong to someone else.
Work through a fictional property call
A controlled tester calls the staging business number and requests a visit for a fictional customer with two properties. The expected path selects the intended agent, asks which property, and creates the approved test request or appointment.
Inspect the room and participant context, agent logs, destination record, and staff queue. A successful call that writes to production should fail the acceptance test even if the conversation sounds correct.
| Fixture | Required result |
|---|---|
| Known fictional customer | Correct approved test record |
| Ambiguous property | Clarification before action |
| Unknown called context | Safe rejection or review |
| Caller hangs up during write | Recoverable uncertain action |
| Staff transfer unanswered | Approved test fallback |
Test failures deliberately
Exercise rejected credentials, unavailable agents, incorrect routing, repeated events, and interrupted remote writes. Document which failures are simulated and which occur through the real controlled dependency. Preserve enough identifiers to locate the first broken boundary.
Full phone-path tests cost more and may be slower than local checks, but they expose configuration and provider differences. Smaller unit or logic tests remain useful for business rules. Use both at the boundary they can actually verify.
Rerun affected fixtures after changes to dispatch, phone configuration, agent names, or business actions. Review multi-number dispatch for client-context isolation scenarios.
Questions about fictional telephony tests
Can all testing happen in a browser room?
Browser testing can verify parts of the agent, but it does not exercise the complete telephone route.
Should test calls be excluded from client invoices?
Apply the explicit agreement and preserve environment attribution. Provider consumption may still occur.
What should the test report include?
Fixture version, route, environment, expected and observed outcomes, relevant identifiers, side effects, cleanup, and unresolved limitations. Avoid claims of production readiness from one successful call.