Voice · RTC · Decision systems

Voice AI Gateway Simulator

A small, tested decision layer that turns simulated voice transcripts into explicit routing actions—without presenting a design study as a production voice platform.

4/4

Automated tests covering routing, workflows, fallback, and invalid events.

Node 20+

Dependency-free runtime using the built-in HTTP and test modules.

3 routes

Queue routing, workflow initiation, and clarification.

Architecture boundary

From call event to controlled decision

SIP caller
Asterisk
ARI adapter
Decision layer
Approved action

The public implementation covers the decision layer. The Asterisk and ARI blocks describe the intended integration boundary and are not included in the repository.

Interactive simulation

Test a transcript

This browser example mirrors the documented deterministic rules. It does not place a call or contact an AI service.

Structured result


      

Implemented

Event validation, keyword classification, deterministic routing, HTTP endpoints, CLI demo, and automated tests.

Not claimed

Live ARI connectivity, SIP registration, audio streaming, ASR, TTS, LLM inference, authentication, persistence, or high availability.

Next milestone

An isolated ARI adapter that converts approved Asterisk events into the public transcript contract without exposing PBX configuration.

Why this project matters

It connects deep Voice/RTC experience with a disciplined path into automation and applied AI: define the boundary, implement the smallest verifiable component, test it, and document what remains.