Back to Projects
Featured Case StudyAI Swarm Agent

Fawn: AI Phone Receptionist

Fawn is an AI-powered voice receptionist for daycare centers. Using Google's Gemini Native Audio, she fields Twilio calls, carries out complex conversations with parents, books tours via Google Calendar, generates structured PDF intake papers, and syncs data to an administrative Director's Dashboard in real time.

Interactive System Simulator

Trigger an incoming call to observe the full-stack, real-time sync telemetry.

IDLE STATE

Call Simulator Ready

Click "Simulate Inbound Call" above

Speech transcript will compile here once call starts.
server.js log
Logs will stream here in real time...

DIRECTOR'S DASHBOARD

Pusher Channel: Connected
Active Calls0
Total Leads23
Tours Scheduled17
Intake Packets12

Recent Inbound Leads

ParentRoomStatus
Marcus BrodyInfantRegistered
Jessica MillerPreschoolIntake Document Generated
David ChenToddlerRegistered

Autonomous Live Action Feed

Classroom Toddler ratio updated: 8/10 kids.
System initialization checks complete.

System Architecture & Pipeline

1. Inbound Carrier

Twilio answers incoming calls and redirects audio packets to our endpoint.

2. WebSocket Server

Node.js server receives raw audio streams from Twilio and establishes bidirectional connection to Gemini.

3. Gemini Live Audio

Native Audio model listens, processes user voice directly, runs tool workflows, and talks back in real time.

4. Pusher Sync

Events (scheduling, leads, intake paperwork) broadcast over WebSockets to trigger instantaneous dashboard updates.

Google Sheets & Calendar SDK

Uses Google Cloud Node.js API with a secure service account credential system. Autonomously updates leads tracker Sheets and schedules tours on the Daycare calendar, keeping operations synced.

Gemini Native Audio API

Uses Gemini 1.5 Flash's Native Audio API rather than traditional TTS-STT pipelines. This directly feeds raw audio waves to Gemini, enabling natural tone-shifting, whisper handling, and ultra-low latency calls.

Intake Document Generation

Takes structured conversational outputs, converts them to JSON schemas via structured Gemini outputs, and generates fully formatted PDF intake documentation ready to be signed.