Gemini Live Agent Challenge Entry

Talk to your
medical records.

Ask about your labs, medications, conditions, and history in natural speech. Exora Voice answers in real time and puts the right data on screen -- without you touching anything.

Not a chatbot. Not a search interface. The AI controls the full display surface -- navigating screens, opening charts, and comparing results side by side as you speak.

Try the Demo

Pre-loaded demo account -- no signup required

Powered by Gemini Live API

Medical records are written for doctors, not patients. Dense clinical language, scattered across dozens of documents, from providers who never share data. Exora Voice lets you have a real conversation with your health history -- and see the relevant data materialize on screen as you ask.

How It Works

One tap. Then just talk.

1

Tap the Mic

A minimal voice bar slides up. The AI greets you and already knows which screen you are on and what data is visible.

2

Ask Anything

Ask about a specific lab result, a condition, a medication, your history. Interrupt at any time. The AI responds in natural speech.

3

See It Live

The AI navigates the app, opens detail charts, and compares results side by side -- all on the main screen, in real time.

Built for the Live Agents category

Every judging criterion addressed.

Barge-in Interruptions

Client-side VAD detects speech the moment you start talking. Audio stops immediately -- no waiting for the server round-trip.

Screen-Aware Context

The AI knows what is on screen at all times. Follow-up questions are answered from screen context without re-fetching data.

AI Controls the Display

Function calling drives real-time app navigation. The AI opens charts, switches screens, and compares panels side by side.

Persistent Conversation

Soft reconnect every four turns preserves full conversation history while keeping latency low. Context never resets mid-session.

System Architecture

Gemini Live API at the center. Real-time, bidirectional, function-calling.

EXORA VOICE -- SYSTEM ARCHITECTURE USER Browser Expo Web (voice.exora.au) MIC AUDIO AI VOICE GOOGLE AI Gemini Live 2.5 Flash Native Audio WebSocket -- Bidirectional Audio TOOL CALLS SCREEN CONTEXT APP LAYER VoiceGlobal navigate_to / get_labs open_panels / get_conditions 12 entity types + multi-panel Live Display Surface Screen navigation | Lab charts | Multi-panel comparison AI controls the full display surface -- user never touches the screen DATABASE Supabase 24 clinical tables -- RLS protected health data fetch GOOGLE CLOUD Cloud Run AI processing pipeline -- Sydney region SESSION MANAGEMENT Soft reconnect every 4 turns -- injects full conversation history -- AudioContext preserved -- latency stays low Client-side VAD barge-in -- screen context injected silently -- tool dedup prevents double-execution

Try the Demo

Log in with the credentials below to explore a pre-loaded health record with voice

Email
voice@exora.au
PIN Code
See testing instructions
Enter the App

The demo account contains a pre-processed health record with lab results, conditions, medications, vitals, allergies, and a healthcare timeline. Tap the mic button once inside the app and start speaking. Try asking about a specific lab result, a condition, or requesting a comparison of two results.

Medical Disclaimer: Exora Voice helps users understand their existing medical records through natural conversation. It does not provide medical advice, diagnoses, or treatment recommendations. All responses are grounded in the user's health record data. Always consult a qualified healthcare professional for medical decisions.