VoiceAssistant
A progressive web app that lets users create, assign, and track tasks entirely through voice — powered by Groq's ultra-low-latency LPU inference engine.
A sleek, mobile-first AI assistant built on the modern T3 stack.
The app was designed around a single principle: voice should be faster than typing. Users speak naturally, and the AI handles transcription, intent parsing, and task execution — all within a single conversational turn. As a PWA, it installs from the browser and runs like a native app with offline capabilities and home-screen presence.
The Application
Voice Interface — Tap to speak and control your workspace with voice commands
Tech Stack
Framework
- Next.js 15 (App Router)
- React 19
- Create T3 App
Styling & UX
- Tailwind CSS v4
- PWA (next-pwa)
- Geist Font
Backend
- tRPC
- Prisma ORM
- PostgreSQL
AI & Voice
- Groq SDK (LPU)
- Whisper STT
- Llama LLM
Core Capabilities
Voice-First Interface
Hands-free interaction powered by real-time speech recognition.
- Browser-native Web Speech API captures voice input with zero setup required.
- Real-time transcription streams words as you speak — no waiting for processing.
- Contextual wake-word detection keeps the assistant responsive without constant listening.
Ultra-Low Latency AI
Groq's LPU inference engine delivers near-instant responses.
- Groq SDK processes speech-to-text via Whisper models at hardware-accelerated speed.
- LLM responses generated through Llama with sub-second latency for natural conversation flow.
- Streaming responses render token-by-token so users see answers forming in real-time.
Type-Safe Full Stack
End-to-end type safety from database to UI with zero runtime overhead.
- tRPC eliminates API boilerplate — procedures are called like local functions with full IntelliSense.
- Prisma ORM generates typed database clients with migration and seeding workflows built in.
- T3 stack convention ensures consistent patterns across the entire codebase.
Progressive Web App
Installable, offline-capable, and feels native on any device.
- Service worker caching enables offline access and instant repeat loads.
- manifest.json configures home-screen install with custom icons and splash screens.
- Mobile-first viewport locked to prevent pinch-zoom — behaves like a native app.