Skip to content

msmel01/waterloo-ai-agents-hackathon2026

Repository files navigation

Valentine Hotline

Built for Voice AI Hackathon Waterloo 2026.

Valentine Hotline is a voice-first AI screening pipeline for dating intros:

Suitor calls in -> AI interview -> transcript + scoring -> Date / No Date

Why It Feels Fast

Valentine Hotline is tuned for ultra-efficient voice interactions using Smallest.ai for low-latency speech handling in the interview loop.

Repo Map

  • backend/ -> FastAPI API, LiveKit agent, async workers
  • frontend/ -> React + Vite app
  • locustfile.py -> load testing profile
  • render.yaml -> deployment setup

Stack

  • Backend: FastAPI, SQLModel/SQLAlchemy, Alembic, Redis, arq
  • Frontend: React, TypeScript, Vite, Tailwind
  • Voice: LiveKit + Smallest.ai STT/TTS + OpenAI conversation engine
  • Scoring: Anthropic Claude
  • Infra: Neon Postgres + Render

Quick Start

  1. API
cd backend
uv sync
uv run uvicorn src.main:app --reload
  1. Voice Agent (new terminal)
cd backend
uv run python -m livekit.agents start agent.main
  1. Worker (new terminal)
cd backend
uv run arq workers.main.WorkerSettings
  1. Frontend
cd frontend
npm install
npm run dev

Environment

  • Configure backend/.env
  • Configure frontend/.env.local
  • API base path is /api/v1

More Docs

  • backend/README.md
  • frontend/README.md

About

Voice-first AI dating screener built for Voice AI Hackathon Waterloo 2026: LiveKit interview agent with Smallest.ai speech, FastAPI backend, and React dashboard for scoring, verdicts, and bookings.

Topics

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

 
 
 

Contributors