SentimentAura

Full-stack live “Sentiment Aura” demo using React + FastAPI with Deepgram Realtime and OpenRouter.

Quickstart

1. Backend (FastAPI)

• Python 3.11+
• Create a virtualenv and install deps:
  • pip install -r backend/requirements.txt
• Set env (create .env in backend/ directory):
  • OPENROUTER_API_KEY=<your_key>
  • DEEPGRAM_API_KEY=<your_key>
  • ALLOWED_ORIGINS=http://localhost:5173
  • Optional:
    • OPENROUTER_MODEL=openai/gpt-4o-mini
    • OPENROUTER_BASE_URL=https://openrouter.ai/api/v1
    • REQUEST_TIMEOUT_SECONDS=12.0
• Run:
  • uvicorn app.main:app --reload --port 8000 --app-dir backend

2. Frontend (React + Vite + TS)

• cd frontend
• npm install
• Create .env.local:
  • VITE_BACKEND_URL=http://localhost:8000
  • VITE_DEEPGRAM_API_KEY=<your_key>
  • VITE_DEEPGRAM_WSS=wss://api.deepgram.com/v1/listen?model=nova-2&encoding=linear16&sample_rate=16000
• npm run dev then open http://localhost:5173

How it works

• Frontend streams mic audio to Deepgram Realtime over WebSocket (16k PCM).
• Finalized transcript segments are sent to FastAPI /process_text.
• Backend calls OpenRouter (provider-agnostic) to extract sentiment score/label and keywords.
• Visualization (p5.js) maps sentiment to color and energy; keywords cause subtle flares.

Scripts

• Backend: uvicorn app.main:app --reload --port 8000 --app-dir backend
• Frontend: npm run dev (in frontend)