Felt Sense is an AI Assistant that helps you navigate your emotional life through patient natural conversations and emotional intelligence.
Felt Sense is a modern web application that provides an AI-powered emotional support companion. It combines speech recognition, natural language processing, LLM intelligence and voice synthesis to create a seamless conversational experience focused on emotional well-being.
- 🎙️ Speech-to-text transcription using Deepgram
- 🤖 Advanced AI responses powered by Claude 3.5 Sonnet
- 🔊 Natural voice synthesis via OpenAI TTS
- 🔐 Secure authentication with WorkOS AuthKit
- 💾 Persistent conversation history
- 🎯 Real-time emotional analysis and support
- Frontend: Next.js, React, TailwindCSS
- Backend: Next.js API routes (serverless)
- Database: Neon (Postgres)
- Authentication: WorkOS AuthKit
- AI/ML:
- Claude 3.5 Sonnet (LLM)
- Deepgram (Speech-to-Text)
- OpenAI TTS (Text-to-Speech)
felt-sense/
├── docker/ # Container configuration
├── prisma/ # Database schema and migrations
├── src/
│ ├── app/ # Next.js app router pages and components
│ │ ├── api/ # API route handlers
│ │ ├── auth/ # Authentication related components
│ │ ├── chat/ # Chat interface pages
│ │ ├── components/ # Reusable React components
│ │ ├── store/ # State management (Zustand)
│ │ ├── page.tsx # Landing page with hero section
│ │ └── layout.tsx # Root layout with metadata
│ ├── lib/ # Core application logic
│ │ ├── api/ # API client functions
│ │ ├── db/ # Database interactions
│ │ └── utils/ # Utility functions for AI services
│ ├── types/ # TypeScript type definitions
│ └── middleware.ts # Auth protection & routing middleware
├── public/ # Static assets
├── .env # Environment variables (gitignored)
└── config files # Various configuration files at root
The application uses Next.js API routes for serverless backend functionality:
-
Authentication (
/api/auth/
)GET /api/auth
- Get current sessionGET /api/auth/callback
- Handle OAuth callback
-
Language Model (
/api/LLM/
)POST /api/LLM
- Process conversations with Claude 3.5 Sonnet
-
Speech-to-Text (
/api/stt/
)POST /api/stt
- Convert audio to text using Deepgram
-
Text-to-Speech (
/api/tts/
)POST /api/tts
- Generate speech from text using OpenAI TTS
- Clone the repository:
git clone https://github.com/your-repo/felt-sense.git
cd felt-sense
- Install dependencies:
npm install
- Set up environment variables:
cp .env.example .env
Required environment variables:
- DATABASE_URL="your-neon-db-url"
- WORKOS_API_KEY="your-workos-key"
- WORKOS_CLIENT_ID="your-workos-client-id"
- ANTHROPIC_API_KEY="your-claude-api-key"
- DEEPGRAM_API_KEY="your-deepgram-key"
- OPENAI_API_KEY="your-openai-key"
- Initialize the database:
npx prisma generate
npx prisma migrate dev
- Start the development server:
bun run dev
or
npm run dev
- Access the application at
http://localhost:3000
.