🧸 Solace.ai
A real-time, voice-based AI companion designed to provide emotional comfort and gentle companionship to children in hospice care.
🎯 Project Overview
Solace.ai is not just a chatbot—it is a voice-first companion built to feel like a calm, caring presence.
It allows children to:
speak naturally (no buttons required) receive warm, emotionally-aware responses engage in continuous, back-and-forth conversation
The system is designed to feel like:
a plush companion, a bedside friend, a comforting voice that stays with you
💡 Why This Matters
Children in hospice care often experience isolation, fear, and emotional distress.
Solace.ai provides:
a constant, gentle presence emotionally safe responses a natural interface (voice) that requires no physical effort 🧠 Core Concept
Traditional systems:
speech → response
Solace.ai:
continuous, emotionally-aware conversation
Key difference:
prioritizes presence over productivity focuses on comfort, not tasks 🏗️ System Architecture 🧠 Backend (AI Brain)
Built with FastAPI, the backend handles the full real-time voice pipeline:
🔁 Core Pipeline Wake word detection Voice activity detection (records until silence) Speech-to-text (Whisper) LLM response generation (OpenAI) Text-to-speech (OpenAI TTS) Audio playback Follow-up listening (multi-turn conversation) ⚙️ Key Features Multi-turn conversation No wake word required after initial activation Interrupt handling User can speak while AI is talking Emotional safety layer Detects distress signals and adapts responses Conversation memory Maintains recent context for natural dialogue Real-time event streaming WebSocket-based updates to frontend 🌐 Frontend (Control Dashboard)
Built with React + Vite
Purpose: Start / stop / reset the system Display real-time system state: listening thinking speaking Show: user transcript AI responses live logs
🧩 Design Philosophy Laptop = Brain Runs all AI components for speed and flexibility Frontend = Interface Provides transparency and control (Future) Raspberry Pi + Plush Physical embodiment of the companion Acts as an interaction trigger device 🔁 Interaction Flow User says wake word System begins listening Speech is recorded and transcribed AI generates a gentle, emotionally-aware response Response is spoken aloud System waits for follow-up Conversation continues naturally 🐒 Personality: Rhasspy
Solace.ai features a soft, child-friendly character:
Rhasspy — a kind monkey with a red cape
Design goals:
warm and calming tone gentle storytelling emotionally supportive responses encourages continued conversation 🚀 Current Status ✅ Core voice pipeline implemented ✅ Backend fully functional ✅ Frontend integrated and polished ✅ Real-time interaction working ✅ Multi-turn conversation enabled ✅ Demo-ready 🔮 Future Improvements 🧸 Physical plush integration (Raspberry Pi trigger system) ❤️ More advanced emotional detection (tone + sentiment analysis) 🏥 Clinical feedback integration (hospice care specialists) 🌐 Deployment for broader accessibility 🎭 Expanded personality behaviors and storytelling depth 🛠️ Tech Stack
Backend
FastAPI OpenAI API (LLM + TTS) Whisper (Speech-to-Text) WebSockets
Frontend
React Vite Tailwind (optional styling layer)
Hardware (Planned)
Raspberry Pi 4 Model B
Bluetooth speaker
Plush companion enclosure
This project is designed with emotional safety as a priority.
Responses are constrained to be: non-harmful age-appropriate comforting The system is not a replacement for human care, but a supportive supplement 📌 Getting Started (Basic)
cd backend pip install -r requirements.txt uvicorn main:app --reload
cd frontend npm install npm run dev
Then open:
http://localhost:5173 🤝 Acknowledgements
Built as part of a hackathon project focused on:
human-centered AI accessibility emotional well-being 📬 Contact
For questions, collaboration, or feedback:
GitHub Issues Direct outreach