Inspiration
Mental health support is often fragmented — a mood tracker here, a meditation app there, a separate journal, and no real connection between any of them. We wanted to build something that brought all of these tools together in one place, backed by AI that actually understands the user's context over time, rather than just another siloed wellness app.
What it does
MindSphere is an all-in-one mental wellness platform. It includes a Mind Journal, Mood Tracker, AI Health Assistant (Lyra), Real-Time Voice conversations, Diet & Nutrition planning with custom recipes, Exercise Plans, Assessments, Appointments, a Disturbance Detector, Meditation & Breathing tools, and a Sleep Tracker — all accessible from a single dashboard.
How we built it
The backend runs on FastAPI, deployed on Railway, with MongoDB Atlas as the database. The frontend is built with React and Vite, deployed on Vercel. The AI Health Assistant uses HuggingFace's Inference API (currently Qwen2.5-7B-Instruct), and real-time voice conversations are powered by a Gemini Live WebSocket relay. Payments are handled through Stripe, and transactional email runs through Resend.
Challenges we ran into
Setting up a reliable local development environment from scratch — FastAPI, React/Yarn, and MongoDB Atlas together — surfaced compatibility issues, including Python version conflicts and SSL certificate errors. In production, we hit MongoDB Atlas SSL/TLS connection errors on Railway, which we resolved through IP whitelisting, along with missing environment variable issues during deployment. We also migrated the core LLM provider from OpenAI to HuggingFace mid-project, while keeping the existing Gemini voice relay working without disruption.
Accomplishments that we're proud of
We're proud of building a genuinely comprehensive mental wellness toolkit as a solo effort — spanning AI conversation, health tracking, and lifestyle tools — and getting it fully deployed to production with working payments, email, and a stable infrastructure pipeline.
What we learned
We learned a lot about managing multi-service deployments (Railway, Vercel, MongoDB Atlas, Stripe, Resend) together, debugging cross-service connection issues, and the practical tradeoffs of swapping LLM providers without breaking existing features like real-time voice.
What's next for MindSphere
Next, we're focused on refining the AI assistant experience, expanding SEO and marketing reach, and continuing to iterate on the platform based on user feedback.
Log in or sign up for Devpost to join the conversation.