Real-time conversation analysis, intelligent client insights, and professional dashboards for coaches.
MVP Live: Live
Vital is a scalable AI Dashboard that turns raw client-coach conversations into structured, actionable intelligence. minimal hallucination from the AI. Coaches get clean, evidence-backed insights. Clients self-update their activity data.
Built for real healthcare professionals.
- Conversation to AI powered Dashboard: Paste client-coach dialogue → get structured JSON output with custom schema + strict instructions.
- Hallucination Control: Lowered temperature + optional fields + proper JSON schema enforcement. Arithmetic (steps, sleep, water intake) will be moved to deterministic functions soon.
- Dual Dashboard:
- Client View: Simple table to log daily activity.
- Coach View: Clean, trackable overview with attached evidence blocks. No more chasing clients.
- Security Layer: Explicit isolation between backend and frontend to prevent direct backend exposure/interception.
- Multi-AI Fallback: Horizontal scaling via multiple free AI providers. One dies or runs out of tokens? Auto-switch. Also lets us route to the best model for medical-grade accuracy.
- Frontend: Next.js (TypeScript)
- Backend: FastAPI (Python)
- AI Layer: Structured output with strict JSON schemas + system prompts
- Architecture: Decoupled frontend + backend with dedicated security boundary
vital/
├── backend/ # FastAPI service
├── frontend/ # Next.js application
└── README.md
Test phase ready.
MVP is fully functional and ready for real healthcare professionals to hammer on it. Several fields still need refinement - we're not pretending it's perfect.
Next Moves (no fluff):
- Move all arithmetic calculations out of AI into pure functions (saves tokens + kills hallucinations)
- Polish remaining dashboard fields
- Production hardening
- Expand multi-AI routing intelligence
Most AI health tools are fancy wrappers that hallucinate and waste everyone's time. Vital is engineered to be different: strict schemas, evidence attachment to allow coaches easily verify AI data, client self-reporting, and coach efficiency first.
cd frontend
npm install
npm run devcd backend
python -m venv venv
source venv/bin/activate
pip install -r requirements.txt
uvicorn main:app --reloadThis is early but serious. If you're a dev who can ship clean code, open a PR.