HaleView is my independent project. I built the frontend and backend around a health profile, activity records, progress charts, and the Hale assistant.
My contribution
The implementation includes profile setup, guest and account access, progress records, and the guidance interface. I also worked on the nutrition and shopping workflows and the account lifecycle.
The frontend uses React and TypeScript. The backend uses Express and SQLite. Docker brings the application and its configuration together for local use.
The main decision
The core calculations do not depend on a language model. Profile metrics, progress, and local guidance should remain available when online AI is off. The user can separately allow online guidance from Hale.
This separation makes the application's behavior easier to inspect. It also keeps an optional external service from becoming a requirement for basic use.
How I approached it
I separated the guest and account storage paths. Guest data stays in the browser; account data is handled by the backend. I then connected profile setup, records, and progress views to those paths.
The portfolio preview uses a synthetic guest profile. It demonstrates the interface and does not show my health data or another person's records.
See the HaleView project page for the interface, stack, and launch instructions.