FitBuddy — Building Better Habits One Week at a Time with Local AI

FitBuddy is a personal habit and lifestyle coach created for a friend who wants to become more active and consistent. The app gathers information about the user's wake and sleep times, activity level, preferred exercise, eating habits, routine goals, and desired length of journey. Using AI, FitBuddy generates a personalized, week-by-week plan to gradually improve the user's habits.

The core of this project is the utilization of an open-weight AI model, Gemma 3:1B, which runs locally through Ollama. This ensures that user data remains on the user's computer rather than being sent to third-party AI APIs. The key aspects of FitBuddy are:

1. Personalization: Focus on giving users the right next step rather than overwhelming them with too much information.

2. Gradual changes: Encourage consistency and focus on small improvements rather than expecting perfection from the start.

3. Progress tracking: Users can monitor their weekly progress, build streaks, and earn rewards for completing weeks.

The tech stack for this project includes HTML, CSS, JavaScript, Node.js, Express, Ollama, Gemma 3:1B, and LocalStorage. The developer learned valuable lessons about running open-weight models locally, connecting applications to local AI models, and handling user information with LocalStorage. Despite the limitations of small models, the project successfully turned a real person's challenge into a usable product powered by open AI.

Future developments may include more reliable habit progression, improved weekly personalization, real feedback from users, collectible weekly stickers, and a longer-term progress dashboard.