FitBuddy

FitBuddy is a free, open-source health tracker for Android, built for Indian diets and daily routines.

Snap a photo or type what you ate — an LLM estimates calories and macros instantly. Review ingredients, tweak weights with live macro recalculation, and save. Bookmark favourite meals as presets for one-tap logging next time.

Features:
• Smart logging — photo or text; AI parses food and exercise automatically
• Meal review — edit dish name, adjust ingredient weights with live macro recalc
• Presets — bookmark meals for one-tap future logging
• Dashboard — daily calorie ring, combined food/exercise log, macro breakdown
• Analytics — custom charts tracking trends over time
• Auto failover — rotates API keys and models on rate limits; stays on your preferred platform
• Offline fallback — works without an AI key via built-in simulator
• Material You — dynamic color theming on Android 12+
• Crash reports — optional anonymous Sentry reports (off by default; opt in via Settings)

AI providers (configured at runtime in Settings):
• OpenRouter — free vision models, automatic model dropdown
• Google Gemini — AI Studio key, OpenAI-compatible endpoint
• Ollama or any OpenAI-compatible server (OpenAI, LM Studio, vLLM, LocalAI) — local or cloud, optional API key

Without a configured provider, the app works offline using the built-in simulator. Photos require a configured AI provider.

Privacy:
• No ads, no tracking by default
• Crash reports (Sentry) are off by default on F-Droid builds; opt in under Settings
• Cloud backup (MongoDB Atlas) is off by default; opt in under Settings → Backup
• AI providers are contacted only when you choose to enable them

Updates on this F-Droid build are delivered by F-Droid. In-app GitHub APK downloads are disabled in this flavor.

Note: The F-Droid APK uses F-Droid's signing key and cannot replace a GitHub or Play Store install in place.

Source: https://github.com/anantdark/FitBuddy
License: GPL-3.0-only

• fix: lay out the AI provider picker as a 2x2 grid
• fix: detect vision models on local OpenAI-compatible servers