Gymnasia is a mobile application built with React Native and Expo that works as an intelligent personal trainer. The app allows you to create and manage workout routines, track your diet and body measurements, and features two AI agents that act as a training assistant.
Both agents run entirely on the device, with no backend, so the user's data never leaves the phone, for privacy and security.
Workout Routines
Create custom routines with exercises, sets, reps and weights. Each routine can be categorized (strength, cardio, flexibility, etc.) and customized with icons. During training, a rest timer with sound alert notifies you when to continue.
Exercise Library
Open-source exercise repository with AI-generated images showing correct execution, in both male and female versions. Exercises include muscle group, required equipment, difficulty level and detailed instructions. Users can also create custom exercises.
Conversational coach (BYOK agent)
Built-in chat that works as a virtual personal trainer, with adapters for OpenAI, Anthropic and Google. It is BYOK (bring your own key): the user supplies the API key and it is stored on the device. The agent has 12 local tools with access to your routine context, and its system prompt is fetched remotely with an offline fallback, so the app keeps working without a connection.
Vision subagent and diet tracking
Daily meal logging with calorie and macronutrient calculation (protein, carbs and fats). A vision subagent estimates macronutrients from food photos, and the app scans barcodes against OpenFoodFacts. Set personalized nutritional goals.
Body Measurements
Track weight and other body measurements with historical data to visualize your progress.
The app has no backend: both agents run entirely on the phone, so the user's data never leaves the device. The project is an npm workspace monorepo containing:
The mobile app uses AsyncStorage for local persistence and Expo SecureStore for API keys. Exercise images are served from the GitHub repository and cached on-device.
How both agents are built, with the architecture graph of each one, is explained in the post architecture of the Gymnasia AI agent.
The project is open source and available on GitHub.
If you like the project, consider giving it a star on GitHub.
How the Gymnasia AI agent is built: a BYOK conversational coach with 12 local tools and a vision subagent that estimates macros from photos, both running entirely on the phone with no backend.
Third and final part of the Deep Research with LangGraph series. Implement the Writer agent that drafts the final report...
Second part of the Deep Research with LangGraph series. Build the Research Supervisor that coordinates several Researche...
Let's talk.
maximofn@gmail.com
Machine Learning and AI specialist. I develop solutions with generative AI, intelligent agents and custom models.
AI agents, powered by LLMs, promise to transform applications. But are they simple executors today or future intelligent collaborators? To reach their...
Learn to create an IA system to execute efficiently on a device
Hugging Face spaces allow us to run models with very simple demos, but what if the demo breaks? Or if the user deletes it? That's why I've created docker containers with some interesting spaces, to be able to use them locally, whatever happens. In fact, if you click on any project view button, it may take you to a space that doesn't work.
Let's talk.
maximofn@gmail.com
Machine Learning and AI specialist. I develop solutions with generative AI, intelligent agents and custom models.
Dataset with jokes in English
Use: Fine-tuning text generation models for humor
Dataset with translations from English to Spanish
Use: Training English-Spanish translation models
Dataset with Netflix movies and series
Use: Netflix catalog analysis and recommendation systems
