FitQuest AI: Turning Screen Time into Outdoor Quests
FitQuest AI, created by Neeraj Bhandari and Mukesh Kumar for Hacktoberfest, uses local open-weight AI model qwen2.5:3b to gamify outdoor fitness activities.
Stock photo for illustration only, not from the actual event
- FitQuest AI turns outdoor activities into bite-sized RPG quests
- Powered by local open-weight AI via Ollama and Pydantic
- Privacy-conscious design with local SQLite conversation history
- Developed collaboratively by Neeraj Bhandari and Mukesh Kumar
In an era where developers and desk workers spend countless hours staring at digital screens, software creators Neeraj Bhandari and Mukesh Kumar have introduced FitQuest AI. Built as a submission for Week 1 of the Hacktoberfest Open-Source AI Challenge under the theme Touch Grass, the project aims to help computer users step away from their monitors and head outside for physical movement.
Unlike traditional fitness tracking applications that encourage users to constantly monitor endless metrics on their mobile screens, FitQuest AI takes a reverse approach. The application utilizes artificial intelligence to generate short RPG-style quests based on walks, runs, hikes, or cycling sessions. Users generate a quest, put their phones away to experience the real world, and return later to log activity, earn XP, unlock achievements, and share milestones with friends.
Stock photo for illustration only, not from the actual event
The backend architecture of FitQuest AI prioritizes user privacy by avoiding external hosted AI inference APIs. Instead, it connects directly to the open-weight model qwen2.5:3b through Ollama. To ensure predictable outputs, Pydantic validates all structured data generated by the model before the application consumes it. Furthermore, conversation history is securely stored locally using SQLite, enabling the virtual coach to utilize workout context and perceived effort for more relevant responses.
Integrating local language models into gamified fitness applications highlights a practical shift toward edge computing and privacy-first software development. By running models locally rather than relying on cloud APIs, developers can protect user location and health data while maintaining full control over core application business logic.
While the language model generates creative quest narratives, core application rules such as XP rewards, level progressions, streaks, and achievements are strictly handled by traditional backend logic. On the frontend, an HTML Canvas-based renderer combines trail photographs, workout statistics, and XP details into a single shareable image directly in the browser, eliminating the need for complex server-side rendering pipelines.
Released under the MIT License, FitQuest AI invites developers to explore its GitHub repository, run the project locally, and contribute feedback. Setup instructions involve cloning the repository, downloading the qwen2.5:3b model via Ollama, and configuring the FastAPI backend alongside the Vite frontend environment.
Source: Dev.to
Found something wrong in this article? Report an issue with this article
Comments
Leave a Comment