The exercises-dataset repository is a comprehensive fitness exercise database containing 1,324 exercises designed to serve as a ready-to-use data layer for fitness applications. The dataset is primarily written in HTML and powers the LogPress app, an AI-assisted workout tracker. Each exercise entry includes an animation GIF, a 180×180 thumbnail image, metadata categorizing the exercise by body part and equipment, and step-by-step instructions available in 10 languages: English, Spanish, Italian, Turkish, Russian, Chinese, Hindi, Polish, Korean, and French.
The repository provides structured data organized in a JSON format located at data/exercises.json, accompanied by a formal JSON Schema file for validation purposes. The dataset covers a wide range of fitness needs, with exercises distributed across multiple body parts including upper arms (292 exercises), upper legs (227), back (203), waist (169), chest (163), and shoulders (143), among others. Equipment variety is substantial, with body weight exercises comprising approximately 25 percent of the dataset (325 exercises), while other common equipment types include dumbbells (294), cables (157), barbells (154), and various machines and specialized equipment.
The repository includes two interactive HTML tools that require no server setup. The index.html file provides a fully client-side exercise browser with live search functionality across all 1,324 exercises, filtering capabilities by category, equipment, and target muscle, and an infinite scroll grid interface. Users can click any exercise card to view full details and instructions in any of the ten supported languages. The setup.html file serves as a developer integration guide, offering database setup instructions with SQL generation for SQL Server, PostgreSQL, MySQL, and SQLite, API integration examples in JavaScript, Python, C#, Java, PHP, Go, and cURL, and LLM-assisted backend generation prompts for popular frameworks including Express.js, FastAPI, ASP.NET Core, Spring Boot, Laravel, and Gin.
The data schema is comprehensive, with each exercise record containing fields for unique ID, name, category, body part, equipment type, target muscle, secondary muscles, instructions in all ten languages, media references, and timestamps. Media assets including thumbnails and animation GIFs are sourced from Gym Visual with proper attribution included in each record. The repository explicitly maintains attribution information and media IDs for all visual content.
GitGenius tracking shows the repository has experienced steady growth, with stargazers increasing from 18,306 to 18,314 since the previous check on August 1, 2026. Issue and pull request response latency shows a median of 9.4 hours and a mean of 36.0 hours across eight tracked items. The most active contributor is hasaneyldrm with 9 recorded events, followed by KATE-xue75 with 2 events and Akianonymus with 1 event. The repository shares overlapping contributors with several other projects including shubhamsaboo/awesome-llm-apps, usestrix/strix, and egonex-ai/understand-anything, indicating active participation in the broader developer ecosystem.
The dataset is licensed under MIT for code and data, with separate terms governing the media assets from Gym Visual. The repository is designed for multiple use cases including fitness and workout planning applications, machine learning projects involving exercise recognition or recommendation, health and wellness research, and educational demonstrations.
