A Non‑AI GitHub Trending Project Worth Checking Out: The Exercises‑Dataset

Browsing GitHub Trending often yields AI‑related projects, but the author highlights a rare non‑AI open‑source repository called exercises‑dataset . The project, created by a Turkish developer, has amassed 6,800 stars in three months and offers a structured fitness‑exercise dataset.

The dataset contains 1,324 exercises, each with detailed step‑by‑step instructions translated into six languages (English, Spanish, Italian, Turkish, Russian, Chinese) and includes hundreds of GIF resources for visual guidance. Although the media files are not bundled in the repository, they can be downloaded from the CDN at static.exercisedb.dev/media/{media_id}.gif, and the author warns about potential copyright issues for commercial use.

Beyond the raw data, the repository provides a developer guide setup.html that automates three major tasks:

Database SQL generation: Supports SQL Server, PostgreSQL, MySQL, and SQLite. Clicking the generate button produces a .sql file with 1,324 INSERT statements ready for import.

API client code: Generates ready‑to‑use client snippets in JavaScript, Python, C#, Java, PHP, Go, and cURL. Users only need to supply the base URL, after which the examples update in real time and can be copied directly into projects.

LLM prompt creation: After selecting a backend framework (Express.js or FastAPI) and a database, the tool outputs a structured prompt that can be fed to ChatGPT, Claude, or Gemini, producing a complete, production‑ready REST API with a single copy‑paste operation.

The article also presents statistics about the exercise distribution: 292 upper‑arm, 227 leg, 203 back, 169 waist, 163 chest, 143 shoulder movements; 325 are body‑weight only, making them suitable for home‑fitness apps. Equipment categories include 294 dumbbell, 157 rope, and 154 barbell actions.

Finally, the author shares the GitHub address ( https://github.com/hasaneyldrm/exercises-dataset) and suggests that developers can hand the link to an AI agent to generate a small fitness app in about five minutes, provided they handle media assets separately.