Webgit
Version Control System over http.
Machine learning requires managing big quantities of information that keeps changing over time and must be utilized concurrently by many users for different purposes, which requires a version control system.
Git has become a standard, but it is designed for managing source code (text), and the information must be cloned to access it (distributed). This means, for example, that for making a training over certain version of the dataset, it is required to download all files.
We would like to design a tool that stores, manages and version files; by using the same protocol than Git (object, commit, branch, merge, tag…), but optimized for the following cases:
- Allow upload and download of large files in a paginated and concurrent manner (optimized for large files).
- Allow the users to manipulate the repository on remote, without the need to download the whole project (remote editing of the files through GET/PUT/PATCH).
- Allows downloading a file quickly, without the need of a checkout or diff calculations (be fast and do not create a bottleneck for a training).
- Complexity:6/10
- Technical specifications:Databases, Control Version System
- CATEGORY 2024-2025
- TAGS Git , go , HTTP




