Every repository gets one primary subcategory and up to two more, from a fixed list (taxonomy v1.0.2) rather than free-form keywords.
Applied generative AI: agents, assistants, and apps built on large language models.
Model development: frameworks, training, research code and classic ML domains.
Running software: containers, Kubernetes, IaC, serverless and platforms.
Shipping and operating software: CI/CD, monitoring, logging, incidents.
Tools developers use to write, build, test and maintain code.
Native and cross-platform apps for phones and computers.
Moving bytes and messages between people and machines.
Programming languages, compilers, runtimes and core libraries.
Operating systems, embedded, GPUs, robotics and high-performance computing.
Software people run for themselves at home or on their own servers.
Creative software: graphics, audio, video, design and games.
Domain software for research, health, finance and other industries.
Distributed ledgers, smart contracts and crypto infrastructure.
Repositories whose main value is knowledge rather than software.
