Software datasets support mining software repositories research and code models. The list includes the Public Git Archive of 182,014 top-bookmarked GitHub repositories, along with derived sets from source{d}: 1.3 billion commit messages, 25.3 million pull request review comments, labeled Java code duplicates and 41.7 million source code identifiers. Libraries.io covers open source package and dependency metadata, and FLOSSmole collects data about open source development.
Newer entries measure the ecosystem: install counts for 64,490 VS Code extensions, AI contribution policies of popular repositories, API pricing and MCP server resource use. The dataset licence does not override the licence of the code inside it, and repositories carry many different licences, so check before training models or redistributing source. Commit and review data can include developers' names and email addresses.