This dataset contains 10,000 raw public GitHub pull request traces. Each JSONL row represents one pull request candidate and joins together PR metadata, discussion, code review, changed files, full diff text, optional auxiliary fields, and retrieval provenance.
The data is designed for mining software engineering workflows of the form:
Pull Request -> Discussion -> Review -> Code Diff -> Merge