AI Policy#
FastStream welcomes contributions made with AI assistance. You are responsible for all code, descriptions, comments, and reviews submitted under your account.
Code and licensing#
Understand and test generated code before submitting it. Check that generated tests exercise the intended behavior and can catch a broken implementation. You must have the right to contribute all submitted material under FastStream's Apache 2.0 license and sign the Contributor License Agreement.
Issues, pull requests, and reviews#
Check AI-drafted text against the code and your own understanding. Explain your changes and respond to review questions in your own words; forwarding an AI answer without checking it is insufficient. Mention significant AI assistance in the PR description.
Reviews must reflect your own examination of the diff. Generic praise, restatements of a PR description, and unchecked AI approvals do not help maintainers assess a change. Maintainers may hide such reviews and restrict accounts that repeatedly post them.
AI can help with language editing and translation. When sharing raw AI output, quote it, identify its source, and explain why it is relevant.
Automated accounts and contribution farming#
We do not accept PRs from automated contribution farms: accounts or groups of accounts used to mass-submit changes across repositories.
Maintainers may consider an account's public contribution history, including mass PR submissions to many unrelated repositories. This applies even when a person operates the account or discloses AI use. Based on that activity, maintainers may reject its PRs to FastStream, restrict its participation, or block the account through the GitHub organization. An individual PR passing tests does not guarantee that we will accept contributions from that account.
Routine maintenance automation approved by maintainers, such as Dependabot, is permitted.
First contributions#
Contributors without write access may have at most one open PR, including drafts, until their first PR is merged into FastStream. After that first merge, this limit no longer applies. The contribution farming policy still applies.
Maintainer discretion#
Maintainers may close or deprioritize contributions when the author cannot explain or test them, or engage with feedback. Every contribution needs a human who takes responsibility for it.
Adapted from AG2's AI Policy, with FastStream rules for contribution farming and first contributions.