How AI SDK maintenance and community facilitation is changing #17243

We are changing how we maintain AI SDK.

More maintenance work will increasingly be supported by our software factory, including reviewing new issues, triaging reports, reproducing bugs, and preparing bug fix pull requests. As part of that shift, we want to be clearer about which contributions are most helpful and how we credit them.

High-quality issues are one of the most valuable ways to contribute. Clear bug reports, minimal reproductions, failing tests, precise documentation feedback, thoughtful feature requests, and community context help us understand problems and improve AI SDK for everyone.

For bug reports, we now distinguish between:

• Production/usage bugs: You encountered a problem while using AI SDK. Please share what you were building, how and why you were using the SDK, what happened, and relevant code, logs, versions, or reproduction steps. When your report leads to a bug fix pull request, we will continue to credit you with co-authorship. This includes problems encountered during development.
• Audit bugs: You used agents to scan repository code and generate bug reports without providing additional usage context beyond what is already available in the repository. We will no longer grant co-authorship for these issues.

Using an agent to investigate or describe a problem you encountered while using AI SDK does not make it an audit bug. Likewise, creating a reproduction for an audit finding does not by itself make it a usage bug. The distinction is the real usage context you contribute.

If you find bugs through an agent audit, please contribute the prompts you used instead. Open a pull request adding a reusable prompt to contributing/audit-prompts/*.md or improving an existing prompt. Explain what it checks, what it found, and what makes it valuable. If it is a new valuable prompt or a meaningful improvement to an existing prompt, we will accept the pull request and grant co-authorship for the prompt contribution. Sharing prompts lets us repeat useful audits and improve them together.

Pull requests are still welcome, especially for documentation fixes, reproduction examples, failing tests, audit prompts, and small focused changes that clarify an issue. Full implementation PRs for bugs are discouraged, because maintainers and automated workflows may be able to prepare those fixes once the issue is clear. We will also credit pull request authors when their work adds valuable input to resolving a usage bug or improves the quality of its fix.

We value our community and all forms of contribution: bug reports, feature requests, documentation help, community projects, provider implementations, examples, discussions, and feedback.

This issue is locked so it can stay focused as an announcement. We will post comments as we evolve our practices. Please use the new Feedback issue type to share feedback, concerns, or ideas about how we can improve AI SDK for the community.