Conflict of interest as a referee?

Disclose your possible conflict to the editor and let them decide. "I had no direct role in the submitted work but am currently collaborating with Drs. Moe, Curly, and Larry on separate projects on plumbing, eye protection, and suit integrity."

It may be that your field is narrow enough that the editor will prefer your mostly impartial review, or they may prefer to find another referee.

This suggestion of course assumes that you think you can offer a fair review whether or not you find the work suitable for publishing. If you think for any reason you would be unwilling to reject the work if it is poor because of your collaborations, you should not agree to review at all.

If they're your current collaborators, you should not review. Not many journals list formal policies on conflict of interest, but here's an example from PNAS (emphasis mine):

When asked to evaluate a manuscript, reviewers and editors must disclose any association that poses a conflict of interest in connection with the manuscript. Recent collaborators, defined as people who have coauthored a paper or were a principal investigator on a grant with any of the authors within the past 48 months, must be excluded as editors and reviewers. Referees and editors are asked to recuse themselves from handling a paper if the conflict makes them unable to make an impartial scientific judgment or evaluation. A referee or editor who has a conflict but believes that it does not preclude his or her making a proper judgment must disclose to the journal the nature of the conflict.

If you still want to review the article, I would contact the editor first, disclose the COI, and ask if they still want you to write a review. If they say no, then just let it go.

This is a conflict of interest. Basically, every policy I've found says that a referee has a conflict of interest if they've recently (for various definitions of "recent") co-authored a paper with one of the authors of the paper under review.

• Allure's answer says that PNAS considers collaboration within the last four years as a COI.

• This Elsevier policy says that being a co-author of one of the authors within the last three years is a COI.

• The IEEE Robotics and Automation Society says five years.

• This ACM policy says four years.

You should either decline to review the manuscript (especially if you're busy or otherwise looking for a reason not to review another paper right now) or advise the editor of your conflict of interest and ask if they still want you to review the paper. In a small field, it may be that nearly all potential reviewers have a conflict, so the editor might still want you to review.