If you run multiple AI agents on the same repo, how do you stop them stepping on each other?
热度趋势
趋势数据积累中
百分比基于当前可用热度信号，而非评论数或独立用户人数。
一位开发者利用多个AI代理在同一个代码库上并行开发2到3个功能，并通过git worktrees来隔离代码。他们想知道其他开发者是否也并行运行AI代理，或者为了避免混乱而选择顺序开发。该开发者正在寻求关于如何在单个代码库上并发操作多个AI代理时，防止代码和数据冲突以及确保正确测试的建议，并质疑自己的方法是否过于复杂。
I use AI to develop, and because it works quite well, I extended it to run multiple agents in parallel. This way I can develop 2 or 3 features at the same time. It works well, and git worktrees do the job of separating the code.
The problem comes when I try to test, or when I ask them to test against a running app. They can't all use the same db and running instance. I still have to do that part manually, one by one, which is annoying and slows the whole thing down.
Curious whether others actually run agents in parallel on one repo, or just do them one at a time to avoid the mess. And if you do run them in parallel, how are you keeping both the code and the data from colliding, and how are you testing them properly? Or is this just me overcomplicating it?