Summary

Hi everyone,

I’m exploring ways to speed up CI runs by leveraging Turborepo’s caching while running tasks in parallel. I have a scenario where I’d like to ignore the task dependency graph for execution purposes, but still want the dependency relationships to be taken into account when calculating cache keys. This way, even though tasks run concurrently, they can still benefit from cache invalidation based on the full dependency tree.

Current Setup

In my turbo.json, I have a task defined like this:

{
  "typecheck": {
    "dependsOn": ["^typecheck"],
    "outputs": []
  }
}

When I run the following command:

npx turbo run eslint --parallel --dry

I noticed that the Dependencies and Dependents fields are empty for each task. Essentially, using the --parallel flag seems to ignore the dependency graph when it comes to scheduling the task execution.

What I’m Trying to Achieve

My goal is to:

• Use the dependency graph for cache key calculation and for determining which packages to run.
  (This ensures that even if a task outputs nothing, its cache key reflects changes in its dependencies.)
• Run the tasks in parallel, ignoring the dependency order.
  (Since tasks like ESLint or typecheck don’t produce output files and I don’t need to wait on dependencies finishing, running them all at once could improve CI resource utilization and reduce overall run time.)

In short, I’d like a configuration or flag that tells Turborepo:

“Calculate cache keys and determine affected packages based on dependencies, but when executing, run everything in parallel.”

Discussion Points

• Is there an existing way (flag or configuration option) to achieve this?
  Perhaps a hidden flag or experimental feature?
• If not, would this be a reasonable feature request?
  I believe many projects could benefit from faster parallel execution on CI where tasks are essentially “fire-and-forget.”
• Workarounds:
  Has anyone found a creative solution—maybe using a combination of task definitions and command-line flags—that mimics this behavior?

Any insights or suggestions would be greatly appreciated. Thanks for your help!

Additional information

No response

Example

No response