Allow use cache; without forcing cacheComponents?

[jacobrobertsbaca]

Hi there! I came across the new use cache feature and was really excited by it. I tried to incorporate as a lightweight way to cache a function call that fetches a large amount of data from a database by marking a server-side function with a use cache; directive. I was a bit perplexed to then see many Uncached data was accessed outside of <Suspense> errors show up across my site and cause errors during build.

After doing some digging and learning more about use cache and PPR, I understand the motivation behind requiring <Suspense /> for dynamic data access, but I can't help but feel that this is a bit of a strong requirement, especially for those who might want to incrementally migrate to this approach. For larger sites like mine which extensively use dynamic rendering, it feels like a tall order to migrate many different parts of the site to this model.

Is there a way to keep dynamically rendering pages in my site, but occasionally cache expensive functions using use cache on a case-by-case basis? Perhaps I'm thinking about this incorrectly, but nothing about the use cache; idiom in isolation seems to imply that uncached, dynamically rendered components should be disallowed. Has there been any thought about having a more flexible option, perhaps e.g. an option to disable the above error messages or to have a useCache config option that allows a separate, less strict version of cacheComponents?

[engripaye]

---

Tested & Trusted

Hi there! I’ve been exploring the new use cache feature and was excited to try it as a lightweight way to cache a server function that fetches a large dataset from my database. However, after marking the function with a use cache; directive, I started seeing many “Uncached data was accessed outside of <Suspense>” errors across my site, which also caused build failures.

After digging deeper into how use cache interacts with PPR and the requirement for <Suspense> boundaries around dynamic data, I now understand the reasoning behind the strictness. Still, it feels like a steep requirement for larger apps that rely heavily on dynamic rendering. Incrementally migrating an entire codebase to wrap every dynamic piece in <Suspense> is quite challenging.

My question is: Is it possible to continue dynamically rendering most pages, but selectively cache expensive functions using use cache on a case-by-case basis? The use cache; directive itself doesn’t suggest that all uncached dynamic data access should become invalid. Has there been any consideration for a more flexible mode — for example:

• a way to disable or relax these error messages, or
• a useCache configuration that allows a less strict version of the cache model?

I may be thinking about this the wrong way, so any clarification on intended usage patterns or upcoming flexibility would be really appreciated.

---