Rate Limit in Next Middleware or TRPC Middleware
What do you think is the best way to rate limit your app with t3 stack.
To me rate limit in next middleware seems much more efficient and logic, since you are on the edge and all of your app traffic is going trough there (static, pages, api, etc..)
But, on the other hands, rate limiting your API with different procedures, for different scope comes with a cleaner organization imo.
Since T3 stack can't run on the edge (unless you don't use prisma or some routes at least)
You are forced to rate limit your API in serverless function.
Maybe I'm getting something wrong, but for me rate limit in next middleware seems better since you can can rate limit every type of routes on the edges (would be insane if you can't pass next-auth context there too 🙏 ).
Is it really something worth the time to consider or is it useless and I should go whatever I want to?
2 Replies
We used this along with a redis instance:
https://github.com/animir/node-rate-limiter-flexible
https://github.com/civitai/civitai/blob/main/src/server/utils/rate-limiting.ts
Which can then be used like:
https://github.com/civitai/civitai/blob/main/src/pages/api/download/models/%5BmodelVersionId%5D.ts
GitHub
GitHub - animir/node-rate-limiter-flexible: Count and limit request...
Count and limit requests by key with atomic increments in single process or distributed environment. - GitHub - animir/node-rate-limiter-flexible: Count and limit requests by key with atomic increm...
GitHub
civitai/rate-limiting.ts at main · civitai/civitai
A repository of models, textual inversions, and more - civitai/rate-limiting.ts at main · civitai/civitai
GitHub
civitai/[modelVersionId].ts at main · civitai/civitai
A repository of models, textual inversions, and more - civitai/[modelVersionId].ts at main · civitai/civitai
Thanks for the repo but the question wasn't about the tools but more about the best approch to implement a rate limit as a whole in this t3 stack