2

LUIS gives 403 response with "Out of call volume quota." message even if I haven't reached 10,000 transactions free per month limit. My current usage is 1431 hits. I would also like to clarify that we are not hitting more than 5 calls a second. It's hardly one call a second.

Anand Padiya
  • 143
  • 8

1 Answers1

0

Are you entirely certain that you are using the correct key, not the starter one? The starter one has a monthly limit of 1000 queries per endpoint per month, which possibly now would be exceeded with 1400+ queries.

See https://learn.microsoft.com/en-us/azure/cognitive-services/LUIS/luis-boundaries#key-limits

Authoring/Starter | 1 million/month | 5/second | 1 thousand/month | 5/second | Authoring your LUIS app

Versus the probably intended

Subscription | F0 | Free tier invalid | 10 thousand/month| 5/second | Querying your LUIS endpoint

Joeppie
  • 438
  • 3
  • 16
  • 1
    Thanks a lot. I had the same issue. It would have been great if those limits would have been mentioned here https://azure.microsoft.com/en-us/pricing/details/cognitive-services/language-understanding-intelligent-services/ – Anand Padiya Jul 19 '18 at 11:48
  • You might want to mark the answer as an answer if it did solve or answer your problem. Also, I did not run into this specific problem, I just Googled for the API limits for LUIS, since Azure is sometimes a bit tricky with those, and came across the page with the more specific information, I never actually used LUIS :) – Joeppie Jul 19 '18 at 13:18
  • @Joeppie I am testing within the Luis.ai app, which has an option for test. However, I run into this error and I used endpoint api, with the same authoring resource key, and it still worked. Is there a way to connect the prediction resource to the "test button" we have within Luis.ai? Any clue? – Lakshmi Narayanan May 28 '22 at 08:54
  • 1
    @LakshmiNarayanan , I know there is typically a disconnect between the 'normal' interface and the actual quota. Reason for this is that it is technically a lot more feasible/cheaper to not 'track' the precise number of requests and making it available in an interface (which quickly goes out of date). My go-to solution would be to ensure that you know how many requests you are firing towards the API and take the appropriate action to throttle requests or gain insight toward how your usage is. (i.e. azure management/reporting portals) – Joeppie May 30 '22 at 13:50