Infrastructure to deploy on production


I would like to know what extra tools do you guys recommend to use together with rasa in order to support more than 200,000 requests per month.

I have just developed some projects combining rasa + flask + ngrok, however it seems not ideal to a large project.

I heard about Amazon Elastic Beanstalk. Can you guys share some combinations to make Rasa scalable, and mention some bottlenecks that I should worry about ?

Best Regards

1 Like

Hi @kaleming, have you taken a look at Rasa X yet? There’s a bunch of deployment options in there that are production ready. For a scalable deployment, I would probably suggest the Kubernetes deployment option. That will handle 200,000 requests per month no problem, and you can always scale it if you get more requests.

I have successfully deployed Rasa X to Kubernetes and am running into scalability issues. The bot works fine if we have a couple of users, but when we have more than 10-20 users the bot really struggles with 20s+ response times. I have checked the pods and cpu utilization and cannot pinpoint the bottleneck. Also the only concerns I have seen online were regarding the tracker store, however I assumed that the Kubernetes set up was going to be scalable out of the box. BTW I am using the AWS EKS set up. Any ideas would be much appreciated.

1 Like

Hey @pavel.hov2 are you still having this issue?

Hi, we would like to know more what is the cause of this as we are building in the same architecture by using EKS.