Rasa X deployment on Windows Server

Hello Rasa family,

I have started my journey with Rasa a couple of months back. Installed it on Windows 10 machine and it worked perfectly fine. Developed the bot according to my use case, exposed the API to be used in an application, tested it and Yay! it worked glitch-less!

Convinced my team, Rasa is the best and we are going ahead with this!

Now when i tried moving it to test server(Windows Server 2012 R2), trouble started!

  1. Tensorflow DLL not found issue - solved this with a workaround
  2. sqlalchemy issue/deployment issue - which needs Multipass , that in-turn needs Hyper-V, which is compatible with Windows 10 only. :frowning: Tried installing Oracle VirtualBox - still no luck!

So my questions are,

  1. What is the best way to deploy in Windows Server? I would like to go with One line deployment- Is it feasible?
  2. Will i encounter anymore hurdles if i go ahead with Windows? Should i just switch to CentOS instead?

Please please help!

@JiteshGaikwad @Juste @akelad @mukherjeesoumya @rctatman

Hi @rashmi.metri, if it’s an option, I would suggest using a Linux based server (ideally Ubunutu or Debian). That will allow you to easily use the one-line deployment script.

If Windows is a must, there are some Windows specific instructions here. Alternatively, the docker-compose installation may be easier on Windows, see this thread: [TOOL/GUIDE] Unofficial RasaX Quick Install script for a pure Windows machine

Thanks for confirming :slight_smile: This is exactly what i suggested to my team.

Faced quite some issues with Windows. Will go ahead with Linux (AVX instructions enabled)

Thanks again :slight_smile: