I also had several issues in getting the environment set up in my windows machine whereas Ubuntu was done with very minimum issues. A few suggestions based on my experience:
- Use a separate conda environment (
$ conda create --name bots_py36 python=3.6
) - Download the requirements.txt from GitHub (
https://github.com/RasaHQ/starter-pack-rasa-stack/blob/master/requirements.txt
) - Try
pip install -r requirements.txt
- If any of the packages failed to install, try them individually (I had to do this as the
Twisted
was failing several times) - If
pip
fails for an individual package, tryconda install
(I did this for Twisted) - Get
spacy
installed usingpip
orconda
(There is a weird minimum version requirement conflict withnumpy
. I currently have 1.14.5) - Optional - Download English lang from spacy using
python -m spacy download en
- I also faced issues with C++ libraries (
cl.exe
). However, after updating the Visual Studio installation in my machine to support C++, it was ok - Tip - Several times, I used
git bash
to execute commands rather than CMD when faced with issues. And, for some reason, I got it working.