Dependency issues when importing haystack

[falconair]

A few weeks ago I was able to pip install haystack, as suggested by the docs. I ran a couple of basic tutorials. Starting yesterday I've been having issues. Please note that I completely uninstalled all haystack packages and attempted to pip install everything again.

When I do form haystack.nodes import BM25Retriever, I get an a no module error for cchardet. I manually installed it via pip install cchardet, which made that error disappear, but now I'm getting the following error:
AttributeError: module 'faiss' has no attribute 'swigfaiss'

Google tells me this is because I may have multiple faiss-cpu packages, one via pip and another via conda.

I needed to use the faiss library for another project so I installed it via their recommended method: conda. That seems to have broken the haystack install! Apparently the conda repo is maintained by Meta and the pip one is not.

This is a mess.

Why wouldn't a pip install of haystack include cchardet and the right version of faiss? Am I missing something or is this a known problem? I didn't post this as a bug because I may just be missing something on my end.

I'd like to avoid having to create different conda environments for haystack and other NLP tools. Even if I broke my original haystack installation with subsequent installations of faiss, langchain, llama-index, etc., the fact that I pip removed and re-installed haystack should fix my installation, right?

[falconair]

As an update, I completely removed faiss and haystack packages, then re-installed haystack and let haystack manage the faiss dependency. That seems to have resolved the issue.