Dylan Freedman

Looking into Apache Tika for this via [tika-python](https://github.com/chrismattmann/tika-python). It does require Java to be installed but seems robust and permissively licensed. Open to another solution that has fewer dependencies, but...

I had this problem and figured out how to make something like this work with the compiler quirks. In essence, the compiler will only make whatever is immediately attached to...

Hi! Firstly: sorry I haven't been able to find time to get to your PR in FastFEC. (Though I have validated there is no perf difference in your version, I...

Not a bad idea! I think the specifying a directory and automatically recursively trawling it is a good idea too. Maybe v0.2 can have default directory as a configurable option...

I haven't seen that one. Are you still getting that error with the latest update (0.1.9)?

Hmm, strange it's working for me. What version of Semantra are you running with (`semantra --version`)? I'm on 0.1.7 (you can update with `pipx upgrade semantra`)

These are some good questions and definitely worthy of a write-up when I find the time. 1. The GPU will automatically be used if torch can find it. How that...

You could try cloning Semantra, editing the requirements to include the GPU source you need, and then `pip install -e .` to install Semantra locally from the repo. Then run...

Thanks for submitting the issue! I don't have Debian/Fedora so a bit hard for me to reproduce. Can you try swapping `annoy-fixed` for `annoy` in the requirements specified in `pyproject.toml`...

Yep, you're absolutely right. This should be granular on a per-file basis. I can look into auto-detecting encoding, but that might be time consuming for ever file, and it might...