currently a pretty barebones mvp. works with a webhook on port 5000 with a function to transcribe all memos which meet requirements if you want. probably a lot of bugs and typos

adds transcripts like this:

1. clone and cd into repo
2. cp .env.example .env
3. edit variables in .env
4. docker compose up -d (maybe docker-compose up -d depending on distro (debian))
5. add webhook

1. fill all the variables in the code/env, modify the filters and stuff
2. get all requirements (ive only tried the nix stuff)
3. run python main.py
4. add the webhook to your memos instance

• run do_all() if u want, this is still pretty rough. there will be a better way eventually

be able to see what i log without having to download the whole video/audio file which can be several gigabytes. also searchability

• figure out a better name for this (super important, obviously)
• make the code not a disaster
  • env vars instead of hardcoded
  • figure out unloading the model in downtime
  • package properly
  • error handling
  • more robust filtering of which memos should be transcribed
  • figure out where to store model in docker image
• docker image
• barebones webui
• add option for external transcription