vigil-llm
vigil-llm copied to clipboard
That PR I keep mentioning.
- Docker builds and runs with local storage (and the container is auto-built on push to the
mainbranch, and pushed to ghcr.io) - Config's all in a Pydantic model now.
- loader.py can do multiple loads in a row so it's not starting up multiple times
- SO much chasing types through flows
- moved testing to use pytest (which meant renaming vigil-server.py)
- ... probably other stuff I've forgotten
If there's anything you'd rather be done a particular way, don't hesitate, we all have our preferred ways of doing things 😄
@yaleman Would you be able to break this PR down into separate PRs that are more feature specific? This is really hard to review and merge as one PR since I have to take all or nothing from this branch
:)