A few observations:
1. Telemetry is enabled by default, and may contain the API and chat queries. I've logged an issue for this along with some suggestions here: https://github.com/khoj-ai/khoj/issues/389
2. It would be advantageous to have configuration in the UI rather than baking it's YAML into the container image. (added a note on that in the aforementioned issue on Github).
3. It's not clear if you can bring your own models, e.g. can I configure a model from huggingface/gpt4all? if so, will it be automatically downloaded based on the name or should I put the .bin (and yaml?) in a volume somewhere?
4. AMD GPU/APU acceleration (CLBLAS) would be really nice, I've logged an issue for this feature request as well. https://github.com/khoj-ai/khoj/issues/390