Project Status: Emphasis on Security and local LLM Performance

[jjuliano]

The current solution we have involves sending personal files which can also be sensitive data (i.e. bank account, etc.) to OpenAI, which is obviously not ideal.

I am currently on the look for a better solution, observing the current trends, and seeing what solution does see fit. Please suggest new approaches and other solutions as well.

With the new approach, I am currently thinking of having an option to use open-source LLM models locally, or via on-prem, or cloud. Also, with this new approach, we will go all-in LLMs, and I will trust the LLMs to do most of the low-level stuff including media-type detection, etc.

Also, the project should be deployable to any personal computers, servers, or companies, in an easy manner, whichever needed this solution. This also highlights the emphasis of easy UI/UX, which can be either through a passive daemon service or a well designed UX and GUI.

And lastly, it's better to use this solution in *nix first fashion, which can either be respecting XDG in Linux, or natively compatible with the macOS Darwin ecosystem. And ensure that it would work seamlessly on Windows. This approach requires moving away from NPM, and having the binaries downloadable. Since, I am also a Golang developer, using Go seems to be the next feasible solution to solve this going forward.

Thank you for the interest of the project.