Description & Goals

In the last Hackweek 2024 project I was playing around with optimization of the image size of SLE Micro 6.0 via custom builds in OBS by limiting the target use-case to Raspberry Pis home IoT. This was image built-time optimization.

This year I would do something similar but during run-time (on the fly) without any prior knowledge. Kind of self-modify itself once deployed. The motivation still is to move from generic distribution to tailored distribution for a given use-case.

So the idea basically is:

1. Start from generic openSUSE Leap 16 or similar running on some real hardware (e.g. some The goal limit the target use-case to simple RPi (4, 400, Pi Zero 2)
2. Boot it and gather all available information about the platform (current kernel configuration in use, which firmwares are loaded, available pci, usb and block devices, list of available kernel modules etc.)
3. Use LLM to adjust the kernel configuration and then re-build the kernel with the new config
4. Re-boot and then remove the unneeded files resulting in tailored distribution for the target hardware it runs on
5. Evaluate if such optimization is useful for anything

The above steps shall be ideally done in AI agentic way using Gemini CLI and local MCP server.

Resources

RPi (4, 400 or Zero 2) or some other custom hardware