Sequence of events:

	1. Started with installing llama.cpp
	2. Then found, it is not capable of web search
	3. Then I wrote the python script and the small UI
	4. Then installed searxng and pointed the python script for tool call
	5. Then found it can't analyze image
	6. Tried a few models - Qwen3.5 9B, Qwen3 VL 4B, Gemma4 E4B. Finally stuck with Gemma 4 E2B
	7. Then tried image generation. Does not work out of the box
	8. Then configured ComfyUI
	9. Tried different models like Realistic, SD 1.5, SD 3.5 and finally stuck with Z-Image
	10. Then tried coding construct, it was easier than I thought
	11. Tried continue.dev but did not work and then finally stuck to opencode with llama.cpp (~/.config/opencode)

The setup is slow but it works
Won't recommend as a local coding assistance
All models were taken q4km