What pi.dev plugin do you suggest for context, compaction and memory management of local models?

Excerpt

I have been battling with my Qwen3.8:27b setup on my rtx 5080 16gb. I am using llama.cpp to run a nvfp4 version of qwen3.8:27b llama-b10699-bin-win-cuda-13.3-x64\llama-server.exe -hf williamliao/Qwen3.8-27B-NVFP4-GGUF:NVFP4 --jinja --chat-template-file d:\qwen\chat_template.jinja --reasoning-format deepseek --host 127.0.0.1 --offline -np 1 --threads 16 --threads-batch 16 --log-verbosity 4 --no-log-prefix --no-log-timestamps --no-mmproj-offload --mmproj D:\OllamaModels\blobs\sha256-ac3714bfdddeca