QuantumChrono/project-trinity ? reverse-engineered prompt

Build me a private AI workspace that can run inside a Kaggle notebook and feel like my own little JARVIS.

I want a simple way to boot either a fast chat mode or a deep reasoning mode, use a 40B model when I need stronger answers, and switch to smaller fast models for everyday chat and vision. It should also have image generation, web search, basic RAG over my own files, and voice input and output, all connected through one web interface I can open remotely.

Please make it work with Kaggle’s free GPU limits, handle the awkward disk and memory restrictions, and keep the setup reliable if the notebook restarts. If you need to check current docs or the best way to wire something together, look it up online. I care most that it’s private, usable, and feels like one unified assistant instead of a bunch of separate tools.