jqi-1/CapsuleLab — reverse-engineered prompt

Build me CapsuleLab, a local first tool that lets AI and data science projects run the same way on any machine.

I want a simple command line flow where I can create a project from a template, check if my machine is ready, build the container, start it, open apps like JupyterLab, Streamlit, Gradio, TensorBoard, or VS Code Server, and see logs. It should handle Docker, GPU checks, ports, volumes, datasets, secrets, Git, and experiment runs without making me remember a bunch of setup steps.

Also include an API and a web dashboard so I can see projects, runtimes, apps, resources, templates, profiles, runs, and remote machines in a friendly UI. Support running on remote servers over SSH, syncing files, and optionally using a small remote agent for Docker management.

Use Python for the CLI and backend, and a React style frontend. Include project templates for basic Python, PyTorch CUDA, Streamlit, RAG research, FastAPI deployment, and open source packages. Look up current docs online if needed.