eliwoodward/server_helper — reverse-engineered prompt

Build me a single file Python command line tool for Linux that lets me talk to an AI model and have it actually work on the server for me instead of making me copy paste commands.

When I run it, it should scan the machine once and save a useful system profile with things like OS, CPU, memory, services, open ports, Docker, firewall, users, logs, and important directories. Then it should let me choose between Anthropic Claude and OpenAI models, ask for an API key if needed, store keys encrypted locally, and test that the connection works.

After that I want an interactive chat where the AI can run shell commands, read files, write files, and search the filesystem until it finishes the task. Include simple chat commands like switching models, clearing history, showing cost or token usage, updating the system profile, updating keys, help, and quit.

Please keep it lightweight, no framework, just one Python file plus a requirements file if needed. It should work on common Linux servers with Python 3.10 or newer.