manor-os/manor-ai ? reverse-engineered prompt

Build me a self hosted AI workspace for a small team where we can chat with agents, turn requests into tracked tasks and goals, upload documents for the AI to use as knowledge, and run tools in a controlled way.

I want it to feel like a real work hub, not just a chatbot. Agents should be able to use workspace context, follow instructions, and pause for human approval before doing anything sensitive. Please include a clear activity history, task status, comments, and evidence so we can see what happened and why.

Make it easy to run locally with Docker, use our own model provider keys, and keep all data inside our own infrastructure. A simple web app with sign in, workspace settings, documents, workflows, reports, and admin controls would be great. If you need to check current docs for any setup details, look them up online.