muhharis99/odysseus — reverse-engineered prompt

Reverse engineered prompt

Build me a self hosted AI workspace that feels like having my own private ChatGPT or Claude on my computer, with a nice web app and mobile friendly design. I want chat with local models or API models, an agent that can use tools like web browsing, files, shell, memory, and reusable skills, plus a simple setup flow where I can start it with Docker, open it locally, and log in with the first admin account.

Please include the main parts shown in the README, chat, agents, model recommendations and downloads based on my hardware, deep research that gathers sources into a clean report, side by side model compare, an editor for documents, memory and skills, email triage, notes and tasks with reminders, and a local first calendar with sync and import export. It should be privacy focused, run on my own hardware, and be easy to configure in Settings for models, search, and email.

If anything is unclear, check the current docs online and make sensible choices.

Want more depth? Deep Reverse