MuhammadUsmanGM/autonoma — reverse-engineered prompt

Build me an AI assistant platform that acts like a digital employee I can run locally. I want to connect it to chat channels like Telegram, Discord, WhatsApp, Gmail, a simple REST chat endpoint, and the terminal, so people can talk to the same agent from different places and it remembers the context.

It should use an LLM provider like OpenRouter or Anthropic, keep useful memories in a local SQLite database, search those memories intelligently, and avoid saving duplicates. Give it safe built in tools for web search, reading and writing files, and running approved shell commands in a protected workspace.

Also build a polished web dashboard where I can watch live activity, see errors and alerts, inspect memories, review past sessions, edit the agent identity, and switch between light and dark mode. Include health checks, metrics, logs, and simple setup with environment variables. Make it easy to run from source, with sensible defaults and example config.