apocas/restai — reverse-engineered prompt

Build me a self hosted AI service platform that I can run locally or deploy with Docker. I want an admin web dashboard where I can create separate AI projects, choose different AI providers like OpenAI, Anthropic, Gemini, Ollama or local models, set prompts and tools, then test everything in a chat playground and expose it through a simple REST API.

It should support document based chat with uploads, searchable knowledge, optional knowledge graph views, and natural language questions over connected databases. Add agents that can use tools, browser automation, and MCP connections. Include a visual block style builder for workflows, plus prompt versions, team access, login security, rate limits, token tracking, cost tracking, latency charts, and project analytics.

Please make it production friendly but easy to start, with sensible defaults, migrations, an included frontend, and clear setup commands. Look up current docs online if you need to.