VoltAgent/voltagent ? reverse-engineered prompt

Build me a full-stack AI agent engineering platform using TypeScript. It should consist of two main parts: an open-source framework for defining the agents and a web-based console for managing them.

The framework part should allow me to create sophisticated AI agents that can remember conversations, use external tools, and even have voice capabilities. I'd like to be able to build multi-agent systems where specialized agents can collaborate on tasks under a supervisor. It's important that they can pull information from my own documents (RAG) and that I can design multi-step workflows, some with human approval steps. Please make it easy to swap between different LLM providers like OpenAI or Anthropic.

The second part is the web console. This should be a dashboard where I can monitor all my agents, see their logs and performance, manage prompts, and test them out in a chat window. It's the central hub for observability, deployment, and making sure the agents stay on track. You can look at modern agent engineering platforms online for inspiration.