LLM platform lock-in is a pain. Here's how I built a custom registry for a truly flexible, multi-model AI agent.
Your architecture might be why the agent is ignoring you. Let's talk about it.
What could possibly go wrong if I let my agent just... be? I had to see. Duh. Told I went to sleep and left the AI alone.
What started as a personal jab at a $5 app unexpectedly became a thriving open-source community.
She read the bug, explored the codebase and shipped the fix.
An end-to-end AI system designed to converse in real time, manage its own memory, and stay fully under my control.
