Ouroboros: Champion Council - AI Model Orchestrator

A local-first VS Code extension for AI development. Orchestrate models, automate complex workflows, rent GPUs, and tap into a decentralized community, all from your editor.

Comment

Inspiration

Ouroboros, also known as Champion Council, was inspired by a year of research into distribution governance, rooted in Stephen Cole Kleene's foundational work on recursion and fixed-point systems. The goal was to translate these powerful, theoretical concepts in a practical, tangible tool for the modern AI developer. How could we create a system that is self-contained, yet infinitely extensible? A tool that is powerful, yet runs entirely on your local machine? This project is the answer.

What it does

Champion Council is a local-first AI orchestration platform built as a VS Code extension. It turns your editor into a command center for AI development, providing: A Council of Models: Load and manage multiple AI models (from HuggingFace or elsewhere) in a "pluggable" slot-based system A Vast Toolset: Access over 140 tools for the entire AI lifecycle, including inference, batch operations, memory management (FelixBag), and advanced diagnostics. Seamless GPU Scaling: When your local hardware isn't enough, rent and manage on-demand GPU power from Vast.ai directly within the editor. Advanced Observability (CASCADE): A unique, built-in system for tracking the provenance and history of AI operations, ensuring reproducibility. Decentralized Community Hub: Using the Nostr protocol, you can share reusable AI workflows, build a reputation, and collaborate with a community of builders.

How we built it

The project has a dual-component architecture:

  1. The VS Code Extension (TypeScript/Node.js): This is the user-facing part, comprising the command palette integrations, status bar indicators, and the main webview panel. It acts as the primary orchestrator.
  2. The MCP Server (Python): A powerful Python backend, packaged as a self-contained "capsule". The extension communicates within it in real-time using a robust Server-Sent Events (SSE) and JSON-RPC bridge. This server exposes the entire 140+ toolset without requiring the user to manage complex Python environments. This design allows us to deliver a massive amount of functionality in a simple, easy-to-install VS Code extension, embodying the "fixed-point" inspiration—a complete system, contained within itself.

Challenges we ran into

Packaging Python: Bundling a complex Python environment (with ML libraries) into a seamless, zero-config component of a VS Code extension was a significant challenge. We solved this by creating a compressed "capsule" that is extracted and run on first use. Real-time Communication: Designing a fast and reliable two-way communication bridge between the Node.js extension process and the Python backend was critical. Our SSE-based protocol proved to be both performant and resilient. Scope & Complexity: Managing the sheer scope of over 140 tools required a disciplined approach to architecture, ensuring the system remained coherent and maintainable.

What's next for Ouroboros

We are just scratching the surface. Future plans include expanding the workflow-sharing capabilities, adding more community and governance features, and continuing to explore the practical applications of recursive systems in software distribution.

Built With

Share this project:

Updates