Consilium multi-LLM collaboration platform with MCP server and Gradio UI

AI Impact Summary

Consilium introduces a structured multi-LLM collaboration framework that uses distinct roles and rounds to reach consensus, effectively turning ensemble reasoning into an executable workflow. It wires a Gradio-based visual roundtable to an MCP server for programmatic integration, and includes a research agent that uses function calls to fetch data from Web Search, Wikipedia, arXiv, GitHub, and SEC EDGAR. The deployed model set spans Mistral Large, DeepSeek-R1, Meta-Llama-3.3-70B, and QwQ-32B, with experimental hooks to Claude Desktop (Cline), Claude Sonnet, and OpenAI o3; MAI-DxO is referenced as a benchmark for multi-LLM panels. This approach will increase compute usage and latency per decision but can improve accuracy on complex, multi-faceted tasks.