video

Readme

Add error handling and fallback device mapping for model loading

Fix transformers version conflict with colpali-engine

Fix model loading issues - pin transformers/torch versions, add trust_remote_code

Remove comments and clean up chat-with-pdf app

Add all popular open-source models to ALL_MODELS for Spaces

Remove main.py, update pyproject.toml and uv.lock for sync with Space

Agentic Demo: LLM Debate & Judge (Track 3), README tag, video placeholder

Switch to pure Gradio MCP server for Hugging Face Spaces (no Modal)

Rewrite: pure Gradio MCP server, no Modal, mcp_server=True, ready for Spaces

Add app.py for Hugging Face Space deployment (Decider MCP)

Decider MCP: user-selectable models, minimal, Gradio UI, docs updated

Minimal MCP server: 2 LLMs, Gradio UI, B200-ready, hackathon submission

docs: clarify project as Modal MCP Server with Gradio UI and B200 support

docs: add project description to README

docs: add submission guidelines and award note to README

fix: use HUGGINGFACE_API_KEY env var

feat: implement expert consensus MCP server

changes

Initial LLM swarm implementation

Optimize for B200 FP4 and parallel inference

current

Modal LLM ensemble MCP: 4x 70B models, consensus, FastAPI endpoint