What's new
- Thinking disabled by default —
--reasoning-budget 0 on all llama-servers. Same answer quality, 2x faster for agentic workloads. API users can still opt in per-request with chat_template_kwargs: {"enable_thinking": true}.
- Blackboard MCP server —
mesh-llm blackboard --mcp exposes blackboard as MCP tools (blackboard_post, blackboard_search, blackboard_feed) for agent collaboration across the mesh.
- Web UI improvements — no hidden thinking in chat responses, automatic retry with backoff on 500/502/503 errors.
- Fly community banner — hosted instances (mesh-llm.com, anarchai.org) show a 'community demo' notice with live busy indicator when the mesh is under load.
- Reasoning compatibility test —
tests/test_reasoning_compat.sh validates the thinking on/off API contract.
Install (macOS Apple Silicon)
curl -fsSL https://github.com/michaelneale/decentralized-inference/releases/latest/download/mesh-llm-aarch64-apple-darwin.tar.gz | tar xz && sudo mv mesh-bundle/* /usr/local/bin/
