Skip to content

feat: add Windows mlx-cpu runtime support - #2298

Open
hvkeyn wants to merge 3 commits into
exo-explore:mainfrom
hvkeyn:main
Open

hvkeyn wants to merge 3 commits into
exo-explore:mainfrom
hvkeyn:main

Conversation

@hvkeyn

@hvkeyn hvkeyn commented Sep 8, 2026

Copy link
Copy Markdown

Summary

  • Enable win32 in the uv resolver and install the mlx-cpu extra so exo can start on Windows.
  • Replace Unix-only pidfile, daemon, rlimit, and pipe-fd APIs with Windows equivalents; make discovery Send so exo-rs builds on MSVC.

Test plan

  • uv sync --python 3.13 --extra mlx-cpu on Windows
  • uv run --frozen exo starts and serves http://localhost:52415
  • Existing macOS/Linux uv sync --extra mlx / mlx-cpu still resolve

Enable win32 in the resolver, replace Unix-only pidfile/daemon/rlimit/pipe APIs, and make discovery Send so exo-rs builds on MSVC.
Store state under LOCALAPPDATA, report nvidia-smi GPUs, expose /api/* Ollama aliases, and add Windows run/log/wipe/firewall scripts.
Restore manual bootstrap peers, listen on IPv4 as well as IPv6, and reject Wi-Fi-only hops for tensor sharding so LAN clusters can form on Windows.
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant