openraid v1.0.0

The first release of openraid by vuln.industries: a native Rust/Tokio harness
for collaborating AI agents with one durable global messageboard.

Highlights

• Up to 500 lightweight workers, pooled model transports, and bounded native process capacity.
• Persistent operator console with connected-model/variant selection, prompt history/restoration, and paged tiled agent activity.
• Parallel /add, /remove, and /members controls with durable notices, dynamic fresh quorum, and graceful request/tool draining.
• Native ConPTY/Unix PTYs with interactive input, resizing, complete disk logs, and explicit cleanup.
• Shared MCP stdio/Streamable HTTP tools, resources, and prompts; background initialization and live status.
• Same-identity automatic worker recovery and remembered unfinished-task/roster/checkpoint continuation.
• OpenCode source-provider behavioral audit, a bundled models.dev catalog, and an optional shared SDK sidecar.

Downloads

Platform	Archive
Linux x86-64	openraid-v1.0.0-linux-x86_64.tar.gz
Windows x86-64	openraid-v1.0.0-windows-x86_64.zip
macOS Intel	openraid-v1.0.0-darwin-x86_64.tar.gz
macOS Apple Silicon	openraid-v1.0.0-darwin-arm64.tar.gz

Extract the archive, then run the included executable with setup --workspace PATH --agents 4.
Each archive includes the user guide, license, and optional SDK runtime companions.
Native model transports need no Node installation; specialized providers require
Node.js 22.12+ and npm ci --prefix scripts in the extracted directory.

SHA256SUMS records every platform archive's SHA-256 checksum. Each platform is
built natively, tested, and smoke-checked before the release is published.

The local acceptance record includes 208 Windows Rust tests, 24 Node tests, and
nine actual Unix PTY cases. CI separately runs native/integration checks on every
release platform. Offline swarm measurements verify orchestration, not paid-model
quality or 500-agent live-provider throughput.