v1.0.0 v1.832.0
parruda/swarmSep 6, 2026 → Oct 13, 202637 days between releases
What's New
Newer release adds Python 3.13 tracking and S3 proxy security headers; the 1.0.0→1.832.0 jump is largely undocumented.
Breaking Changes
- No breaking changes are listed in the newer notes, but the 831-version gap may contain undocumented breaking changes.
New Features
- Record each instance's Python version for tracking.
- New instances start on Python 3.13 by default.
- Variable value type support is widened.
Improvements
- S3 proxy objects are served with nosniff and a sandbox CSP for security hardening.
- Python runtime version is now tracked per instance.
Bug Fixes
- Unavailable OAuth sign-in options are now badged in the resource form.
- Variable value widening no longer breaks live servers' cached queries.
- S3 proxy responses include security headers to mitigate content sniffing.
Upgrade Notes
The two release notes reference different upstream repositories (hindsight vs windmill), so verify you are comparing the correct artifacts. The 1.0.0→1.832.0 jump is massive and intermediate changes are undocumented; review all releases in between. New instances default to Python 3.13, so confirm runtime compatibility.
The Two Releases
v1.0.0
Sep 6, 2026· 2 months ago
The first major release of the speech-to-speech project consolidates the runtime around a single Realtime engine and replaces implicit mode flags with three explicit commands: serve, talk, and local. It adds several new speech recognition and synthesis backends, strengthens realtime conversation lifecycle handling (ordering, interruptions, cancellation, cleanup), and refreshes the Apple Silicon/MLX setup. Setup guidance and diagnostics are also improved, with transcript logging now opt-in for privacy.
Full release notesv1.832.0
Oct 13, 2026· 10 days ago
Introduces Python version tracking and improves S3 proxy security by adding CSP and nosniff headers.
Full release notes