Skip to content

v0.9.43

Choose a tag to compare

View all tags
@github-actions github-actions released this 14 Jul 09:00
· 63 commits to main since this release
v0.9.43
6642b61

🛟 Office-wide fallback brain (opt-in)

When a teammate's model provider goes sustainedly overloaded — repeated 5xx (GLM/Z.AI's 529 under load is the classic) — the office can now re-run that same task on a fallback brain you chose, instead of leaving it to die on the retry loop.

  • Set it once in Settings → CONNECT → 🛟 สมองสำรอง — pick any connected provider + an optional model. Applies office-wide.
  • Off by default. With no fallback set, nothing changes: an overloaded brain just retries hard, exactly as before.
  • Fires only after the overload is sustained (not a one-off blip), only onto a provider that's actually connected, and never loops back onto the down brain or fails over twice for one task.
  • Only server overload/unavailability (5xx) triggers it. Bad auth (401/403) still fast-fails; rate/usage limits (429) still pause-and-resume — none of those switch brains.

The failed-over run keeps the original task and still reports back to whoever delegated it, with a one-line note that it switched brains and why. This is the careful version of the auto-failover reverted in v0.9.40.

Known issue (still under investigation)

The macOS "stuck on the boot logo" render bug from v0.9.42 is not fixed here — waiting on boot-log output from an affected Mac.

Update via the in-app 🔄 banner or bagidea update.

🤖 Generated with Claude Code

Assets 7
Loading