1.22.0 (2026-10-11)
Features
- analyze: per-facet analysis status and targeted runs (#1840) (7dbc57c)
- analyzer: per-facet requests and a file AudioSource in the worker (#1841) (5d06d37)
Bug Fixes
- analyzer: recycle idle lean worker (#1874) (9132a32)
- app: publish listener track progress to media sessions (#1880) (77377f4), closes #857
- broadcast: let listeners share a username (#1877) (13271e2)
- controller: measure cloud mp3 clips instead of guessing (#1871) (3d335f7)
- dj: ground spoken place names in on-air location (#1885) (d95deee)
- docker: support Blackwell GPUs with cu128 analyzer wheels (#1878) (7334185), closes #1340 #1738
- doctor: make markdown code legible and clarify settings (#1883) (9d1e568), closes #1704
- jingles: pin uploaded audio to mixer sample rate (#1882) (1350cd4), closes #1688
- loudness: no boost for a track whose peak is unknown (#1844) (adcb609)
- mcp: publish tool output schemas without a $schema dialect (#1838) (992c98b)
- picker: preserve predecessor flow context in agent re-picks (#1884) (b3e3957), closes #1820
- selection: artist spacing sees an act inside a multi-artist credit (#1847) (2f804b6)
- skills: subwave-worktree-dev reads the root .env, not controller/.env (#1845) (b934c39)
- tools: audiomuse-import writes genres, not the generated genre column (#1879) (eed3613), closes #707