fix(zerogpu): persist quota ledger so a restart stops claiming phantom minutes

fix(zerogpu): size GPU lease to remaining quota instead of a fixed 120s

fix(api): expose /v1/warmup on the served Gradio app

feat(models): add Qwen3.8-27B Turbo Coder to the ZeroGPU catalog

fix(api): harden empty-content handling for ZeroGPU chat

fix(api): recover Qwen tool calls on ZeroGPU

fix(cron): reserve ten ZeroGPU minutes for interactive use

test: send FLOW_API_KEY bearer in phonetic e2e client fixture

security+perf: enforce studio auth and cache singletons

fix(farm): derive transcript genre/persona from corpus category

feat(farm): feed harvested transcript content into live topics

fix(harvest): use YouTube captions instead of GPU-bound transcription

fix(studio): keep song structure when trimming, add remote MiniMax route

fix(pipeline): stop reporting success over broken harvest stages

feat(farm): wire harvested Moldovan content into song topics

Fix song farm duration: size lyrics to hit 2:30-3:30

Add Song Farm control center widget + remote MiniMax retry

chore(security): set visibility to private, sanitize internal Tailscale IPs, and drop qa_evidence tracking

fix(hub): add model alias resolution and persist canonical ibrowse qa evidence

feat(lyrics-baker): streaming SSE arena, Pydantic schemas, dictionary expansion

test: cron-level regression for quota stop saving resume state and aborting album

fix: classify ZeroGPU quota stops surfaced via Gradio queue output errors and async-jobs failed status

fix(cron): cap Albina album at 6 tracks (subset rotatie)

fix(cron): persist exact track order in resume state

albina bondar character + genre plan + lyrics diversity fixes

fix(dispatcher): preserve first quota error over generic HTTP 500

fix(quota): expire stale quota_exhausted_observed flag via TTL

feat(karaoke): anti-drift monotonic alignment, confidence scoring and 3 display modes (word, line, scroll)

feat(moldovan-qwen): cron quota guard + resume, junk audit, baker alternatives fix

fix(vault): derive audio_file from audio_filename/audio_url for legacy song manifests

fix(burst): route UNET through ModelSamplingSD3 shift=8 per official Comfy-Org Wan2.1 template

feat(burst): --max-cost hard cap refuses VM creation when estimate exceeds budget

Add burst --detach mode, macOS notify, and manifest auto-generation

Add Hyperstack burst I2V scene renderer with offline unit tests

feat(vault): orphan karaoke video reconciliation and sidecar audio fallback

feat(ui): add karaoke video playback modal and render triggers to track cards

fix(vault): persist karaoke video fields into song manifest on standalone render

fix(lyricist): dual anti-prompt-leak defense (prompt rule 0 + parser filter)

lyricist: uncensored-first LLM chain + short punchy composition rules

studio: expose karaoke_video in vault/music + flag karaoke videos in vault/videos

feat(karaoke): link rendered karaoke video into song manifest

feat(karaoke): default-on auto karaoke video for every new song

Merge remote-tracking branch 'origin/main' into deploy_main

feat(karaoke): word-synced lyrics video engine + vault sidecar features

fix(deps): unpin qwen-tts family; feature-gate Breeze TTS 2

test: pass FLOW_API_KEY bearer header in space_client fixture when set

feat(zerogpu): add Breeze TTS 2 experimental voice + remove image gen by policy

freight: QLoRA pilot complete - 98.1% holdout CE improvement, adapter on private HF repo

feat(freight): dataset pipeline — clean, dedup, redact, split, version

feat(telemetry): record completion tokens for GPU throughput metric