fix: sync file progress and CUDA graph reuse fixes

fix: sync streaming and PE fixes while preserving ZeroGPU integration

Slow-PE UX: drop client PE watchdog, hide fallback PE; mobile example bottom sheet

Mobile UI pass + serve comment sync

Bind mp4-mode result download to its own session via rec id

mobile team gallery: hero + 2x2, the hero at full res (matches the maintenance page)

rotation: video-track resize re-picks orientation while idle (target dims and crop mask were start-time-only)

mobile topbar: one-line brand (hide the tagline, nowrap); banner text-wrap kills the orphan char

mobile: keep the two portrait panels side by side; wrap chip rows and grow touch targets; debug readouts behind the Profiling toggle; team gallery becomes a 2-col grid on small screens

maintenance page english-only (hero + 2x2 grid); merge session-init into the warmup stopwatch (measured 10-15s, not 30s); trim stale comments

grace-period wording: left 0 shows time-is-up instead of a zero countdown

quota banner: the 60s->30s->shared-pool ladder and the 60s-balance advice, zh/en; app pool relabeled shared (it is not owner-billed)

lease_left repaints only over its own hint (prefix test replaces four state flags); drop a self-evident comment

lease countdown: fork ticks lease_left each second; client shows remaining during warmup and run

client: staged startup progress with per-phase timers (model warmup vs session setup); drop ellipses from hints

acquisition ladder: visitor 60/30/10 then app pool 60/30/10 on quota refusals; client shows pool/seconds/2x for wait+grant; lease expiry is a normal end, not an error; busy vs quota split; /ws honors maintenance

terminal error message is sticky: session_timeout/close no longer wipe it

wan555-equivalent acquisition: one visitor-billed @spaces.GPU per session, nothing else

visitor quota pays for usage (wan555 pattern): mint identity restored; free waiting room + visitor-then-pool single schedule

gradio-pattern acquisition: free in-app waiting room, single app-pool schedule per turn, visitors never billed; drop the mint/probe machinery

gpu states render in the output overlay; the banner belongs to the resolution notice

remove the dead gpu_busy state

billing-free waiting: poll enable_queue=false probes; gpu_wait/gpu_ready/quota states in the client

visitors never wait in a billed queue: app pool first, instant-refusal fallback only, any queue timeout aborts

queue congestion: skip to the app pool instead of re-burning visitor reservations; live queue status in the client

quota exhaustion: structured error + bilingual banner instead of a raw traceback

hide the MP4 upload source on the Space (webcam only)

res banner: bing-fei -> er-fei

res banner: no-PE note, firmer wording; team note: native 1248x720 + Claude Fable 5 emphasized

team strip: note that Claude Fable 5 PE is not available on the Space

team gallery: strip above the fold + lightbox originals, served only where /team.json exists

sync from local: PE template overhaul + anthropic backend; low-VRAM mode (new lowvram.py); leaner FP8/dit internals; graph-path set_timesteps skip; UI kv-reset 600 + PE model label

print diagnostics in full (drop error/log truncation); align joyomni_streaming with repo HEAD (drop dormant skip-warmup gate)

drop the handshake readiness gate (fast first click fell back to anonymous); unify .gitattributes with main

speak the zerogpu-headers postMessage handshake with the huggingface.co parent (gradio protocol); /mint stashes the parent-minted quota headers, /ws redeems the ticket

claim user identity over GET /mint (edge translates __sign on HTTP only) and hand /ws a one-time ticket; visitor tier now runs user-scoped

forward HF __sign JWT on the WS URL so logged-in users bill their own quota; log/report token identity; add ?probe=1 identity check

revert gradio-lease experiment: any Blocks instance breaks the FastAPI-demo startup path (4 cycles verified; reference streaming Space uses the no-Blocks topology). Keep the boot speedup + ladder.

native gradio lease path (LTX-style): frontend claims the GPU via a real gradio queue endpoint (ZeroGPU handshake = personal billing/PRO priority); direct ladder kept as 6s fallback

final quota design: per-visitor IP billing with app-pool fallback (user-level billing unreachable over raw WS — scheduler 401s non-handshake tokens); trim comments

three-tier quota ladder: logged-in user JWT (personal/PRO pool) > per-IP visitor token > shared app pool; frontend fetches the user JWT via hf.co/api/spaces/<id>/jwt before connecting

bill per-visitor ZeroGPU quota (expose X-IP-Token via LocalContext in the session thread); surface server errors in the UI instead of a dead hourglass

sync to 3478e4b: drop write-only state; surface downlink encode time in profiling

sync to e756766+9ee7f46: shared gate detection, fps-scaled gate timings, session hardening UI; adapt engine to new gate API

mirror deploy/static