fix(spaces): pin torchaudio==2.10.0 to match HF base torch

fix(spaces): module-level @spaces.GPU so ZeroGPU's startup detector finds it

fix(spaces): pin custom nodes to SHAs (also fixes LTXVideo's master default)

fix(spaces): _git_clone uses init+fetch so SHA refs work

fix(spaces): use \$HOME instead of /data (no persistent-storage add-on needed)

fix(spaces): pin sdk_version to 5.50.0

fix(spaces): suggested_hardware uses zero-a10g (ZeroGPU)

chore: rename Space title to 'LTX 2.3 Studio'

feat(spaces): registry entries for renamed BF16 aliases + seed inputs + bootstrap

feat: API-format workflows + sidebar nav + responsive UI

fix(backend): pass cache_args to PromptExecutor + dump worker tracebacks

fix(backend,models): runtime corrections from end-to-end testing

ci: optional deploy-on-main to HF Space

ci: run unit tests + ruff lint on every push

feat(app): generate handler — async streaming, status banner, video output

feat(app): per-mode forms with media inputs, presets, advanced accordion

feat(app): Gradio shell with sidebar nav and empty mode tabs

feat(ui): categorized lora_chrome — camera dropdown, detailer, IC, pose

feat(ui): preset_bar + status_banner components

feat(backend): submit() async generator with progress hooks + ZeroGPU

feat(backend): ComfyUILibraryBackend skeleton + event dataclasses

feat(tools): refresh_models materializes every required model

feat(models): ensure_models with HF cache symlinks + missing-file fallback

feat(models): walk_workflow_for_models scans loader nodes

feat(models): MODEL_REGISTRY mapping filenames to HF repos

fix(workflow): support dict-keyed widgets_values for VHS audio/video nodes

feat(modes): A2V + Lipsync + Keyframe + Style parameterize_fn

feat(modes): T2V + I2V parameterize_fn with stage maps

feat(modes): Mode dataclass + empty MODE_REGISTRY skeleton

feat(workflow): set_input + validate over node graph

feat(workflow): load_template returns fresh deep copy per mode

data: extracted mode-specific workflow templates from master

feat(tools): extract six mode templates from master workflow JSON

docs(plan): correct ComfyUI import paths and pin runtime clone to submodule SHA

chore: vendor ComfyUI as git submodule pinned to working commit

test: pytest fixtures (master_workflow, canonical_inputs, fake_hf_cache)

docs: add MIT LICENSE

docs: README with Spaces front matter and local quickstart

fix(setup): replace `|| true` with explicit existence check for refresh_models.py

chore: idempotent setup.sh — venv, submodule, custom nodes, models

chore: pin runtime + dev dependencies and configure pytest/ruff

docs(claude): expand project guidelines with TDD discipline, commit format, common pitfalls

docs: implementation plan — 27 tasks across 8 phases

docs: design spec for v1 — Gradio + bundled ComfyUI library backend

Initial commit