Prompt: repeat "a cute" per animal for clearer multi-animal scenes

Fix place picker to 6 columns (was hardcoded to 5)

Add boat + space shuttle places (18 total → 3 rows of 6)

Expand picker: 24 animals (lookalike-grouped) + 16 places

README: app flow screenshots + hackathon note; fix gen concurrency

Docs: publish LoRA trainer + add README fine-tuning story

Integrate trained NumZoo LoRA into the app

Fix ZeroGPU initialisation: spaces before torch, two-phase pipeline load, guidance_scale

Prep captions for LoRA training: trigger word + content-only

Regenerate training image 011 — replace human girl with chibi butterfly

Switch dataset model to Qwen-Image + add 54-image training set (LFS)

Dataset: FLUX.2-dev + app-aligned prompts with multi-animal/place combos

Docs: update CLAUDE.md + README for FLUX.2-klein-4B

Make HF cache setup bulletproof — never crash import, fall back to ephemeral

Fix poisoned HF cache: fresh path + self-healing wipe

Fix cache cleanup: wrap removals in try/except, also clear .locks placeholder

Fix HF Spaces startup: restore spaces import + clear stale cache placeholders

Restore startup pre-download for FLUX.2-klein-4B on HF Spaces

Fix: restore @spaces.GPU decorator required by ZeroGPU

Upgrade image model to FLUX.2-klein-4B + align training pipeline

Restore icons in status bar and enforce consistent panel width

Improve level experience: better status bar + level 5+ harder difficulty

add samples and banner

Add banner image to app header with styled title

Fix wrong-answer feedback; add numeric keyboard on mobile

Fix My Collection title card in dark mode

Revert emoji pre-selection to 1 animal + 1 place

Improve UI workflow: clean quiz, collection title card, no spinners

Remove streak from prompt; support up to 3 animals and places

Fix collection display: replace JS polling with server-side render

Add training dataset generator + fix MPS dtype for local inference

Remove startup model pre-caching — let from_pretrained cache in container

Fix .locks dir placeholder — also repair cache_root/.locks/models--*/

Fix: repair entire HF cache tree of placeholder files at startup

Fix FileExistsError on cache path: unlink placeholder file before makedirs

Change cache path to /data/hf_cache to avoid storage git conflict

Fix ENOTDIR by validating every cache path level; add diagnostics

Fix corrupt cache detection (lexists) and suppress asyncio GC noise

Fix corrupt cache entry and asyncio fd warning on startup

Redesign game loop: level-based flow with reward at end of each level

Revert share=True — not supported on HF Spaces

Enable share=True for public Gradio link

Fix FileExistsError on /data/huggingface/hub mkdir

Async pre-generation, fix loader, fix collection display

Run snapshot_download in thread to suppress asyncio fd warning

CLAUDE.md: document audioop-lts Python version constraint

Remove audioop-lts from requirements — only needed on Python 3.13+

scripts: remove hardcoded paths, clean up run.sh kill logic

Use persistent storage (/data) for model cache on HF Spaces Pro

Remove snapshot_download at startup — crashes container (16Gi limit)