Merge remote-tracking branch 'origin/main'

Update engine.py

fix(painter, modal): fp16-safe VAE, rembg session reuse, scaledown window

feat(engine): split play_turn into text+image phases, add resume and session_info

feat(orchestrator): anti-repeat retry, error fallback, scout routing

feat(state): milestone50, remember_fact, beat guard, new_character ordering

feat(memory): structured budget fit, milestone-50 hint, pacing nudge

feat(prompts): force grammar fields, scout_schema, action_note, lru_cache

feat(schemas, config): add remember_fact directive, ViewState.history, IMAGE_VAE

fix(llm): _hf_gen_kwargs, truncated-JSON repair, ModalLLM warmup, debug-gated logs

refactor(utils): extract shared helpers into visualnovel/utils.py

chore(llm): clean comment

fix(llm): Try to catch the text after the last <think>

fix(llm): when using action (such as look around), remove the thinking

Fix(schema): non restrictive emotion mapping from LLM response

fix(TransformersLLM): strip Qwen3 think-tags, add LLM logging, atomic

fix(tts): handle empty audio from kokoro_onnx.create() gracefully

fix(TransformersLLM): handle missing/fenced JSON with retry and fallback

fix(zerogpu): persist game state to /tmp after every mutation for worker

fix(zerogpu): persist phase state, because zerogpu run each function in

fix(TransformersLLM): inject resolved schema skeleton into prompt for

fix(ZeroGPU): apply the same fix of zerogpu model's loading for the

feat(ZeroGPU) : Fix HuggingFace logs error.

chore(reformat): reformat existing code with ruff.

Feat : Music and Relations

Feat : Save System

Auto-download Kokoro model files if missing at startup

Make IMAGE_MODEL/LORA defaults backend-aware

Merge branch 'main' of https://huggingface.co/spaces/build-small-hackathon/Hackathon-IA-VisualNovel

Journal

Auto-detect HF Space env; fix image dependency check

Merge origin/main into feature/update-vocal_system

Update TTS with "voice description mapper"

save

Add TTS

Fix backdrop character bleed: guidance scale + cleaned style prompt

Add STT fixes, SDXL-Lightning backend, voice UX and rembg sprite transparency

Lots of thing

Fix quiet_stderr for windows user

Initial commit: Reorganized project structure