fix

fix: quantization_config 역직렬화 오류를 config.json 패치로 완전 해결 (로컬 통합 테스트 통과)

fix: keras 3.12.1 + tf 2.21.0 복원 (로컬 환경 일치) + quantization_config 역직렬화 오류 3단계 fallback 추가

fix: tensorflow 2.15 (keras 충돌 제거) + 에러 메시지 500자로 확장

fix: local model load fallback + score grade color fix

Fix SyntaxError in css string

Remove Soft theme (blue label bug), unit back in label parentheses

Unit in info field, enlarged section title

Fix placeholder encoding, enlarge title

Add unit suffix to input field labels (만원, 일, 개)

UI: dark mode, placeholder, 2-col inputs

Fix RAG KeyError L10210000; Add grade-colored HTML score display

Enhance LLM personal features context

Fix Keras serialization mismatch by matching local TF/Keras versions

Fix Gemini model configuration and R/A/G display

Fix scikit-learn version mismatch and visualize R/A/G steps

fix

Hotfix: Gemini 404 and enhanced error visibility

Final stable: hf_hub_download and dictionary-based chat

Emergency fix: Remove type='messages' and restore Tuple format for absolute Gradio compatibility

Fix model load traceability and Chatbot data format

Robustness update: Detailed error reporting and stable chatbot format

Final stable sync: Tuples for Chatbot and HF health check

fix

Fix HF deployment errors: Gradio 5 upgrade and 0.0.0.0 binding

Clean up scratch files and finalize robust deployment

fix

fix

Fix NameError in app.py and ensure internal loan count calculation works

Remove L10210000 from UI and calculate internally

Corrected feature labels, implemented loan sum logic, and fixed API key config

rem

Fix feature mismatch and typos in credit_app.py

Clean: Synced code for Hugging Face (binary files handled via API)

UX: Improve all input labels with full descriptions, units, and example values

Build Fix: Add build-essential tools and loosen pip version constraints

Critical Fix: Use Gradio 5.2.0, tuple-based Chatbot, and Docker ENV vars to resolve all crash issues

Fix: Final memory optimization to prevent RUNTIME_ERROR by deferring embedding model loading

Fix: Upgrade to Gradio 5.1.0 and enable queue to resolve TypeError and sharing issue on Docker

Final: Project-wide Gemini integration and Dockerization for production

Optimize: Final refinement for Hugging Face Docker SDK (0.0.0.0:7860)

Dockerize: Add Dockerfile and switch Spaces SDK to docker for ultimate stability

Ultra-Stable: Defer ALL heavy imports to function call to prevent boot crash (500 error)

Stabilize: Downgrade to Gradio 4.44.1 and implement Lazy Loading to fix 500 error

Fix: Expand UI features to 14 to match prediction model and fix No API found error

Hotfix: Add monkeypatch for HfFolder ImportError

Fix: Downgrade Gradio to 5.4.0 to resolve HfFolder and dependency conflict

Fix: Resolve HfFolder ImportError by pinning huggingface-hub to 0.24.0

Unify LLM to Google Gemini project-wide and cleanup Ollama codes

Switch LLM to Google Gemini 1.5 Flash (Google AI Studio)