# ShuiLi ZhiXue Quiz System - OPTIONAL dependencies # NOT installed by start.bat by default. These power scanned-image OCR, # local RAG engine, and video speech transcription. They are large (they pull # torch / onnx / native binaries). The app degrades gracefully without them. # # === 一键安装全部AI组件(推荐) === # install_ai_components.bat # 或手动: # "D:\Python311\python.exe" -X utf8 -m pip install -r requirements-ai.txt # # === 本地RAG引擎(扫描教材全链路:MinerU + BGE-M3 + Qdrant + Reranker + CLIP) === # 详见 requirements-ai.txt # # Lightweight local OCR (ONNX runtime, no GPU needed) rapidocr-onnxruntime>=1.3.0 onnxruntime>=1.16.0 pypdfium2>=4.0.0 scikit-image>=0.21.0 # # Deep-learning OCR (large; pulls torch / torchvision, several GB) easyocr>=1.7.0 # # Tesseract OCR engine: ALSO install the Tesseract binary for Windows and add it # to PATH (the pytesseract wheel alone is not enough). pytesseract>=0.3.10 # # pdf2image: ALSO install poppler for Windows and add its bin/ to PATH. pdf2image>=1.16.0 # # Video / audio speech-to-text for mnemonic extraction (ctranslate2 based) faster-whisper>=1.0.0