minivectordb webtextcrawler requests aiohttp duckduckgo-search text-util-en-pt PyMuPDF==1.22.5 xmltodict torch transformers numpy<2 onnx onnxruntime==1.18.0 onnxruntime-extensions faiss-cpu scikit-learn thefuzz[speedup] git+https://github.com/FlagOpen/FlagEmbedding.git@7a5eed7459f93b286ca2b46ed8e8d5795b4f9168#egg=FlagEmbedding