The documentation describes that FAISS builds shared & static libraries but these do not appear to end up in the installed Python packaging. Rather, the Python conda packaging only installs the swig-converted artifacts.
If a codebase, dependent on FAISS, is being built against the libfaiss.so and/or libfaiss_gpu.so artifacts, this leaves developers with the option of forcing contributors to build FAISS themselves, or including the libs in the codebase.
Further clarification on the ideal way of packaging against FAISS C++ API's would be ideal.
The documentation describes that FAISS builds shared & static libraries but these do not appear to end up in the installed Python packaging. Rather, the Python conda packaging only installs the swig-converted artifacts.
If a codebase, dependent on FAISS, is being built against the libfaiss.so and/or libfaiss_gpu.so artifacts, this leaves developers with the option of forcing contributors to build FAISS themselves, or including the libs in the codebase.
Further clarification on the ideal way of packaging against FAISS C++ API's would be ideal.