Closed maoberpriller closed 11 months ago
The problem was due to missing wheels for the PyMuDF, FAISS-cpu, and tiktoken libraries, which causes pip to try and build from source. Unless Windows Visual Studio is installed on your machine, you will get these errors. This appears to only be an issue with python version 3.12 and above. Downgrading to python 3.10 or 3.11 resolves this issue. However, you will still get a failure on the llama-cpp-python library as it can only built from source. If you don't plan on using a local model, you can just skip it by commenting it out of the requirements.txt file. The app will run without llama support, but the OpenAI models will work just fine.