This /python folder is for running Huggingface-like AI models locally as micro-services, fetched on localhost by the main WorkBuddy C++ runtime. Ideally we want everything in C++ natively for realtime but can't remake every AI model, sadly the AI space is 90% Python.
Python version = 3.12
conda create --name workbuddy python=3.12
conda activate workbuddy
pip install -r python/requirements.txt
python python/socket_server.pyEnvironment Variables come from the workspace root .env outside of /python. .env is shared by C++ and Python.
https://github.com/pyannote/pyannote-audio
This model performs speech segmentation and diarization that can filter groups of speakers into categories. Useful for speech recognition commands and transcriptions.
Configure HUGGINGFACE_ACCESS_TOKEN in the .env file.
Accept the gated access user conditions on HuggingFace: