development quickstart

Important Note: Make sure you have Python 3.10.9. For compatibility reasons.

Note: path is relative to project root folder

1. code and dependency

git clone git@github.com:ClinicianFOCUS/FreeScribe.git

# if using CPU
pip install client_requirements.txt

# if using GPU
pip install client_requirements_nvidia.txt

If you encountered a compilation error when installing llama_cpp_python, try this tutorial https://www.youtube.com/watch?v=r-05yuXTEPE

It will show you how to build and install Llama.cpp from source, along the way you will solve environment/dependency issues.

2. __version__ file

place a __version__ file in src/FreeScribe.client folder with content like v0.0.9466.alpha.

Otherwise, you'll see error like:

End result be like:

It is generated from release. We create the file manually when developing.

https://github.com/ClinicianFOCUS/FreeScribe/blob/main/.github/workflows/release.yml

      - name: Create Version Text File for PyInstaller
        run: |
          $tag = '${{ github.ref }}' -replace 'refs/tags/', ''
          echo $tag > .\scripts\__version__
        shell: pwsh

You can find the latest releases here: https://github.com/ClinicianFOCUS/FreeScribe/releases/tag/v0.0.9466.alpha

3. Create folder src/FreeScribe.client/models, place LLM model there manually

The app will be using a local whisper model for speech-to-text service and an LLM model for note generation.

STT model will be downloaded automatically, while the LLM model needs to be downloaded manually.

Download it here: https://huggingface.co/bartowski/gemma-2-2b-it-GGUF/blob/main/gemma-2-2b-it-Q8_0.gguf, place the file in path src/FreeScribe.client/models to eliminate the error:

4. start client

Activate your virtual environment, go to path src/FreeScribe.client, start the app:

python client.py