Create a OpenAI compatible text-embddding server from open source text embedding model.

First, create new virtual environment by conda and activate it.

conda create -n openai-text-embedding python=3.11
conda activate openai-text-embedding

Then, install the dependencies.

pip install -r requirements.txt
# Or use mirror
pip install -r requirements.txt -i https://pypi.tuna.tsinghua.edu.cn/simple

Clone the repository, then modify the model_path.json file to specify the path of your model or the respository of your model.

{
    "model_path": "path/to/your/model"
}

Then, run the server by

uvicorn text_embedding:app --reload

You can open the http://localhost:8000/docs to test the server.