Custom Embedding models
Using custom embedding models in Reor.
Reor will automatically download the embedding model you choose from the dropdown. However, you can also attach your own local embedding model or specify a custom remote model.
Attach a Local Embedding Model
To attach a local embedding model, you'll need to download it from Hugging Face using Git Large File Storage (LFS).
1. Install Git LFS
Follow this link to install Git LFS.
2. Download a model
To download a model now, you will need to git clone it. Only "Xenova" ONNX embedding models are supported. You can find all of those models here.
To download the model "Xenova/bge-m3" for example, run the command:
git clone https://huggingface.co/Xenova/bge-m3
3. Attach to Reor
Go to Settings -> Embedding Models -> Attach Local Model and select the folder where the model is located.
Allow Reor to download the model for you
You can also let Reor download the model for you: specify the Hugging Face "Repo" in settings, go to Embedding Models -> Attach Remote Model and enter the model name. For example, Xenova/bge-m3