HumanSignal/label-studio-ml-backend

Can I use API-key instead of deploying a ML model?

Closed this issue · 2 comments

Niemj8919 commented

It seems that I have to download and deploy a machine learning model locally, and then let Label-Studio use it. Can I directly transmit the data to a LLM model (such as ChatGPT, Claude) using an API key and get the annotation results?

ethanporcaro commented

The model has to run through a backend, but there is a backend configuration to just forward to openai

  • 👍1
makseq commented

In Label Studio Enterprise we have Custom Javascripts that can be injected to your labeling configuration. Using them you can run requests to LLM models directly.

However, LS doesn't provide direct access to LLM models without LS ML Backends.