Connect IBM watsonx.ai to Label Studio

Overview

IBM watsonx.ai provides a suite of foundation models for natural language and multimodal AI tasks. With Label Studio, you can integrate watsonx models directly into your labeling workflow to generate pre-annotations, refine model outputs, and accelerate iterative training.

How it works

Label Studio connects to watsonx.ai through a custom ML backend that runs locally or in your cloud environment. You can use the watsonx_llm example backend (Docker or Python) to send task data from Label Studio to your watsonx.ai project, retrieve model predictions, and display them as interactive drafts in the labeling interface.

Optionally, the same service can sync completed annotations to WatsonX.data through webhooks, making it easy to track human feedback and build continuously improving datasets.