MLflow

Track experiments and register models in BullSequana AI through Studio and the BSQAI API.

Agentic Friendly

MLflow holds experiments, runs and registered models in BullSequana AI. You reach it only through the Models area of Studio in the Portal (Experiments and Model repository) and through the BSQAI API. Both work on your organization's MLflow and apply the MLflow permissions: what you create is yours until you share it.

The Global organization uses the platform's MLflow. Any other organization uses its own MLflow instance, and an organization without one has no MLflow. GET /v1/mlflow/capabilities tells whether the active organization has MLflow.

Dev environments and pipelines have no direct MLflow access. They get no MLflow address or credentials, the MLflow UI is not offered, and the MLflow SDK cannot reach the tracking server. To record work from a dev environment, use the bsqai_mlflow client that the Python JupyterLab image includes, or call the BSQAI API with an API key, as the guides below show.