Skip to content

MLflow (experiment tracking + artifacts)

1. Overview

mlflow is an Atlas service family in the apps category. Its implementation and service-owned documentation live under services/mlflow/.

2. Role In Atlas

Atlas uses this service according to its manifest, topology row, SOURCE settings, dependencies, and runtime data-flow declarations.

3. Tracks And Category

  • Category: apps
  • Kind: container
  • Tracks: all, ml-eng, trading

4. Access

  • Kong aliases: mlflow.localhost
  • Port variables: MLFLOW_PORT

5. Configuration

  • SOURCE variables: MLFLOW_SOURCE
  • Default SOURCE values: disabled
  • Available SOURCE values: container, disabled

6. Dependencies And Topology

  • Required dependencies: supabase, minio
  • Optional dependencies: jupyterhub
  • Runtime calls: supabase, minio

7. Source Values

SOURCE Variable Default Values
MLFLOW_SOURCE disabled container, disabled

8. Runtime Integration

The manifest data-flow list declares runtime calls to supabase, minio. The topology row supplies aliases and port surfaces used by the generated gateway and service references.

9. Architecture

10. Operations

Use ./start.sh to configure this service through the wizard or pass the matching SOURCE flag when the service is source-configurable. Use ./stop.sh to stop the active Atlas project.

11. Source Documentation