Skip to main content

mlflow

Catalog date: 2026-06-24.

The platform's model registry + experiment tracker. Owned by alphaswarm_models — every Predictor, AlphaBacktestExperiment, walk-forward run, and finetune trainer registers here.

Identity​

FieldValue
Service idmlflow
Rolemlops
Imageghcr.io/mlflow/mlflow:v3.14.0-full (compose and Kustomize both pin this line — compose comment: "One tested MLflow 3.14.0 line across every deployment surface"; deployments/kubernetes/base-services/mlflow/ also builds a custom ghcr.io/alpha-swarm-ai/alphaswarm-mlflow:3.14.0-pg-s3-v1 migration-job variant)
Port5000
Storageobject store for artifacts (MinIO / S3 / GCS / ADLS depending on cloud); tracking-store backend differs by surface — compose uses MLFLOW_BACKEND_STORE_URI: sqlite:////mlflow/artifacts/mlflow.db (SQLite), while the Kustomize base-services/mlflow/ Deployment uses MLFLOW_BACKEND_STORE_URI pointed at Postgres

Deployment surfaces​

SurfaceWhere
Composeservice mlflow in alphaswarm_platform/compose/docker-compose.yml
Kustomizedeployments/kubernetes/base-services/mlflow/ — Deployment + Service + ExternalSecret-backed credentials
MLOps overlayreachable through mlops/ when paired with Argo Workflows + Dagster

Dependencies​

Upstream:

  • postgres — tracking store.
  • minio / s3 / gcs / azblob — artifact store.

Downstream:

  • alphaswarm-core, alphaswarm-worker — every Predictor / Skill / walk-forward / finetune flow registers runs here.
  • alphaswarm-ml-mcp — read paths surface through the data.ml.* MCP slice.

Operations​

  • Auth: behind the cluster ingress; the in-cluster URL is the only path. Local dev exposes http://localhost:5000 for browser inspection.
  • Pruning: no alphaswarm/tasks/cleanup/ directory or mlflow_prune.py file was found in alphaswarm, and no MLflow reference exists in alphaswarm/tasks/retention_tasks.py — flagging the earlier pruning-task claim as unverified/likely stale rather than restating a path that does not exist.
  • Run tagging: every run is tagged with the originating experiment_id + test_id per AGENTS rule 34 so audit queries can correlate ML runs with strategy / backtest activity.

See also​