Reproducible data-science stack: 3 multi-stage, non-root, pinned images (JupyterLab / papermill batch runner / MLflow server) plus a 4-service compose file, 23-variable .env, Makefile, smoke-test notebook and README — 16 files.
docker compose up -d
DockerfileDockerfile.batchDockerfile.mlflowcompose.yamlrequirements.lab.txtrequirements.batch.txtrequirements.mlflow.txtrequirements.user.txt.env.examplejupyter_server_config.py.dockerignoreMakefilescripts/run_batch.shnotebooks/00_smoke_test.ipynbREADME.mdCHANGELOG.mdThis page is the working piece. The full pack has everything below.
3 multi-stage images · 4 compose services · 33 pinned library versions · 23 settings in one .env — every FROM line carries an exact version tag, so the rebuild you run next year resolves the
Measured on this shelf: a rival paid Docker data-science stack sells at USD 19 today, which is the price here. We publish no 'a consultant would charge X' figure because we have not measured one.
Send it to me