Snowflake-Labs/ml-pipelines — reverse-engineered prompt

Build me a reusable Snowflake MLOps starter repo in Python that teams can copy into their own GitHub project.

I want it to help deploy machine learning workflows into Snowflake as scheduled tasks, with separate dev, staging, and prod environments. Include setup scripts that create the Snowflake roles, databases, warehouses, stages, service users, and load a small demo dataset. Add a simple versioned feature store where I can define entities and feature views in a config file, then deploy them locally.

Please include a template for creating new ML projects, an example project, shared ML utility code, and scripts to deploy a project, run its DAG right away for testing, and clean up project or feature resources. Add GitHub Actions so feature branches deploy to staging and main deploys to prod using Snowflake secrets.

Use Python 3.10 and uv, and write clear README instructions for setup, local development, secrets, and common commands. Look up current Snowflake docs if needed.