Transformer Pipeline (Engineering View)

Role

In this codebase, “transformer” often refers to the transform stage inside an ETL pipeline (e.g. raw → windows, or windows → feature vector), not to a separate standalone pipeline. The Features pipeline is the main place where “transformation” from windows to feature rows happens.

---

Where It Fits

• ETL architecture: Each pipeline type has a Transformer component (see ETL architecture) that implements the “raw → useful shape” step.
• Feature vectorization: The features pipeline uses a FeatureVectorizer (and optionally temporal context) to produce one feature row per window; that is the main “transformer” for ML.

---

Full Reference

• For the features transform and config: Features pipeline and Features.
• For the generic ETL transform interface: ETL architecture.