Questions teams ask first.

The Pipewright core is free and open source. You pay Google Cloud for what your environment uses, in your own project and on your own bill, so your credits and commitments apply. An idle starter environment costs little: the launcher scales to zero, and pipeline runs, endpoints and online stores cost money only while they exist. Pipewright Enterprise and the Pipewright team’s services are priced per organisation; ask us.

No. By default Pipewright uses Google’s managed services: Agent Platform Pipelines, Ray on Agent Platform and Agent Platform endpoints. Teams that already run GKE can use Kubeflow Pipelines, KubeRay and KServe in a private cluster instead, with the same use case files.

No. Pipelines stay plain Kubeflow Pipelines code. Pipewright compiles them, supplies the parameters, identities and outputs, and releases them. The examples show the conventions.

No. Pipewright runs inside your project: pipelines, models, endpoints, monitoring and evidence all stay there, encrypted with your keys. In the private blueprint nothing is reachable from the internet.

Yes. Switch on synthetic data in a use case and Pipewright learns it inside the environment that holds the real data, with no outbound calls. Keys and personal data are never learned. Each candidate is scored on fidelity, on how well a model trained on it does on real data, and on privacy attacks (exact copies, records too close to real ones, singling out, linkage and inference), and only one that passes every gate is published. Moving it to another environment needs every approver group your platform team requires, and leaves an evidence pack. Without any real data, Pipewright can also write rows from the data contract alone.

Not unless you ask it to. By default drift opens a decision case for people to approve, decline or snooze. A "candidate" mode trains a challenger automatically but never promotes it without a person.

By promotion, not by building it again. pipewright promote takes the release you tested, and the model you approved, to production: images copied by their digest, pipelines checked against the tested ones, the model's files verified one by one. Production's own policies apply first, and a production approver still decides when the model starts serving. Anything that differs from what was tested is refused, and every promotion is recorded and can be rolled back without rebuilding.

Yes. A use case names the features it needs, from BigQuery feature tables or groups registered in the Agent Platform Feature Store, and Pipewright reads each value as it was at the row's cut-off, so nothing from the future reaches training. Batch scoring reads features the same way. Your features stay with their owners: Pipewright copies nothing, and your platform team decides which feature datasets each team may read. Every model's evidence records exactly which features it trained on.

Yes, in the environments where you train. A step asks for a Ray on Agent Platform cluster in its YAML and gets one for that run only, sized within limits your platform team sets. Pipewright creates it with your keys and network, runs the step's Ray job on it and deletes it afterwards, so nothing is left running. Ray needs the VPC Service Controls perimeter in dry run, so it runs in development and test, while production keeps its perimeter enforced and receives the models you trained there by promotion. While developing, data scientists keep their own standalone clusters for notebooks; pipelines never depend on them.

Yes, with Fleet View in Enterprise. One read-only screen shows every use case in every environment: the live release and its digest, the champion and the candidates waiting, run failures, drift and retrain decisions, the evidence controls and each environment's security posture. A collector reads metadata only, never your data, and each viewer signs in with their Google identity and sees only their own teams.

Yes. Every model version records its lineage: the tables and features it learned from, the run, release and configuration that built it, its evaluation and gates, who approved it and how, its evidence pack and where it serves. pipewright lineage prints it in the core; in Enterprise, Fleet View draws it, with the model’s whole history across environments and the retrain cases between versions.

Yes, with BigQuery ML. A SQL team writes a short YAML file and its SQL; Pipewright trains, evaluates, registers, monitors and proposes retrains, with the same approvals and evidence as any model. The data and the model stay in BigQuery, and every job has a byte limit.

Yes. The Pipewright core is open source under the Apache License 2.0, on GitHub: use it, change it and run it in your own Google Cloud. Pipewright Enterprise adds the private-by-default blueprint, evidence packs, governed synthetic data releases, Fleet View and support.

Google Cloud only: on Agent Platform Pipelines, or on Kubeflow Pipelines in a private GKE cluster, with the same YAML and the same releases on both.

With their own Google identity. Approver groups are registered by the platform team; membership is checked by Google Cloud IAM, and every decision is recorded with the person's name and reason.