Data Product Runtime Artifacts

Floe’s alpha data product deployment path should be understood as an artifact flow.

Recommended Alpha Shape

1. Data product source lives in a product repository.
2. CI runs dbt checks and uv run floe platform compile --spec <product>/floe.yaml --manifest <platform>/manifest.yaml --output <target>/compiled_artifacts.json --generate-definitions.
3. CI builds a product runtime image containing product code, dbt files, compiled artifacts, and Dagster definitions.
4. CI publishes the image to the organization registry.
5. The organization deployment path updates the Dagster code location or release values to use that image.
6. Dagster launches Kubernetes work and emits OpenLineage and OpenTelemetry evidence.

Handoff Patterns

Pattern	Use When	Floe Contract
GitOps PR	Platform changes are deployed from a GitOps repo	CI proposes image tag and values changes
CI deploy job	Your release workflow can deploy after approval	CI deploys the image and records evidence
Service catalog request	Platform team owns production deployment	CI publishes artifact metadata and requests deployment
Release train	Production changes move in scheduled batches	Artifact digest and evidence are promoted together

Lower-Level Primitive: floe-jobs

charts/floe-jobs can render Kubernetes Jobs and CronJobs for dbt, dlt, and custom workloads. Use it when your Platform Engineer has approved that pattern. Do not treat it as the default self-service alpha story until a complete product workflow is documented and validated.