Metadata Stores API ποΈ
Metadata stores are the lineage backbone of FlowyML. They record every pipeline run, the status of each step, input/output artifact references, and timing information. FlowyML ships with a lightweight SQLite backend for local development and supports PostgreSQL and cloud-managed alternatives for production deployments. All backends implement the MetadataStore interface documented below.
Metadata Stores track pipeline runs, step status, and artifact lineage.
Base Metadata Store
Bases: ABC
Base class for metadata storage backends.
Functions
list_assets(limit: int | None = None, **filters) -> list[dict]
abstractmethod
list_pipelines() -> list[str]
abstractmethod
list_runs(limit: int | None = None) -> list[dict]
abstractmethod
load_artifact(artifact_id: str) -> dict | None
abstractmethod
load_run(run_id: str) -> dict | None
abstractmethod
query(**filters) -> list[dict]
abstractmethod
save_artifact(artifact_id: str, metadata: dict) -> None
abstractmethod
SQLite Metadata Store
See Also
- Artifact Stores API β the companion storage layer for binary step outputs
- Plugins Overview β how metadata-store backends are registered as plugins