Skip to content

Documents · Source engineering and validation

Source engineering and validation

For everyday use, start with the main noun references: elevation, roads, and streamflow. Each has an executable real-data vignette. This developer directory preserves acquisition history.

What changes when the input is terrain, a mapped road network, or observations from a streamgage? These three contained projects test that question using real provider data. They reuse CubeDynamics' grammar and source QA, while keeping the source-specific meaning visible.

The original proofs below are retained as historical access evidence. The new installed candidates add bounded retries, raw snapshot replay, pinned range reads and stricter release gates. They are still not production catalog registrations; read the current scope and remaining gates.

These are experimental project-owned nouns, not additions to cubedynamics.data.list_sources() or production serving revisions. Their Python modules live under examples/source_projects/; install this checkout editable and run from the repository root. Each has its own command, tests, report, real-data figures and stop boundary. No website tab or generic source framework was added.

What the proofs establish

Small real requests, explicit limits, inspectable provenance, preserved native semantics, structural/numerical checks, and reviewed diagnostic figures. They do not establish spatial completeness, independent ground truth, broad provider reliability, or suitability for a particular scientific decision.

Raw upstream samples remain in ignored artifacts. Small reviewed figures and machine-readable evidence are published here. Rerunning a rolling query may produce different content; retain the raw response and checksum when exact replay matters. Publication evidence is checked offline; live provider checks are separately marked online and integration.

Custom nouns · Custom verbs · Source lifecycle