geointel/docs/specs/DATA_LIFECYCLE.md

Data Lifecycle

Original dataset lifecycle

1. User or system registers dataset.
2. Original file/reference is preserved.
3. Dataset enters UPLOADED or CREATED.
4. Validation extracts basic file and geospatial metadata.
5. Metadata is persisted.
6. Dataset becomes READY, REQUIRES_CRS or failure state.

Derived dataset lifecycle

1. Operation is requested from a ready parent dataset.
2. AnalysisRun is created.
3. Job produces derived artifact.
4. Derived dataset references parent dataset and analysis run.
5. Derived metadata is extracted.
6. Derived layer can be rendered/exported.

Detection lifecycle

1. Detection run is created.
2. Inputs and parameters are stored.
3. Model/import produces candidate features.
4. Features are georeferenced.
5. Features are validated.
6. Features are persisted as detections.
7. Optional QA/QC compares them to reference features.
8. Detections may be exported.

Segmentation lifecycle

1. Segmentation run is created.
2. Raster and model parameters are stored.
3. Model/import produces masks.
4. Masks are georeferenced.
5. Optional polygonization creates vector features.
6. Features and/or mask paths are persisted.
7. QA/QC and export may run.

Deletion and archival

• Deleting a project should not immediately physically delete files unless cleanup is requested.
• Archival removes items from active lists but keeps reproducibility.
• Physical cleanup must never leave dangling database references.