Skip to content

Stability and compatibility

Section 22 · From docs/INVENTORY_REFERENCE.md on GitHub

What you can rely on across 0.6.x releases:

  • The field names and types in this document. New fields and new metadata keys may be added; consumers should ignore keys they do not know.
  • asset_type, edge_type and relationship values are never renamed within 0.6.x. New values may appear (a new collector can introduce a new asset type).
  • The exported file names and their top-level keys.
  • The public Python API: the signatures listed here and the package exports of cloudg.inventory (InventoryMapper, InventoryResult, RelationshipLinker, DependencyGraph, AWSDeepInventoryCollector, AzureDeepInventoryCollector, GCPDeepInventoryCollector, OrganizationTopology, discover_organization).

What you should not rely on:

  • id values, which are random per run. Correlate on arn.
  • The order of assets, edges, metadata keys and unsorted counts.
  • The exact metadata keys of a type staying identical between versions; the catalog shows what is produced today.
  • Names starting with an underscore, even when re-exported for backwards compatibility.

Secrets are never part of any structure: parameter values, environment variable values, passwords, VPN pre-shared keys, Direct Connect authentication keys, connection strings and Cloud Control properties matching the secret pattern are dropped at collection time, and GCP URL aliases keep only scheme://host. Trust policy conditions and resource policy principals are kept, because they define the relationships.