# Capabilities
## Properties
**schema_introspection** | **bool** | The connector can enumerate a dataset's columns authoritatively. `false` means it has no schema authority at all (a headerless CSV, a schema-less collection) -- not that the dataset is missing; the mapping is then reported as unvalidated (a warning) rather than as every field being absent. |
**stable_paging** | **bool** | The connector can page a dataset by a unique, total order, so batches are a stable partition with no skipped or duplicated rows. `false` makes it unusable as a source and blocks the run. Whether one particular table has such a key (e.g. a primary key) is a separate, per-table runtime error. |
**nested_source_paths** | **bool** | The connector can resolve a dotted `ColumnMapping.source_column` (`profile.address.city`), not just a top-level name. A relational connector declares `false`, and any nested mapping entry is then an `Error`-severity issue naming the offending path. |
**writable** | **bool** | The connector can be a migration target at all. `false` is what \"this target is read-only\" is read from -- distinct from the per-table, per-connection question of whether the current role holds `INSERT`, which is probed separately and reported through `MappingReport`. |
**transactional_batch** | **bool** | A batch is applied atomically -- all rows or none. `false` does not block a run, but is warned about: a failure can leave a partially applied batch to reconcile by hand. |
**resumable_cursor** | **bool** | The connector's page cursor is durable across a process restart. This is only the connector half of checkpoint/resume; the engine does not persist run state yet, so no run is resumable today regardless of this value. The dry-run report says which half is missing. |
[[Back to Model list]](../README.md#documentation-for-models) [[Back to API list]](../README.md#documentation-for-api-endpoints) [[Back to README]](../README.md)