<span id="两套连接器元数据协议"></span>

# Two connector metadata protocols

/connectors/db and /connectors/structured/v1 are both registered and used separately by current SDKs. An older path is not necessarily deprecated. Changing the URL alone does not migrate the protocol.

| Item | /connectors/db | /connectors/structured/v1 |
| --- | --- | --- |
| Source type | Derived from the saved connector_id. | Request includes source_type, checked against the connector type. |
| Database/schema lists | String arrays in databases or schemas. | name/display_name object arrays and endpoint_results. |
| Table/object lists | Table-name strings in tables. | objects with database/schema/table/collection/object_kind. |
| Field structure | Six-field columns array. | schema_snapshot, endpoint results, and applicable mapping/hash objects. |
| Usage prerequisites | The inspected legacy service checks active status, Worker type mappings, and configuration. Modern loading-purpose checks do not automatically apply. | Also checks source type, loading purpose, and corresponding capabilities/flags. |
| Errors | Some legacy handler parameter/operation errors still return HTTP 200. Check the business code. | Follow the modern API error/capability protocol rather than copying legacy statuses. |

Modern APIs are already covered in the existing Reference. Six legacy entry points are still used by dedicated SDK/CLI clients and require compatibility contracts and mappings. The six routes do not represent six additional modern source capabilities. Without a source deprecation declaration, do not describe them as deprecated or scheduled for removal.

<span id="matrixone目标协议"></span>

## MatrixOne target protocol

/connectors/mo/database/list and /table/list use a saved MatrixOne connector's external target configuration to read actual target metadata. Permissions include export_task.create and connector.use. This is neither the current workspace Catalog database list nor another alias for arbitrary structured sources.

Listing a target does not prove that the target table schema is writable, fields are mapped, or export has completed. Database addresses and credentials come from connector configuration, not resubmission in these two requests.
