Reusable Connections keep provider credentials out of Job payloads and let files move server-to-server without passing through your application. External imports are normalized into a ConverterHQ-owned File before conversion; outputs can then be delivered to any supported provider.
Connections
Connection secrets are encrypted at rest and never returned by the API. Use POST /v1/connections/{id}/test before production workflows and rotate or revoke a Connection instead of copying credentials into saved Jobs.
Canonical input staging
import/s3, import/gcs, import/azure and import/sftp first fetch the source and persist it as an owned ConverterHQ File. Downstream conversion Tasks read that owned File, giving retries, retention, checksums and ownership one consistent boundary regardless of the source provider.
Multi-cloud outputs
Conversion outputs are ordinary ConverterHQ Files and may be exported with export/s3, export/gcs, export/azure or export/sftp. Input and output providers do not need to match, so an S3-origin workflow can deliver to GCS, Azure Blob or SFTP and vice versa.
S3 security
For S3 Connections, prefer AssumeRole with ExternalId to long-lived access keys. Scope the role to only the buckets/prefixes ConverterHQ needs and separate read/write prefixes when your workflow permits it.
Import/export Tasks
The DAG contract stays consistent across providers. Imports produce Files; exports consume Files. External export attempts are checkpointed so an uncertain provider side effect is not silently repeated unless the retry mode explicitly allows replacement.
Failure handling
Cloud authentication failures, missing objects and permission errors are deterministic until the external state changes. Inspect the Task error before retrying, test the Connection independently, and avoid retry storms against a provider that is denying access.