A common partner-integration shape: a CSV lands on an SFTP server on a schedule, and its rows have to become Salesforce records without duplicating anything when the job runs twice.
A scheduler triggers the flow. The SFTP connector lists the inbound directory and reads each new file. A FlowScript transform parses the CSV and maps columns onto Salesforce fields. The Salesforce connector upserts against an external ID field, which is what makes a re-run safe — matching records update instead of duplicating. Processed files move to a separate directory so the next run ignores them.
A Spring Boot application with a scheduled entry point, an SFTP client built on sshj, the Salesforce connector configured for JWT Bearer authentication, and retry with backoff around both remote calls. It runs on your infrastructure with no dependency on FlowSetu.
Idempotency through upsert rather than insert, moving files only after a successful load, and deciding what happens to a row Salesforce rejects — failing the whole file is rarely what anyone wants, and silently dropping the row is worse.