P S.

That’s why we’re feeding the CSV data to the consumer. For now, I am using the same workaround suggested by James H. i.e. to read the CSV in a parent pipeline and send the offsets one at a time to the child pipeline to handle the read.

Patrick T. Agreed, Kafka shouldn't be treated as a database. We normally rely on group IDs, but in this case, the downstream target specifically requested re-processing for a select list of offsets.

Got it, understood 👍 In my first HTTP Client (where I use the GitLab API), the response doesn’t return a location or filename. So I populated the $['content-location'] value from pipeline parameters, and it worked even though that value isn’t an actual upload location. Thanks Jocelyn A. for the quick and prompt response!

Jocelyn A. Thanks, the above method worked like a charm. Please correct me if I’m wrong, but my understanding is that the $['content-location'] field can have any value—it’s required mainly because the actual file content is transferred via the binary output view of the previous Snap.