I've written a tool to do this not once but twice now! The first was for ingesting netflow data, the second from ingesting AWS VPC flow logs. In both cases this was storing 'flows' rather than raw packet information.

I'd be super interested in tooling to help with this. I'm running – literally right now – a Python script that is writing about 100k flows per second into Postgres (TimescaleDB). It is batch rather than realtime, and it geocodes the flows on the way in.