what's your strategy for handling duplicate messages in message queues

Hey everyone, I'm stuck on something with our distributed system. I need to figure out how to make sure our message processing is idempotent. You know, when we pull messages from queues like Kafka or RabbitMQ, sometimes they get delivered more than once. We absolutely have to stop our consumers from doing the same work twice. So, what approaches do you guys use to spot and deal with these duplicate messages? I'd love some practical tips and stories from actual experience.