You are building a lightweight chat client with the Foundry SDK for a deployed language model. During testing, users complain that the assistant forgets what was said earlier in the same conversation and treats every question as if it were the first. Which change to your client code should you make so the model can maintain the flow of the conversation?