Thread Truncation With New Assistants API

Due to the large context window of GPT-4 Turbo combined with the new assistants API, a single message can end up costing a crazy amount if a thread is full of messages. For most uses, I don’t need the full context, just the last ~3 or 4 messages. Unless I missed it, there doesn’t seem to be a way to truncate the thread size through the API. Anyone have any ideas/ways they’ve found to get around this?