用确定性压缩对话上下文，降低 token 消耗
Token compression for AI contexts, reducing token consumption by compressing conversation exchanges before they enter the LLM context window.
