用于构建智能客服Agent，使用大语言模型回访用户并记录语音，对回访中的同类问题进行分析，按高回复比例进行replay。提示词将记录与回放（Record & Replay）技术应用于智能客服，以“经验”赋能Agent，提升可靠性、效率与用户体验。

中文版提示词

请构建一个智能客服Agent：使用大语言模型回访用户，并对语音进行记录；对回访中出现的同样问题，由大模型进行分析，按照高回复比例进行replay。用“经验”赋能Agent——将系统的记录与回放（Record & Replay，R&R）技术应用于智能客服Agent，通过记录Agent与用户之间的交互过程以及系统内部状态，并在需要时精确重现这些交互，从而解决大语言模型驱动的AI Agent所面临的不确定性、高计算资源需求、隐私问题和性能限制等固有挑战，显著提升智能客服系统的可靠性、效率和用户体验。

英文版提示词

Build an intelligent customer-service Agent: use a large language model to call back users and record the voice; for recurring questions during callbacks, the LLM analyzes them and replays according to the highest-response ratio. Empower the Agent with "experience"—apply the Record & Replay (R&R) technique to the customer-service Agent, recording the interactions between the Agent and users as well as the internal system state, and accurately reproducing these interactions when needed. This addresses the inherent challenges of LLM-driven AI agents—uncertainty, high computational cost, privacy concerns, and performance limits—significantly improving the reliability, efficiency, and user experience of the customer-service system.

🛠️ **适用 AI 工具**：ChatGPT、Claude、DeepSeek、Kimi、通义千问、Gemini