📄 中文摘要
大多数人工智能聊天应用程序都默默假设一个模型可以满足所有需求，但这种想法并不成立。在构建Chymera时，目标是让系统能够识别用户提问的类型，并将其发送到最适合回答的模型，而不是将用户锁定在单一的语言模型中。文章讲述了构建这一路由层的过程，包括初次尝试中的错误以及最终有效版本的实现。通过这种方式，用户可以更流畅地与系统互动，而无需考虑使用哪个模型。该方法旨在提高聊天机器人的响应准确性和用户体验。
如何构建意图分类器以在多个大型语言模型之间路由消息
出处: How I Built an Intent Classifier to Route Messages Across Multiple LLMs
发布: 2026年3月27日
📄 English Summary
How I Built an Intent Classifier to Route Messages Across Multiple LLMs
Most AI chat applications operate under the flawed assumption that a single model is sufficient for all tasks. The author aimed to address this issue while building Chymera by creating a system that identifies the type of question being asked and routes it to the most suitable model, rather than locking users into one language model. The article details the development of this routing layer, including initial mistakes and the characteristics of the final working version. This approach allows users to interact with the system more seamlessly without needing to consider which model to use, ultimately enhancing the accuracy of responses and user experience.