Ask HN:各位正在构建 AI 智能体的开发者,你们是如何提升其运行速度的?

1作者: arkmm6 个月前
由于需要跨多个系统进行协调以及链式调用大型语言模型(LLM),如今许多智能体的使用体验都可能感觉非常缓慢。我很想知道大家是如何解决这个问题的: * 你们都是如何识别智能体中的性能瓶颈的? * 哪些类型的更改为你们带来了最大的速度提升? 就我们而言,我们编写了一个用于分析性能的工具,以便识别缓慢的 LLM 调用 - 有时我们可以为该步骤换用更快的模型,或者我们意识到可以通过消除不必要的上下文来减少输入 token 数量。对于需要外部访问的步骤(浏览器使用、API 调用),我们已经转移到快速启动的外部容器 + 线程池以实现并行化。我们还尝试了一些 UI 更改来掩盖一些延迟。 大家还在使用哪些其他的性能增强技术?
查看原文
Because of the coordination across multiple systems + chaining LLM calls, a lot of agents today can feel really slow. I would love to know how others are tackling this:<p>- How are you all identifying performance bottlenecks in agents?<p>- What types of changes have gotten you the biggest speedups?<p>For us we vibe-coded a profiler to identify slow LLM calls - sometimes we could then switch out a faster model for that step or we&#x27;d realize we could shrink the input tokens by eliminating unnecessary context. For steps requiring external access (browser usage, API calls), we&#x27;ve moved to fast start external containers + thread pools for parallelization. We&#x27;ve also experimented some with UI changes to mask some of the latency.<p>What other performance enhancing techniques are people using?