请问HN：在构建大型语言模型应用时，你们是如何处理用户上下文的？
我一直在使用大型语言模型（LLMs）构建各种应用，每当我需要用户上下文时，最终都得手动搭建一个上下文管道。
当然，模型可以很好地推理和回答问题，但它对用户是谁、来自哪里或在应用中做了什么一无所知。如果没有这些信息，我要么让模型提出尴尬的初始问题来弄清楚，要么让它猜测，而通常猜测是错误的。
所以我不断重建相同的设置：跟踪事件、丰富会话、总结行为，并将这些信息注入到提示中。
这使得应用变得更加有帮助，但过程非常繁琐。
我希望能有一种简单的方法来获取会话摘要或用户上下文，这样我就可以直接放入提示中。类似于：
```javascript
const context = await getContext();
const response = await generateText({
system: `这是用户上下文：${context}`,
messages: [...]
});
```
以下是我如何使用这个的几个例子：
- 在支持方面，我传入他们查看的文档或他们访问的错误页面。
- 在营销方面，我总结他们的旅程，比如“点击广告” → “阅读博客文章” → “查看定价页面”。
- 在销售方面，我强调一些行为，以判断他们是初创公司还是企业。
- 在产品方面，我将会话分类为“困惑”、“探索方案”或“准备购买”。
- 在推荐方面，我从最近的活动中生成嵌入，并利用这些信息更准确地匹配内容或产品。
在所有这些情况下，我通常会注入一些信息，比如最近的活动、时区、货币、流量来源以及我能收集到的任何信号，以帮助引导用户体验。
还有其他人遇到过同样的问题吗？找到更好的解决方案了吗？
我正在考虑围绕这个问题构建一些东西，最初是为了自己解决这个问题。我很想听听其他人是如何处理的，或者这对你来说是否有用。
查看原文
I've been building stuff with LLMs, and every time I need user context, I end up manually wiring up a context pipeline.<p>Sure, the model can reason and answer questions well, but it has zero idea who the user is, where they came from, or what they've been doing in the app. Without that, I either have to make the model ask awkward initial questions to figure it out or let it guess, which is usually wrong.<p>So I keep rebuilding the same setup: tracking events, enriching sessions, summarizing behavior, and injecting that into prompts.<p>It makes the app way more helpful, but it's a pain.<p>What I wish existed is a simple way to grab a session summary or user context I could just drop into a prompt. Something like:<p>const context = await getContext();<p>const response = await generateText({
system: `Here's the user context: ${context}`,
messages: [...]
});<p>Some examples of how I use this:<p>- For support, I pass in the docs they viewed or the error page they landed on.
- For marketing, I summarize their journey, like 'ad clicked' → 'blog post read' → 'pricing page'.
- For sales, I highlight behavior that suggests whether they're a startup or an enterprise.
- For product, I classify the session as 'confused', 'exploring plans', or 'ready to buy'.
- For recommendations, I generate embeddings from recent activity and use that to match content or products more accurately.<p>In all of these cases, I usually inject things like recent activity, timezone, currency, traffic source, and any signals I can gather that help guide the experience.<p>Has anyone else run into this same issue? Found a better way?<p>I'm considering building something around this initially to solve my problem. I'd love to hear how others are handling it or if this sounds useful to you.