TE
科技回声
首页24小时热榜最新最佳问答展示工作
GitHubTwitter
首页

科技回声

基于 Next.js 构建的科技新闻平台,提供全球科技新闻和讨论内容。

GitHubTwitter

首页

首页最新最佳问答展示工作

资源链接

HackerNews API原版 HackerNewsNext.js

© 2025 科技回声. 版权所有。

Show HN: I created a chatbot trained on Bear Blog's docs

5 点作者 HermanMartinus大约 2 年前

2 条评论

HermanMartinus大约 2 年前
I get a decent number of emails every week that could easily be solved by users reading the docs. The questions generally revolve around DNS record setup, which can be complicated for the average user.<p>I&#x27;ve tried my best to make the docs as easy to read as possible, but as Paul Graham said, you can bet on laziness, and it&#x27;s much easier to send me an email.<p>Therefore I&#x27;ve trained a chatbot on Bear&#x27;s documentation. It&#x27;s currently using GPT 3.5 since 4 is 29x more expensive and 3 does a pretty good job at parsing information if that information is structured well. The only downside is that it has the ability to go off topic.<p>One thing to note, the Bear docs are super simple and once compressed all fits within the 1000 token context of GPT3.
oulipo大约 2 年前
Nice! Have you tried using Markprompt too?