HackerNews中文版

嘿，HN，我开发了 CodeContext API（https://codecontext-api-production.up.railway.app），旨在解决我在构建 AI 编码代理时遇到的一个问题：知识截止日期。我的代理因为训练数据过时而不断产生虚假的 API，比如 LangChain 或 Next.js。手动抓取文档网站的每个请求既慢又昂贵（令牌费用累积很快！）。 CodeContext API 是一个针对 1000 多个流行库文档的语义搜索层。输入：{"query": "react hooks"} 输出：干净的 JSON，包含最相关的代码片段和解释。延迟：亚秒级（针对 RAG 进行了优化）。为什么要使用这个？节省令牌：不要将整个手册都放入上下文中。只获取你需要的内容。准确性：为你的代理提供最新版本的实际文档。速度：无需构建或维护自己的抓取工具。如果你想测试延迟，主页上有一个实时演示（无需注册）。我很想听听你们对 API 结构的反馈，以及你们希望看到哪些其他库被索引！

查看原文

Hey HN,I built CodeContext API (https://codecontext-api-production.up.railway.app) to solve a problem I kept hitting while building AI coding agents: The Knowledge Cutoff.My agents kept hallucinating APIs for libraries like LangChain or Next.js because their training data was stale. Scraping documentation sites manually for every request was slow and expensive (tokens add up!).CodeContext API is a semantic search layer over the documentation of 1000+ popular libraries.Input: {"query": "react hooks"} Output: Clean JSON with the most relevant code snippets and explanations. Latency: Sub-second (optimized for RAG). Why use this?Save Tokens: Don't dump the whole manual into the context. Fetch only what you need. Accuracy: Give your agent the actual docs from the latest version. Speed: No need to build/maintain your own scrapers. There is a Live Demo on the homepage (no signup required) if you want to test the latency.I'd love your feedback on the API structure and what other libraries you'd like to see indexed!

我构建了一个API，使大型语言模型（LLMs）能够即时访问1000个库的文档。