CodeCortex – 面向 AI 编码助手的持久知识图谱存储库
1 分•作者: costea•5 个月前
我启动了一个名为 CodeCortex 的开源实验。<p>这个想法很简单：<p>AI 编码工具在每次会话中都会反复重新学习相同的代码库结构。<p>这会浪费 token，并且经常遗漏架构依赖关系。<p>CodeCortex 探索了一种不同的方法：构建一个代码库的持久知识图谱，并随着代码库的演进而不断演进。<p>AI 代理不再重新扫描代码库，而是查询这个图谱。<p>项目还处于早期阶段，但代码库现已公开。<p>很想听听使用 AI 编码工具进行开发的各位的想法。<p>https://github.com/costeamarius/codecortex
查看原文
I’ve started an open-source experiment called CodeCortex.<p>The idea is simple:<p>AI coding tools repeatedly re-learn the same repository structure every session.<p>This wastes tokens and often misses architectural dependencies.<p>CodeCortex explores a different approach:
build a persistent knowledge graph of the repository that evolves with the codebase.<p>Instead of rescanning the repo, AI agents query the graph.<p>Still early, but the repo is now public.<p>Curious to hear thoughts from people building with AI coding tools.<p>https://github.com/costeamarius/codecortex