1.4 现有方案为什么不够

既然上下文这么重要，那有没有现成的方案？有，但都不够。

Git 管代码版本。 Git 记录的是"谁在什么时候改了什么"。它回答的是"怎么变成这样的"这个问题，而不是"现在 AI 需要知道什么"。你可以从 Git 历史里推断上下文，但那是人的工作，不是 AI 能高效做的事情。

Obsidian 管个人知识。 Obsidian 是"人学了什么"的整理。它按人的认知结构组织，而不是按 AI 的任务需求组织。一个笔记可能包含多个主题的混合，一个任务可能需要跨多个笔记的信息。知识点和任务上下文之间，有一层转换工作要做。

AI 聊天记录管对话。 聊天记录记录了曾经的问答，但问答是线性的、有始有终的。一个项目有多个方向、多个阶段、多个并行任务，全都混在一条对话流里，不可分割，不可复用。

Prompt 模板管指令。 很多人写 Prompt 模板，试图把上下文固化到模板里。但模板是静态的，项目是动态的。今天加了新功能，明天改了技术选型，模板跟不上变化。而且模板越长，维护成本越高。

这几个方案各管一个维度，但没有一个管"AI 在当前任务中需要知道什么"这个维度。

这不是谁的错。因为这些工具在设计时，AI 还不是协作对象，它们的用户是人。人可以根据上下文碎片自己拼出全貌，AI 不行。

那需要什么？

需要一个专门为 AI 设计的上下文管理系统。它应该：

  • 以 AI 为第一读者，而不是人
  • 按任务组织，而不是按文件或个人知识
  • 结构化的，可被 AI 路由和索引的
  • 可持久化、可版本化、可跨会话复用的
  • 模型无关、工具无关的

听起来很复杂？其实很简单。一个目录结构，几份约定文件，一套工作流，就够了。这就是接下来要讲的 AI Context Workspace。

Comments

Post a Comment

Popular posts from this blog

How to turn off Sass warning prompts in Nuxt.js projects

Background of the Issue When developing projects with Nuxt.js, it’s common to encounter a large number of Sass-related warning messages in the console. These warnings mainly include: Warnings about the order of mixin declarations. Warnings about the use of Legacy JS API. Although these warnings do not affect the operation of the program, they can degrade the development experience by cluttering the console. Today, we will explore how to elegantly address this issue. Solutions 1. Disable Warnings via Configuration In the nuxt.config.js file, we can modify the Sass configuration to suppress these warning messages: export default { build: { loaders: { scss: { sassOptions: { quietDeps: true , // Suppress warnings from dependencies logger: { warn: function ( message ) { return null ; // Silence warning messages }, debug: function ( message ) { return null ; // Silence ...
Read more

Configuring SSH Access to a Docker Container via an Alternative Port

Configuring SSH Access to a Docker Container via an Alternative Port If you want to configure SSH access to a Docker container using an alternative port, such as port 6666 , follow the steps below carefully. This guide assumes that you have administrative access to your Docker host and are familiar with basic Linux commands. Step 1: Run the Container with SSH Service Start by running a new container from your image with port 6666 on the host mapped to port 22 inside the container: sudo docker run -it -p 6666 : 22 --name ssh-container testimage Make sure to replace testimage with the actual name of your Docker image. The --name option gives the container a specific name for easier management. Step 2: Install OpenSSH Server in the Container Once inside the container, update the package list and install the OpenSSH server: apt-get update && apt-get install -y openssh-server Step 3: Set the Root Password Set a password for the root user, which will be required when l...
Read more

Quickly Set Up a Cloud Database Using MongoDB Atlas

Quickly Set Up a Cloud Database Using MongoDB Atlas MongoDB Atlas is the official fully managed cloud database service from MongoDB, simplifying the deployment, management, and scaling of MongoDB in the cloud. This article will guide you on how to quickly create a MongoDB Atlas cluster and set up basic connection configurations. Creating a MongoDB Atlas Cluster Sign Up & Log In If you do not have a MongoDB Atlas account yet, head over to MongoDB Atlas to sign up. After logging in, click on “Build a Cluster” to start creating a new cluster. Select Cluster Configuration To meet the needs of most development and testing scenarios, you can choose the free tier (M0) shared cluster, which is usually sufficient for small applications or personal projects. When choosing a cloud provider, select based on your preference. Here we chose Azure, but AWS and GCP are also great choices. Specify Geographic Location For geographic location selection, choose according to your u...
Read more