专注于分布式系统架构AI辅助开发工具(Claude
Code中文周刊)
当前位置：Toy Tech Blog 前沿哨所 正文

VS Code Codex Plugin Configuration: A Guide to Customizing Gateway Model List Display

2025-12-19 分类：前沿哨所 阅读(5) 评论(0)
智谱 GLM，支持多语言、多任务推理。从写作到代码生成，从搜索到知识问答，AI 生产力的中国解法。

This article explores how to configure custom gateway model list display in the VS Code Codex plugin. The author attempted to use the [model_provider.xxx] and [profiles.modelname] configuration methods mentioned in GitHub’s official documentation but encountered difficulties in practice. The article details the problems encountered during the configuration process and seeks community assistance. For technical professionals using VS Code for AI model development, understanding how to correctly configure model list information is crucial for improving development efficiency. This article provides reference ideas and discussion directions for solutions to developers facing similar issues.

Original link:Linux.do

赞(0)
未经允许不得转载：Toy Tech Blog » VS Code Codex Plugin Configuration: A Guide to Customizing Gateway Model List Display
分享到
免费、开放、可编程的智能路由方案，让你的服务随时随地在线。

相关推荐

评论 抢沙发

取消

置顶推荐

快讯

  • Claude-Mem 中文版来了：让 AI 终于有了"记忆"

    TL;DR

    Claude Code 原生没有跨会话记忆。Claude-Mem 解决了这个问题。现在，它有中文版了。

    项目地址：https://github.com/cfrs2005/claude-mem

    一、问题：AI 的"金鱼记忆"

    用过 Claude Code 的人都知道一个痛点：

    每次新会话，AI 都是一张白纸。

    昨天你花了两小时教它理解项目架构。今天打开，它全忘了。

    这不是 Bug，是设计。大模型本身没有持久化存储。每次对话都是独立的。

    就像和一个失忆症患者合作——能力很强，但每天都要重新自我介绍。

    二、方案：给 AI 配一本日记

    Claude-Mem 的思路很简单：

    既然 AI 没有记忆，就给它造一个。

    它做了三件事：

    1. 自动记录：每次对话的关键信息，自动存入本地数据库
    2. 智能压缩：不是存原文，而是提取要点，节省空间
    3. 按需召回：新会话开始时，自动加载相关上下文

    类比：给 AI 配了一本日记本。每天写日记，需要时翻阅。

    Claude-Mem 架构图

    三、为什么要中文版

    原版 Claude-Mem 是英文的。

    对于中文开发者来说，有几个问题：

    1. 提示词是英文：AI 的记忆摘要、搜索结果都是英文输出
    2. 文档是英文：安装配置要啃英文 README
    3. 调试信息是英文：出问题时看日志很痛苦

    中文版解决了这些：

    • 所有提示词汉化，记忆摘要用中文生成
    • 完整中文文档
    • 中文日志输出

    对中文用户更友好。

    四、安装指南

    4.1 前置条件

    • Node.js 18.0.0+
    • Claude Code 最新版
    • SQLite 3

    4.2 安装步骤

    第一步：安装原版插件

    在 Claude Code 中执行：

    /plugin marketplace add thedotmack/claude-mem
/plugin install claude-mem

    重启 Claude Code。

    第二步：克隆中文版

    git clone https://github.com/cfrs2005/claude-mem.git
cd claude-mem
npm install

    第三步：编译并同步

    npm run build-and-sync

    这一条命令完成：编译 → 同步到插件目录 → 重启服务。

    完成。

    五、使用场景

    场景一：项目上下文保持

    你在开发一个复杂项目。

    没有 Claude-Mem：每次都要重新解释项目结构、技术栈、代码规范。

    有了 Claude-Mem：AI 记得你的项目，直接进入工作状态。

    场景二：知识积累

    你教会 AI 一些特定领域知识。

    没有 Claude-Mem：下次全忘。

    有了 Claude-Mem：知识持久化，越用越聪明。

    场景三：工作流延续

    昨天调试到一半，今天继续。

    没有 Claude-Mem：从头开始。

    有了 Claude-Mem：AI 知道昨天做到哪了。

    六、技术原理

    Claude-Mem 的核心是一个本地 SQLite 数据库。

    写入流程： 1. 对话结束时，触发记忆压缩 2. 提取关键信息（决策、发现、错误、代码变更） 3. 生成结构化摘要 4. 存入本地数据库

    读取流程： 1. 新会话开始 2. 根据当前项目/上下文，检索相关记忆 3. 注入到系统提示词 4. AI 获得"记忆"

    隐私保障： - 所有数据存在本地 - 不上传任何服务器 - 你完全控制自己的数据

    七、与原版的区别

    特性 原版 中文版
    提示词语言 英文 中文
    文档语言 英文 中文
    日志输出 英文 中文
    记忆摘要 英文生成 中文生成
    核心功能 完整 完整

    功能完全一致，只是语言本地化。

    八、常见问题

    Q：会影响 Claude Code 性能吗？

    A：影响很小。记忆压缩是异步的，不阻塞主流程。

    Q：数据存在哪里？

    A：本地 SQLite 文件，路径可配置。

    Q：能和原版共存吗？

    A：不建议。中文版会覆盖原版的提示词文件。

    Q：支持哪些操作系统？

    A：macOS、Linux、Windows（WSL）。

    九、小结

    Claude-Mem 解决了 Claude Code 的记忆问题。

    中文版让中文用户用得更顺手。

    如果你： - 经常用 Claude Code 开发 - 受够了每次重新解释项目 - 希望 AI 能"记住"你教它的东西

    试试这个项目。

    GitHub：https://github.com/cfrs2005/claude-mem

    参考链接

    • 原版项目：https://github.com/thedotmack/claude-mem
    • Claude Code 官方文档：https://docs.anthropic.com/claude-code

  • One Year of AI Product Development: Lessons and Insights from Four Products

    The author reflects on their journey developing four AI products in one year, from Readecho in 2024 to Wentutu, PrompterHub, and Heartsticker in 2025, sharing ten key reflections. These insights include matching product form with use cases, balancing technical learning with business goals, prioritizing operations over features, focusing on core value, considering commercialization from the start, aligning feature design with operations, driving business value through growth, how platform choice affects product potential, the importance of building a user community, and the symbiotic relationship between personal branding and product growth. These experiences offer practical product development strategies for indie developers and entrepreneurs, emphasizing how to avoid common pitfalls and achieve sustainable product growth in the AI era through hands-on experience.

    Original Link:V2EX Share & Discover

  • GPT-5.2-Codex Powers Handwriting Practice Website Development with Impressive Results

    Recently, a developer successfully created a handwriting practice website using the latest GPT-5.2-Codex update. The author had previously attempted to debug using Gemini 3 and Claude 4.5 with unsatisfactory results, but GPT-5.2-Codex delivered impressive outcomes. The website is now live at https://lianzi.wangwangit.com, and the author invites readers to try it out and share their experiences. This case demonstrates the practical application of AI models in website development, offering valuable insights for tech enthusiasts and highlighting the practical value of cutting-edge AI technology.

    Original link:V2EX Share & Discover

  • Qoder+Obsidian: A Practical Solution to Boost Local Knowledge Base Search Efficiency

    This article introduces a practical method to enhance the search efficiency of Obsidian knowledge bases. As a popular knowledge management tool, Obsidian's native search functionality often fails to meet user needs. By combining with Qoder's Agent service, users can achieve more precise searches within their local knowledge base. The article provides detailed operational steps: simply open Qoder in your Obsidian storage folder to quickly find the content you need. Additionally, the article mentions that Qoder is currently running a credit giveaway event, offering users an opportunity to try this solution. This combined approach is particularly suitable for users who need to efficiently manage large numbers of Markdown documents, significantly improving knowledge retrieval efficiency.

    Original Link:V2EX Share & Discover

  • AI-Powered Chrome Extension Removes Gemini Watermarks

    This article introduces an AI-developed Chrome extension specifically designed to remove the star watermarks from Gemini-generated content. The tool processes images using local computation, requiring no internet connection or reliance on AI models, ensuring user privacy and security. The author has optimized the algorithm to solve the poor performance issues of traditional methods, and the extension is now available on the Chrome Web Store for users to try directly. The extension provides an efficient and convenient solution for Gemini-style visible watermarks, suitable for scenarios that require cleaning up AI-generated content. The article also mentions related discussion topics, encouraging users to further exchange technical details. This tool demonstrates the innovative value of AI in cutting-edge technology applications, providing a practical utility for handling AI-generated content.

    Original Link:Linux.do

  • Complete Guide to Building a WeChat Bot with AstrBot + WeChatPadPro

    This article provides a detailed guide on how to build a WeChat bot by combining AstrBot and WeChatPadPro, supporting integration with large model services like OpenAI and DeepSeek. Starting from deploying AstrBot, to configuring the WeChatPadPro environment, and then adding the bot connection, it offers complete operational steps. Solutions for WeChat login security verification issues are also provided. Suitable for developers and tech enthusiasts to learn and quickly build their own AI-driven WeChat bot.

    Original Link:Linux.do

最新评论

热门标签

AI(540)人工智能(283)Gemini(266)claude(173)GitHub(141)google(81)Chatgpt(72)openai(69)AI工具(65)article(65)大模型(63)谷歌(62)开源项目(54)开源工具(53)model(52)AI模型(52)大语言模型(49)tool(48)agent(46)code(45)开源(42)users(40)cursor(39)提示词(38)deepseek(38)Guide(37)Claude Code(37)development(35)api(34)Antigravity(34)

十年稳如初 — LocVPS，用时间证明实力

10+ 年老牌云主机服务商，全球机房覆盖，性能稳定、价格厚道。

老品牌，更懂稳定的价值你的第一台云服务器，从 LocVPS 开始