All-In Copilot

English

🚀 SDK & CLI for building VS Code Chat extensions with custom LLM providers.

✨ Features

One-click CLI Generation - Interactively create extension projects
Multi-provider Presets - Built-in configurations for GLM, DeepSeek, Qwen, MiniMax, OpenAI, Anthropic
Dynamic Model Fetching - Automatically fetch available models from API
Ready-to-use Templates - Copy templates to quickly create custom Copilot extensions
Lightweight SDK - Core SDK has no VS Code dependencies, works in any Node.js environment

Architecture

all-in-copilot/
├── packages/
│ └── sdk/ # Core SDK
│ └── src/
│ ├── core/ # Types, model fetcher
│ └── vscode/ # VS Code provider helpers
│
├── templates/ # Extension templates
│ ├── base-template/ # Base template for custom providers
│ ├── glm-template/ # GLM (智谱AI) example
│ ├── minimax-template/ # MiniMax example
│ ├── kimi-template/ # Kimi K2 (Moonshot) example
│ ├── mimo-template/ # Xiaomi MiMo example
│ └── aliyun-coding-template/ # Aliyun Model Studio Coding Plan example
│
└── cli/ # Project generator CLI
└── src/
└── index.ts

Quick Start

Method 1: Use CLI (Recommended)

# Install CLI globally
npm install -g @all-in-copilot/cli

# Create a new project interactively
all-in-copilot

# Or with short command
aic create my-copilot

Method 2: Copy Template

# Copy template
cp -r templates/glm-template my-copilot
cd my-copilot

# Edit configuration
vim src/config.ts

# Install dependencies
npm install

# Compile and test (F5 in VS Code)
npm run compile

CLI Commands

all-in-copilot              # Interactive mode
all-in-copilot create NAME  # Create project with prompts
all-in-copilot list         # List available presets
all-in-copilot help         # Show help

Available Presets

Preset	Provider	API Format
glm	GLM (智谱AI)	OpenAI
minimax	MiniMax	Anthropic
kimi	Kimi (Moonshot)	Anthropic
mimo	Xiaomi MiMo	Anthropic
deepseek	DeepSeek	OpenAI
qwen	Qwen	OpenAI
openai	OpenAI	OpenAI
anthropic	Anthropic	Anthropic
custom	Custom	OpenAI

Provider Configuration

Edit src/config.ts to customize your provider:

export const PROVIDER_CONFIG: ProviderConfig = {
	id: "provider-id",
	name: "Provider Name",
	baseUrl: "https://api.example.com/v1/chat/completions",
	apiKeySecret: "extension-name.apiKey",
	family: "provider-family",
	apiMode: "openai", // 'openai' | 'anthropic' | 'gemini' | 'ollama'
	supportsTools: true,
	supportsVision: false,
	defaultMaxOutputTokens: 4096,
	defaultContextLength: 32768,
	dynamicModels: true,
	modelsCacheTTL: 5 * 60 * 1000,
};

export const FALLBACK_MODELS: ModelConfig[] = [
	{
		id: "model-1",
		name: "Model 1",
		maxInputTokens: 30000,
		maxOutputTokens: 4096,
		supportsTools: true,
		supportsVision: false,
	},
];

// Optional: Filter which models to display
export function filterModels(models: ModelConfig[]): ModelConfig[] {
	return models.filter((m) => m.id.includes("chat"));
}

SDK Usage

import {
	convertToOpenAI,
	convertToolsToOpenAI,
	processOpenAIStream,
	fetchModelsFromAPI,
	estimateTokens,
} from "@all-in-copilot/sdk";

// Dynamic model fetching
const providerConfig: ProviderConfig = {
	id: "my-provider",
	name: "My Provider",
	baseUrl: "https://api.example.com/v1",
	// ... other config
};

const models = await fetchModels(providerConfig, { apiKey: "your-api-key" });
console.log("Available models:", models);

// Use OpenAI-compatible provider
const provider = new OpenAIProvider(providerConfig, { apiKey: "your-api-key" });
for (const model of models) {
	provider.registerModel(model);
}

const response = await provider.complete({
	model: models[0].id,
	messages: [{ role: "user", content: "Hello!" }],
});

Development

Build SDK

cd packages/sdk
npm run build

Build VS Code Extension

cd templates/minimax-template
npm run compile

Watch Mode

# Terminal 1
cd packages/sdk && npm run watch

# Terminal 2
cd templates/minimax-template && npm run watch

CI/CD & Publishing

GitHub Secrets Required

Add the following secrets in your repository settings:

Secret	Description
`VSCE_PAT`	VS Code Marketplace Personal Access Token
`OVSX_PAT`	Open VSX Registry Token (optional)

Workflows

Workflow	Trigger	Description
CI	Push/PR to `main`	Build SDK, CLI and all templates, run tests
Release	Tag push (`v*`) / Manual	Publish stable or pre-release to marketplace
Pre-release	Push to `pre-release/*` / Manual	Publish pre-release version
Publish Single	Manual only	Manually publish a single extension

Release Process

Stable Release:

# 1. Update version
cd templates/glm-template
npm version patch  # or minor / major

# 2. Commit and create tag
git add .
git commit -m "chore: bump glm-template to v0.2.0"
git tag glm-template-v0.2.0
git push origin main --tags

Pre-release:

# Method 1: Use tag
git tag glm-template-v0.2.0-beta.1
git push origin --tags

# Method 2: Use branch
git checkout -b pre-release/glm-template
git push origin pre-release/glm-template

# Method 3: Manual trigger via GitHub Actions

Tag Naming Convention

Pattern	Triggers	Example
`v*`	Release all templates	`v1.0.0`
`<template>-v*`	Release specific template	`glm-template-v0.2.0`
`-pre`, `-alpha`, `-beta`, `-rc`	Pre-release	`v1.0.0-beta.1`

Templates

Template	Base URL	API Mode	Dynamic Models
GLM	`https://open.bigmodel.cn/api/paas/v4/chat/completions`	OpenAI	✅
MiniMax	`https://api.minimaxi.com/anthropic/v1/messages`	Anthropic	❌
Kimi	`https://api.moonshot.cn/anthropic`	Anthropic	❌
MiMo	`https://api.xiaomimimo.com/anthropic/v1/messages`	Anthropic	❌
Aliyun Coding Plan	`https://coding.dashscope.aliyuncs.com/apps/anthropic/v1/messages`	Anthropic	❌
Base	Custom	OpenAI/Anthropic	✅

How Dynamic Model Fetching Works

Provider API (/models)
       ↓
   fetch + cache
       ↓
 filterModels()
       ↓
VS Code Model List

Extension calls /models endpoint on provider's API
Response is parsed and converted to ModelConfig[]
Models are cached for modelsCacheTTL milliseconds
If fetch fails, fallback to FALLBACK_MODELS
Optional filterModels() function filters displayed models

中文

🚀 用于构建带有自定义 LLM 提供商的 VS Code Chat 扩展的 SDK 和 CLI。

✨ 功能特性

CLI 一键生成 - 交互式创建扩展项目
多供应商预设 - 内置 GLM、DeepSeek、Qwen、MiniMax、OpenAI、Anthropic 配置
动态模型获取 - 自动从 API 获取可用模型列表
即用型模板 - 复制模板即可快速创建自定义 Copilot 扩展
轻量 SDK - 核心 SDK 无 VS Code 依赖，可在任何 Node.js 环境使用

架构

all-in-copilot/
├── packages/
│ └── sdk/ # 核心 SDK
│ └── src/
│ ├── core/ # 类型定义、模型获取
│ └── vscode/ # VS Code 提供者助手
│
├── templates/ # 扩展模板
│ ├── base-template/ # 自定义提供者的基础模板
│ ├── glm-template/ # GLM (智谱AI) 示例
│ ├── minimax-template/ # MiniMax 示例
│ ├── kimi-template/ # Kimi K2 (月之暗面) 示例
│ ├── mimo-template/ # 小米 MiMo 示例
│ └── aliyun-coding-template/ # 阿里云百炼编程计划示例
│
└── cli/ # 项目生成 CLI
└── src/
└── index.ts

快速开始

方法 1：使用 CLI（推荐）

# 全局安装 CLI
npm install -g @all-in-copilot/cli

# 交互式创建新项目
all-in-copilot

# 或使用简短命令
aic create my-copilot

方法 2：复制模板

# 复制模板
cp -r templates/glm-template my-copilot
cd my-copilot

# 编辑配置
vim src/config.ts

# 安装依赖
npm install

# 编译并测试（在 VS Code 中按 F5）
npm run compile

CLI 命令

all-in-copilot              # 交互模式
all-in-copilot create NAME  # 使用提示创建项目
all-in-copilot list         # 列出可用预设
all-in-copilot help         # 显示帮助

可用预设

预设	提供商	API 格式
glm	GLM (智谱AI)	OpenAI
minimax	MiniMax	Anthropic
kimi	Kimi (月之暗面)	Anthropic
mimo	小米 MiMo	Anthropic
deepseek	DeepSeek	OpenAI
qwen	Qwen (通义千问)	OpenAI
openai	OpenAI	OpenAI
anthropic	Anthropic	Anthropic
custom	自定义	OpenAI

提供者配置

编辑 src/config.ts 来自定义你的提供者：

export const PROVIDER_CONFIG: ProviderConfig = {
	id: "provider-id", // 提供者 ID
	name: "Provider Name", // 显示名称
	baseUrl: "https://api.example.com/v1/chat/completions", // API 地址
	apiKeySecret: "extension-name.apiKey", // API 密钥存储键
	family: "provider-family", // 模型系列
	apiMode: "openai", // API 模式：'openai' | 'anthropic' | 'gemini' | 'ollama'
	supportsTools: true, // 是否支持工具调用
	supportsVision: false, // 是否支持图像
	defaultMaxOutputTokens: 4096,
	defaultContextLength: 32768,
	dynamicModels: true, // 是否动态获取模型
	modelsCacheTTL: 5 * 60 * 1000, // 模型缓存时间
};

export const FALLBACK_MODELS: ModelConfig[] = [
	{
		id: "model-1",
		name: "Model 1",
		maxInputTokens: 30000,
		maxOutputTokens: 4096,
		supportsTools: true,
		supportsVision: false,
	},
];

// 可选：过滤要显示的模型
export function filterModels(models: ModelConfig[]): ModelConfig[] {
	return models.filter((m) => m.id.includes("chat"));
}

SDK 使用

import {
	convertToOpenAI,
	convertToolsToOpenAI,
	processOpenAIStream,
	fetchModelsFromAPI,
	estimateTokens,
} from "@all-in-copilot/sdk";

// 动态模型获取
const providerConfig: ProviderConfig = {
	id: "my-provider",
	name: "My Provider",
	baseUrl: "https://api.example.com/v1",
	// ... 其他配置
};

const models = await fetchModels(providerConfig, { apiKey: "your-api-key" });
console.log("可用模型:", models);

// 使用 OpenAI 兼容的提供者
const provider = new OpenAIProvider(providerConfig, { apiKey: "your-api-key" });
for (const model of models) {
	provider.registerModel(model);
}

const response = await provider.complete({
	model: models[0].id,
	messages: [{ role: "user", content: "Hello!" }],
});

开发

构建 SDK

cd packages/sdk
npm run build

构建 VS Code 扩展

cd templates/minimax-template
npm run compile

监视模式

# 终端 1
cd packages/sdk && npm run watch

# 终端 2
cd templates/minimax-template && npm run watch

CI/CD 和发布

需要配置的 GitHub Secrets

在仓库设置中添加以下 Secrets：

Secret	描述
`VSCE_PAT`	VS Code 插件市场个人访问令牌
`OVSX_PAT`	Open VSX 注册表令牌（可选）

工作流

工作流	触发条件	描述
CI	Push/PR 到 `main`	构建 SDK、CLI 和所有模板，运行测试
Release	Tag 推送 (`v*`) / 手动	发布正式版或预发布版到插件市场
Pre-release	Push 到 `pre-release/*` / 手动	发布预发布版本
Publish Single	仅手动	手动发布单个插件

发布流程

正式发布：

# 1. 更新版本号
cd templates/glm-template
npm version patch  # 或 minor / major

# 2. 提交并创建 tag
git add .
git commit -m "chore: bump glm-template to v0.2.0"
git tag glm-template-v0.2.0
git push origin main --tags

预发布：

# 方法 1：使用 Tag
git tag glm-template-v0.2.0-beta.1
git push origin --tags

# 方法 2：使用分支
git checkout -b pre-release/glm-template
git push origin pre-release/glm-template

# 方法 3：通过 GitHub Actions 手动触发

Tag 命名规范

模式	触发	示例
`v*`	发布所有模板	`v1.0.0`
`<template>-v*`	发布指定模板	`glm-template-v0.2.0`
`-pre`, `-alpha`, `-beta`, `-rc`	预发布	`v1.0.0-beta.1`

模板

模板	Base URL	API 模式	动态模型
GLM	`https://open.bigmodel.cn/api/paas/v4/chat/completions`	OpenAI	✅
MiniMax	`https://api.minimaxi.com/anthropic/v1/messages`	Anthropic	❌
Kimi	`https://api.moonshot.cn/anthropic`	Anthropic	❌
MiMo	`https://api.xiaomimimo.com/anthropic/v1/messages`	Anthropic	❌
Aliyun Coding Plan	`https://coding.dashscope.aliyuncs.com/apps/anthropic/v1/messages`	Anthropic	❌
Base	自定义	OpenAI/Anthropic	✅

动态模型获取工作原理

提供者 API (/models)
       ↓
   获取 + 缓存
       ↓
 filterModels()
       ↓
VS Code 模型列表

扩展调用提供者 API 的 /models 端点
响应被解析并转换为 ModelConfig[]
模型被缓存 modelsCacheTTL 毫秒
如果获取失败，回退到 FALLBACK_MODELS
可选的 filterModels() 函数过滤要显示的模型

Acknowledgments / 致谢

This project is inspired by and references the following excellent open source projects:

本项目受到以下优秀开源项目的启发和参考：

Core References / 核心参考

Project	Description
vscode-copilot-chat	Microsoft's official GitHub Copilot Chat extension - the authoritative reference for VS Code Chat API usage / 微软官方 GitHub Copilot Chat 扩展 - VS Code Chat API 使用的权威参考
vscode	Visual Studio Code source code - for understanding VS Code extension APIs / VS Code 源代码 - 用于理解 VS Code 扩展 API

Community Projects / 社区项目

Project	Description
oai-compatible-copilot	OpenAI-compatible Copilot - early exploration of custom model integration / OpenAI 兼容 Copilot - 自定义模型集成的早期探索
huggingface-vscode-chat	HuggingFace's VS Code Chat extension / HuggingFace 的 VS Code Chat 扩展
addi	AI-powered development assistant with MCP integration / 带 MCP 集成的 AI 开发助手
ChatGLM-vscode-chat	ChatGLM VS Code extension - GLM model integration reference / ChatGLM VS Code 扩展 - GLM 模型集成参考

AI Providers / AI 服务提供商

Special thanks to the following AI providers for their excellent APIs:

特别感谢以下 AI 服务提供商提供的优秀 API：

智谱AI (Zhipu AI) - GLM series models / GLM 系列模型
Moonshot AI (月之暗面) - Kimi K2 models / Kimi K2 模型
MiniMax - M2 series models / M2 系列模型
Xiaomi MiMo (小米) - MiMo models / MiMo 模型
OpenAI - GPT models / GPT 模型
Anthropic - Claude models / Claude 模型
Google - Gemini models / Gemini 模型

Documentation / 文档参考

Contributing / 贡献

Contributions are welcome! Please feel free to submit a Pull Request.

欢迎贡献！请随时提交 Pull Request。

License / 许可证

MIT

Name		Name	Last commit message	Last commit date
Latest commit History 212 Commits
.github/workflows		.github/workflows
.vscode		.vscode
assets		assets
cli		cli
packages/sdk		packages/sdk
templates		templates
.gitignore		.gitignore
.prettierignore		.prettierignore
.prettierrc		.prettierrc
.vscode-test.mjs		.vscode-test.mjs
.vscodeignore		.vscodeignore
AGENTS.md		AGENTS.md
LICENSE		LICENSE
README.md		README.md
eslint.config.mjs		eslint.config.mjs
package-lock.json		package-lock.json
package.json		package.json
pnpm-lock.yaml		pnpm-lock.yaml
pnpm-workspace.yaml		pnpm-workspace.yaml
tsconfig.json		tsconfig.json

Folders and files

Latest commit

History

Repository files navigation

All-In Copilot

English

✨ Features

Architecture

Quick Start

Method 1: Use CLI (Recommended)

Method 2: Copy Template

CLI Commands

Available Presets

Provider Configuration

SDK Usage

Development

Build SDK

Build VS Code Extension

Watch Mode

CI/CD & Publishing

GitHub Secrets Required

Workflows

Release Process

Tag Naming Convention

Templates

How Dynamic Model Fetching Works

中文

✨ 功能特性

架构

快速开始

方法 1：使用 CLI（推荐）

方法 2：复制模板

CLI 命令

可用预设

提供者配置

SDK 使用

开发

构建 SDK

构建 VS Code 扩展

监视模式

CI/CD 和发布

需要配置的 GitHub Secrets

工作流

发布流程

Tag 命名规范

模板

动态模型获取工作原理

Acknowledgments / 致谢

Core References / 核心参考

Community Projects / 社区项目

AI Providers / AI 服务提供商

Documentation / 文档参考

Contributing / 贡献

License / 许可证

About

Resources

License

Uh oh!

Stars

Watchers

Forks

Releases 31

Packages 0

Uh oh!

Contributors

Uh oh!

Languages

Packages