返回目录
开发工具 完整应用

dsh-vscode

Lixxx1/dsh-vscode

A Claude Code/Codex-style VS Code sidebar for DeepSeek Harness. 像 Claude Code 、Codex 一样,在 VS Code 侧边栏中使用 DeepSeek Harness。

Stars
7
Forks
0
Issues
0
更新
今天

PROJECT TOPICS

项目标签

PROJECT README

README

DeepSeek Harness for VS Code

Use DeepSeek Harness as a coding assistant in the VS Code sidebar.

The goal is the same kind of workflow offered by Claude Code and Codex in VS Code: keep an assistant beside your editor while you work, with the current project already in context.

English | 简体中文

CI MIT License Status

DeepSeek Harness in the VS Code sidebar

Features

  • Chat with DeepSeek about the project currently open in VS Code.
  • Keep the assistant visible in the right sidebar while editing code.
  • Create new conversations or continue existing project sessions.
  • Choose the Model and Reasoning Effort separately.
  • Read streamed Markdown responses and expandable tool cards without leaving the editor.
  • Review tool approvals and answer DeepSeek's follow-up questions in the conversation.
  • Attach images to prompts, and stop running tasks directly from the sidebar.
  • Open the same conversation in an editor tab when you need more space.

Install

Install the official DeepSeek Harness CLI first:

npm install -g @deepseek-ai/dsh

Then build and install the extension:

git clone https://github.com/Lixxx1/dsh-vscode.git
cd dsh-vscode
corepack pnpm install --frozen-lockfile
corepack pnpm run check
corepack pnpm run package
code --install-extension dsh-vscode.vsix

Requires VS Code 1.100 or newer and Node.js ^22.19 or >=24.

Use

  1. Open a trusted project folder in VS Code.
  2. Select DeepSeek Harness in the right sidebar. If it is hidden, find it under Other Views.
  3. Use the key button to configure DEEPSEEK_API_KEY.
  4. Choose a conversation, Model, and Reasoning Effort, then start working.

For implementation details, see Architecture. To contribute, read CONTRIBUTING.md.

License

MIT

CLASSIFICATION EVIDENCE

分类依据

项目类型完整应用
功能分类开发工具
规则置信度

系统优先读取 GitHub Topics,再与站内分类词典和词根规则比对。