Search...Search plugins and themes...
⌘K
Sign in
  • Get started
  • Download
  • Pricing
  • Enterprise
  • Account
  • Obsidian
  • Overview
  • Sync
  • Publish
  • Canvas
  • Mobile
  • Web Clipper
  • CLI
  • Learn
  • Help
  • Developers
  • Changelog
  • About
  • Roadmap
  • Blog
  • Resources
  • System status
  • License overview
  • Terms of service
  • Privacy policy
  • Security
  • Community
  • Plugins
  • Themes
  • Discord
  • Forum / 中文论坛
  • Merch store
  • Brand guidelines
Follow us
DiscordTwitterBlueskyThreadsMastodonYouTubeGitHub
© 2026 Obsidian

BuddyBridge Deep

BuddyBridgeBuddyBridge43 downloads

An unofficial bridge for DeepSeek Harness via the Agent Client Protocol (ACP): multi-turn chat, streaming responses, thinking blocks and tool calls.

Add to Obsidian
  • Overview
  • Scorecard
  • Updates1

第三方 Obsidian 插件 · 通过 ACP 桥接 DeepSeek Harness (DSH) Unofficial Obsidian plugin · Bridges DeepSeek Harness via ACP

在 Obsidian 里直接和 DeepSeek Harness 对话:流式回复、思考过程、工具调用、多轮会话,并自动复用 DSH Web 的 API Key 与地址,开箱即用。

Chat with DeepSeek Harness inside Obsidian — streaming replies, thinking blocks, tool calls, multi-turn sessions, and zero-config reuse of your DSH Web API key & base URL.


功能 Features

功能 Features
流式聊天面板(停止 / 重试) Streaming chat with stop / retry
可折叠思考块与工具调用卡 Collapsible thinking blocks & tool-call cards
每对话一个 DSH 会话,多轮上下文连贯 One DSH session per conversation, multi-turn context
上下文注入:笔记路径 / 全文 / Vault 根路径(可开关) Context injection: note path / full content / vault root (toggles)
工具调用权限弹窗 Permission modal for tool calls
断链自动重连(指数退避) Auto-reconnect with exponential backoff
双语界面(简体中文 / English,跟随 Obsidian) Bilingual UI (zh / en, follows Obsidian)
自动读取 DSH Web 凭据与地址,零配置 Auto-reads DSH Web API key & base URL — no manual setup
会话分叉 / 附加笔记 / 上下文占用指示 Fork conversations / attach notes / context usage meter
设置与聊天记录导出 / 导入 Export / import settings & chat history
一键后端配置脚本 One-click backend setup script

安装 Installation

1. 前置:DeepSeek Harness + ACP profile

本插件通过标准 ACP (Agent Client Protocol) stdio JSON-RPC 与 DSH 通信,需要本机已安装 DeepSeek Harness(dsh)并装配 ACP profile。

已装 dsh 的用户可一键装配(推荐):

powershell -ExecutionPolicy Bypass -File scripts/setup-acp.ps1

脚本会自动安装配套的 @deepseek-ai/dsh-acp-demo 到 ~/.dsh/profiles/acp、编译 koffi 原生库并生成 cordis.yml,最后打印实际启动命令,形如:

node C:\Users\<you>\.dsh\profiles\acp\node_modules\@deepseek-ai\dsh-acp-demo\lib\bin.js -c C:\Users\<you>\.dsh\profiles\acp\cordis.yml

注意:dsh --profile acp 不会加载 ACP 服务器——demo 包不是 profile bundle,必须用上面的独立 bin 命令。

2. 安装插件本体

  1. 复制 main.js / manifest.json / styles.css 到 <你的Vault>/.obsidian/plugins/buddybridge-deep/
  2. Obsidian 设置 → 第三方插件 → 开启「BuddyBridge Deep」
  3. 设置页 →「DSH ACP 命令」填好启动命令 → 点「诊断连接」验证

配置 Configuration

设置 → BuddyBridge Deep:

项 说明 Setting
DSH ACP 命令 启动 ACP stdio 服务器的命令 ACP start command
DeepSeek API Key 留空自动读取 DSH Web(~/.dsh/.credentials.yaml) Auto-read from DSH Web when empty
DeepSeek Base URL 留空自动读取 DSH Web(~/.dsh/settings.yaml),如火山方舟 Auto-read from DSH Web when empty
请求超时 单次请求超时秒数 Request timeout
上下文注入 笔记路径 / 全文 / Vault 根路径开关 Context injection toggles
外观 主色调 / 字体大小 Primary color / font size
界面语言 简体中文 / English / 跟随 Obsidian Language
管理 最大对话数 / 导出 / 导入 / 重置 Max conversations / export / import / reset

零配置凭据:DSH Web 已能对话的话,插件会直接复用它的 API Key 与 Base URL(读 ~/.dsh/.credentials.yaml 与 ~/.dsh/settings.yaml),无需在插件里重复填写。

开发 Development

npm install          # 安装依赖
npm run build        # tsc 类型检查 + esbuild 打包 → main.js
npm test             # jest:239 个用例(含 ACP 客户端端到端集成测试),80% 覆盖率门禁
npm run dev          # 监听模式(esbuild --watch)

独立冒烟脚本(不依赖 Obsidian):

node scripts/acp-smoke.mjs --command "dsh --profile acp" --cwd "C:\your\vault"

架构 Architecture

src/
├── core/                # 共享抽象层:bridge-adapter / stream-chunk / vault-tools
│   ├── dsh-env.ts       # DSH 全局 .env 读写(~/.dsh/.env)
│   └── dsh-shared.ts    # 自动读取 DSH Web 凭据与地址(credentials.yaml / settings.yaml)
├── bridges/deep/        # DeepSeek 适配层:ACP JSON-RPC stdio 客户端 + 命令解析
├── chat/manager.ts      # 会话管理
├── settings/            # 设置页、确认/权限弹窗、诊断、一键后端配置
├── views/chat.ts        # 聊天视图(流式、思考/工具块、错误卡重试)
├── context.ts / io.ts   # 上下文注入 / 导出导入
└── main.ts              # 插件入口

已知限制 Known Limitations

  • 官方 @deepseek-ai/dsh-acp 自动化仅支持 fresh sessions:无 session/load / session/list / session/fork(插件重载后每对话新建会话)。
  • 权限「始终允许」依赖服务器是否提供 allow_always 选项(官方默认仅 allow-once / reject-once)。
  • 思考过程是否可见取决于 ACP 服务器是否发送 agent_thought_chunk。

License

MIT

HealthExcellent
ReviewCaution
About
Chat with DeepSeek Harness inside Obsidian using streaming replies, collapsible thinking blocks, tool-call cards, and multi-turn sessions. Auto-read your DSH Web API key and base URL, keep one DSH session per conversation, inject note or vault context, and auto-reconnect with exponential backoff.
AIIntegrationsInterface
Details
Current version
0.2.0
Last updated
3 weeks ago
Created
3 weeks ago
Updates
1 release
Downloads
43
Compatible with
Obsidian 1.7.2+
Platforms
Desktop only
License
MIT
Report bugRequest featureReport plugin
Author
BuddyBridgeBuddyBridgeben4202121
GitHubben4202121
  1. Community
  2. Plugins
  3. AI
  4. BuddyBridge Deep

Related plugins

Agent Client

Chat with Claude Code, Codex, Gemini CLI, and more via the Agent Client Protocol — right from your vault.

Fast Note Sync

Real-time sync of your vaults across server, mobile, and web; shareable with anyone; supports REST and MCP integrations to build your personal AI knowledge base.

Smart Composer

AI chat with note context, smart writing assistance, and one-click edits for your vault.

Claude Sidebar

Run Claude Code in your sidebar.

Claude Code IDE

Connect your vault to Claude Code and share editor context with the CLI.

Hider

Hide interface elements such as tooltips, status bar, titlebar, and more.

BRAT

Easily install a beta version of a plugin for testing.

Claudian

Embeds Claude Code/Codex and other local Agents as AI collaborators in your vault.

LanguageTool Integration

Advanced grammar and spell checking, powered by LanguageTool.

Smart Connections

Find related notes and excerpts while writing. Your AI link building copilot displays relevant content in graph + list view. A local embedding model powers semantic search. Zero setup. No API key.