Mark Ku's Blog
Podcast ConversationAI dialogue version of this article · Mandarin audio

Introduction

This week has seen a massive explosion in the open-source community. dsh-desktop has taken the top spot with a whopping 19k stars, packaging local AI UIs into a one-click installable desktop powerhouse. Today, Mu-Yan is also going to show you a mind-blowing, high-tech project called decimen-optical-transfer, which surprisingly uses dynamic QR codes displayed on a screen for "optical file transfer"—completely without internet or pairing. Plus, what is qm, the tool designed to solve the chaos of AI collaboration? I'll tell you all about it in a moment!


This Week's Open Source Picks

1. decimen-optical-transfer: An "Optical File Transfer" Tool That Instantly Turns Screens and Cameras into an Air-Gapped Network Bridge

  • What it is: It solves the pain point of how to transfer files in or out of an air-gapped environment where there is absolutely no network, USB drives are banned, or you're just too lazy to turn on AirDrop. It allows the sender to display dynamic QR codes on a screen, while the receiver simply scans them with a camera. No network or Bluetooth pairing is required to securely transfer files.
  • This Week's Buzz: This is a brand-new project launched within the last 30 days, rapidly soaring past 6,286 stars on GitHub (approx. 6.3k ★).
  • Key Features:
    • Pure Optical Transfer & Fountain Code Fault Tolerance: Utilizing the Luby transform fountain code algorithm, it can automatically reconstruct files even if a few QR code frames are missed during transmission. It supports a single file limit of up to 64 MB, with real-world testing showing a stable and continuous throughput of 418.5 KB/s from desktop to mobile.
    • SHA-256 Verification & Animation Export: Before transmission completes, it automatically performs a SHA-256 integrity check to ensure the file is not corrupted. It can also export the entire transmission process as a looping APNG or PNG sequence animation for offline storage.
    • PWA Offline Support: As a Progressive Web App (PWA), it runs completely offline after the initial load and supports 12 languages, including right-to-left (RTL) layouts.
  • Quick Start: The fastest way is to directly open the https://decimen.app web version, which is completely installation-free. If you want to self-host, you can clone the repository and run:
    git clone https://github.com/bashalarmistalt/decimen-optical-transfer.git
    cd decimen-optical-transfer
    npm install && npm run dev
    
    It is built with TypeScript and Vite, using zxing-cpp under the hood for QR code decoding.
  • Comparison: No matter how lightweight tools like LocalSend, PairDrop, or qrcp are, they all require both devices to be on the same local network. Decimen, however, directly swaps the transmission medium to "light." Therefore, in extreme scenarios like air-gapped environments or places with no Wi-Fi, it is currently the most elegant and only viable solution.
  • Who It's For: Ideal for engineers who frequently work in air-gapped networks or have extreme privacy and security requirements. One important "gotcha" to note: starting from version v0.4.0, its license has changed from MIT to AGPL-3.0. If your company plans to embed it into commercial products, be sure to double-check the legal terms first.
  • Link: GitHub Portal

2. qm: The Multi-User AI Agent Isolated Collaboration Platform That Ends Workspace Pollution

  • What it is: When teams start adopting AI coding agents, the most common pain point is that "everyone shares a single agent, leading to workspace file pollution, and tangled permissions and background tasks." qm is specifically designed for this, upgrading agents to a multi-user architecture so teams can collaborate safely and cleanly.
  • This Week's Buzz: It has garnered 14,144 stars within 30 days (GitHub 14.1k ★), is licensed under MIT, and supports complete self-hosting.
  • Key Features:
    • Independent Persistent Sandboxes: Every member gets their own isolated sandbox. Files, keychains, permissions, and background tasks are completely independent and won't conflict, yet everyone can still co-operate the same agent within Slack channels or group messages.
    • Pluggable Model Layer & Security Postures: The model layer uses a harness architecture, already supporting Pi, OpenCode, Codex, and Claude Code. It offers three security levels: Strict (all actions require approval), Auto (default, includes content moderation), and Dangerous.
    • Background Automation & Web App Deployment: Built-in automation tasks triggered by cron, watch, and webhooks, and it can even deploy internally developed web apps to designated team members.
  • Quick Start: You don't even need to manually clone the source code. You can generate a deployment repository in your current directory with a single CLI command:
    npm exec --yes --package=@yc-software/qm@latest -- qm init . --org <slug> --target <fly-or-aws>
    
    Then run npm install and follow the interactive wizard to complete the infrastructure and credential setup.
  • Comparison: Tools like Claude Code, OpenCode, or Codex are essentially single-user terminal tools; sharing them among multiple users usually requires clunky, manual account separation. qm is not meant to replace them, but rather to act as a multi-user collaboration harness. Compared to SaaS solutions like Devin, qm has no cloud version—it runs entirely in your own Fly or AWS account, keeping your code and private data completely in your hands.
  • Who It's For: Tech leads and development teams who want to safely introduce AI coding agents to their workflow without uploading sensitive code to third-party SaaS platforms.
  • Link: GitHub Portal

3. dsh-desktop: A Zero-Friction, One-Click Install Desktop Experience for DeepSeek Harness

  • What it is: To run DeepSeek Harness (DSH) locally, you usually have to mess around with runtime environments, set up a local Web UI, and host services yourself. For engineers who hate configuring environments, the barrier to entry is too high. dsh-desktop packages this entire suite into a native Windows and macOS desktop app, allowing you to download and start playing with a single click.
  • This Week's Buzz: The undisputed star-count champion of the week, racking up an incredible 19,589 stars within 30 days (GitHub 19.6k ★), licensed under the MIT license.
  • Key Features:
    • Out-of-the-Box Native Desktop Experience: Built-in system tray controls, native window management, and automatic background service startup—absolutely no terminal commands required to configure the environment.
    • Extreme "Everything is a Plugin" Architecture: The design philosophy of DSH is carried out to the fullest here. Even the desktop shell's windows, system tray, terminal, and auto-update features are implemented as DSH plugins, utilizing the exact same composition mechanism as third-party plugins.
    • Built-in Plugin Marketplace & Stable Upgrades: Features an integrated plugin marketplace where you can browse and install new features with a single click. It pins and maintains the upstream DSH version, with the desktop client only responsible for layering desktop-specific services, making the upgrade path incredibly straightforward.
  • Quick Start: Go straight to the dshdesktop.cn official website to download the Windows x64 installer or macOS Universal DMG. Drag it into your Applications folder like any normal software, launch it, and you're good to go.
  • Comparison: Compared to manually running the local Web UI of DeepSeek Harness, this desktop version is a "hand-holding, hassle-free bundle" of the official experience. However, Mu-Yan must point out one important caveat: this is an independently community-maintained project. The README explicitly states it has no affiliation with the official DeepSeek team. It is completely free under the MIT license; if anyone tries to charge you for this software, reject them immediately.
  • Who It's For: Perfect for engineers and AI enthusiasts who want to experience the power of DeepSeek Harness on their own machines without spending time wrestling with development environments.
  • Link: GitHub Portal

Wrapping Up

The three tools introduced this week—whether it's decimen using "light" as a transmission medium, or qm solving the pain points of multi-user AI collaboration—all showcase the incredibly creative engineering mindset of the open-source community. Mu-Yan highly recommends giving decimen a try; watching your screen flash and seeing a file transfer straight to your phone feels absolutely magical. That's all for this week's "Open Source Weekly." See you next week!

Author

Mark Ku

擁有 10+ 年經驗的資深軟體工程師,現為 AI 應用 Builder,專注於大型平台架構與簡化複雜系統設計,從電商系統到訂閱與收費平台,結合 AI Agent、AI 整合與自動化開發,打造高效率且可持續演進的產品技術基礎。Read More

Found this useful?

The author's free tools, daily podcasts and newsletter are all here.

Mark Ku · This article is licensed under CC BY 4.0. Credit the author and link back to the original when reusing it.

Comments

Subscribe to Newsletter

Subscribe to get new posts delivered instantly — never miss a tech share.

By submitting, you agree to receive emails. You can anytime.

Popular Posts

View all
Mark Ku
··602

Oracle Cloud Always Free Tier: Linux Host and Static IP for a $0 Cloud Solution

Oracle Cloud Always Free Tier: Linux Host and Static IP for a $0 Cloud Solution
Mark Ku
··490

Say Goodbye to Postman's Fee Trap! A Hands-on Guide to Bruno, the Open-Source Git-Native API Testing Powerhouse.

Say Goodbye to Postman's Fee Trap! A Hands-on Guide to Bruno, the Open-Source Git-Native API Testing Powerhouse.
Mark Ku
··333

A Free, Open-Source, Notion-like Knowledge Base — A Complete Guide to Deploying and Backing Up Outline Wiki

A Free, Open-Source, Notion-like Knowledge Base — A Complete Guide to Deploying and Backing Up Outline Wiki
Mark Ku
··264

Training Your Own AI Voice: Hardware Requirements, Open-Source Model Comparison, and LoRA Fine-Tuning

Training Your Own AI Voice: Hardware Requirements, Open-Source Model Comparison, and LoRA Fine-Tuning
Mark Ku
··221

Building an Efficient API Management Platform: Deploying Kong Gateway from Scratch - Part 1

Building an Efficient API Management Platform: Deploying Kong Gateway from Scratch - Part 1
Mark Ku
··215

Setting Up Samba on Ubuntu to Share Folders with Windows 11

Setting Up Samba on Ubuntu to Share Folders with Windows 11
🛠️ 用螢幕播 QR 碼也能離線傳檔?decimen 狂吸萬星的背後秘密|開源好物週報 - Mark Ku's Tech Notes