Qwen Code v0.19.5 Changelog

This release is all about making the AI assistant smarter and easier to use, especially if you interact with it through a web browser on your phone or computer. It gives you more control over the AI's "brainpower," fixes several glitches that made typing feel slow, and adds behind-the-scenes security upgrades for people running their own servers.

What's new

• Control AI brainpower: A new /effort command lets you pick how much computing power the AI uses to solve a problem. You can choose to have it think super hard, or give a quick, fast answer.
• Secure web servers: If you host this software online for others to use, you can now easily turn on HTTPS (the secure, encrypted connection websites use) by pointing to security certificates.
• Better mobile experience: Using the web interface on your phone feels much more like a native app now, with smoother scrolling and screens that fit your phone properly.
• Smoother menus: List menus and pop-ups in the web interface got a big upgrade. They are easier to navigate with your keyboard and work better for people using screen readers.
• Chat timeline: There's a new, slim sidebar timeline in the web chat that helps you see your conversation history at a glance.
• Team approvals: If you have AI "teammates" working on tasks that require a plan, a human team leader now has to approve that plan before the AI can move forward.
• DingTalk integration: The software can now automatically send proactive messages to DingTalk (a popular chat app) to keep teams updated.
• Skill details: When you look at the list of "skills" (special commands the AI can use), it now shows a description and difficulty level so you know exactly what they do.

Improvements

• Faster typing: The web interface no longer freezes or lags when you type while the AI is adding things to the chat.
• Safer background tasks: When background workers print error messages, the system now automatically hides your private passwords and secret keys so they don't leak.
• Better connection tools: The software is now smarter about finding external tools (called MCPs). If a tool doesn't respond right away, it will patiently wait and try again instead of just failing.
• Smoother chat channels: The way the software sends out streams of messages in the background has been upgraded to prevent glitches and timing errors.
• Clearer task statuses: Chat adapters now show exactly what stage a task is in (like starting, running, or finished).

Fixes

• Invisible edits fixed: If you made a change to a file that only added or removed blank spaces, the system used to say "No changes detected." It now correctly shows what you did.
• No more ghost scrolling: Fixed a glitch where old text would randomly flash on the screen when scrolling up in the web chat.
• Web chat loads faster: The web chat no longer freezes while waiting for external tools (MCPs) to load. It also waits to create a chat session until you actually type your first message.
• Commands show up sooner: Shortcut commands (like /review) now appear in the menu before you type your first message, so you know they are there.
• Smart scroll button: The "scroll to bottom" button in the web chat now only appears when you actually need it (when there is content hidden below your screen).
• Disconnected typing: Fixed a confusing bug where you could keep typing in the chat box even after your internet connection dropped.
• YOLO mode fixed: "YOLO mode" (a setting that lets the AI run without asking for permission) no longer interrupts you with annoying approval pop-ups for external tools.
• Memory speed-up: The system no longer wastes time loading memory prompts when there is nothing in its memory to look at.

Breaking changes

No breaking changes in this release.