Qwen Code v0.19.6 Changelog

This update is all about making the AI smarter when handling big, multi-step projects, while also making the app look and feel much smoother—especially on mobile devices. It’s a great upgrade for developers who use Qwen Code to manage complex workflows and want fewer crashes and glitches.

What's new

• Nested Sub-Agents: The AI can now spawn "sub-agents" (mini helper AIs) to handle smaller parts of a big task. These helpers can even spawn their own helpers, up to a limit you set. The web interface now shows them as a tree diagram so you can easily track what's going on.
• Data Visualization Skill: The AI now has a built-in skill to help create charts and visual data displays.
• Vision Model Picker: You can now easily choose different "vision" models (AIs that can understand images) directly from the web dashboard.
• Smarter Slash Commands: The menu for typing quick "/" commands is now taller, shows how many options are in each group, and supports "fuzzy search" (finding what you want even if you mistype it).
• Better Background Tasks: You can now set time limits for specific tool actions, and you can configure exactly when repeating background jobs (like scheduled tasks) should expire.
• Workspace Memory: Added new ways for the AI to "forget" old information and "dream" (process and organize memories) without being tied to a specific chat session.
• Version Display: The app's version number is now visible at the bottom of the web sidebar.

Improvements

• Faster Startup: Added "bootstrap fast paths," which is a fancy way of saying the app skips unnecessary steps to launch much quicker.
• Smaller Memory Footprint: When the AI looks at images or other mixed media in your chat history, it now uses way less data to remember it, keeping things running smoothly.
• Better Mobile Loading: When you switch chats on your phone, the screen no longer stutters, and it gives you clearer loading feedback.
• Improved Streaming: The app can now handle large streams of data from the background (NDJSON streams) more efficiently, and the idle timeout (how long the app waits before assuming a connection died) has been increased so things don't disconnect too early.
• Chat Bot Upgrades: The QQ Bot now sends messages in clean Markdown format by default and manages reply chains better so they don't get stuck.

Fixes

• Mobile Visual Glitches: Fixed a bug where switching sessions on mobile caused the screen to jump around (jank). Also fixed a layout issue where certain menus were stretching weirdly.
• macOS Security Path: Fixed an issue on Mac computers where the app was looking for its security permissions (seatbelt profile) in the wrong folder, which could cause it to fail to run properly.
• Sub-Agent Crash: Fixed a crash that happened if a helper AI tried to use a specific placeholder tag (${hook_context}) when no hook (a trigger script) was actually set up.
• Disabled Skills: Fixed a bug where the app was still trying to use skills that you had explicitly turned off in your settings.
• VS Code False Alarms: Fixed the VS Code extension's secret scanner so it stops flagging normal code patterns as security threats.
• Security Hardening: Added stronger security checks to the QQ Bot's gateway, made its logging safer (so it doesn't accidentally leak private info), and fixed how it handles state changes.

Breaking changes

No breaking changes in this release.