Qwen Code v0.19.10 Changelog

This release is all about making the software better for people juggling multiple coding projects at once. Whether you're using the web interface or connecting through code, it's now much easier to keep your work organized, see exactly what the AI is doing behind the scenes, and recover smoothly when internet connections or logins glitch out.

What's new

• Multi-workspace support: You can now connect and manage multiple project folders at the same time. The system keeps their data separate and can even show them side-by-side in a split view.
• Web Shell review panel: The web interface now has a side panel where you can inspect edited files, view differences (diffs), and check on scheduled tasks without leaving your chat.
• Interactive questions in SDK: If you're writing code that uses this tool, your AI can now pause and ask the user a question, wait for an answer, and then keep going.
• DingTalk mentions: If you use DingTalk (a chat app), the bot can now tag the person who sent a message in its reply.
• Prompt stashing: If you type a question but aren't ready to send it, you can press Ctrl+S. The tool saves it just for that project and brings it back next time.
• Git branch chip: The text box where you type now shows which Git branch you're working on, so you always know where your code is saving.
• Custom colors: You can now organize your chat sessions in the web interface using custom color codes.
• Table reader: If the AI generates a huge Markdown table, you can double-click any cell to pop open a window that makes it easy to read and copy the full text.

Improvements

• Smarter session recovery: If your internet drops, a login expires, or the background system restarts, the tool is much better at saving your chat history and picking up right where you left off without crashing.
• Better subagent tracking: "Subagents" are like helper-AIs working in the background. They now show a clear timeline of what they're doing, the exact commands they are running, and where their logs are saved.
• Faster model switching: The menu for choosing which AI model you want to use has been redesigned to show clearer info and switch faster.
• Smarter code reviews: When the AI reviews your code, it now double-checks its own findings before posting them to make sure it's not giving you bad advice.
• Adjustable timeouts: You can now set a time limit for background commands, so if a command gets stuck, it won't freeze your whole session forever.
• More developer controls: Programmers using the TypeScript or Python SDKs can now control how much "effort" the AI puts into thinking, check account usage stats, and choose how the tool communicates behind the scenes.
• Surviving restarts: If you add a new workspace while the tool is running, it remembers it even if the system restarts, so you don't have to set it up again.

Fixes

• Truncated thinking fixed: Previously, if you tried to expand the AI's "thinking" process while it was typing, it only showed a cut-off preview. Now it shows the full thought process in real-time.
• Auto-reconnecting workers: Chat connections (like DingTalk and Feishu) used to break if the background system restarted. They now reconnect automatically without you needing to fix them.

Breaking changes

No breaking changes in this release.