Qwen Code v0.19.7 Changelog

This release is all about making the software more reliable and easier to use. It matters a lot to developers who use this AI coding assistant every day, because it introduces a "fallback" system that keeps the AI running even if the main servers are busy, while also adding powerful new dashboards to monitor what the AI is doing behind the scenes.

What's new

• Automatic Server Backup: If the main AI model gets overloaded and crashes, the software will automatically switch to a backup model so your work isn't interrupted.
• Web Dashboard Upgrades: The web interface (web-shell) now has a dedicated page to show "daemon status" (what the background helper program is doing), complete with time-series charts to track performance over time.
• Better Session Management: You can now organize your chat sessions from the web sidebar—meaning you can archive, unarchive, or delete them to keep your workspace clean.
• Custom Tool Timeouts: You can now set a specific time limit for how long the AI is allowed to try using a single tool before it gives up.
• Multi-Folder Support: The command-line interface (CLI) can now safely handle workspaces that pull files from multiple different folders at once.
• Smarter Mentions: In the web interface, when you tag a file or tool using the "@" symbol, it now shows little icons and custom panels to make things easier to read.

Improvements

• Faster File Searching: When searching for files, the system now skips over folders you've told it to ignore right from the start, rather than searching them anyway and filtering them out at the end. This makes searching much faster.
• Faster Settings Loading: The command-line tool now saves your workspace settings in its memory and only updates them if the file actually changes, speeding up how fast the program boots up.
• Faster Auto-Fixes: The automated system that fixes code errors (the autofix pipeline) has been streamlined. It skips duplicate steps and runs tests more efficiently.
• Smoother Bot Reviews: The bots that automatically review code on GitHub will no longer waste time reviewing old, closed, or stale pull requests.

Fixes

• Login Glitches: Fixed an issue where changing your API key (your access password) would cause a permanent "401 Unauthorized" error, locking you out. Also fixed a false warning that popped up during pre-checks.
• Scroll Lock: Fixed a bug where long AI responses would cause the command-line screen to lock, forcing you to scroll all the way to the top.
• Windows Text Fix: Added a special code prefix for the Windows command prompt (cmd.exe) so it correctly reads text characters (UTF-8) without garbling them.
• Branch Checkout Crash: Fixed a problem where the auto-fix tool would accidentally mess up your tracked files when switching between different project branches.
• Timeout Bug: Fixed an issue where setting a request timeout to "0" (which usually means "turn the timeout off") would make the program crash immediately instead of waiting forever.
• File Reading Rules: The system now correctly remembers when you've attached a file using the "@" symbol, treating it as "read" so it doesn't trigger rules that stop the AI from reading unread files.
• Thinking Mode Fix: Fixed an issue where the AI's "thinking" mode was accidentally turned on for servers that don't support it.
• Context Window Math: Corrected the math used to figure out how much text a custom AI model can remember at once (the context window).

Breaking changes

No breaking changes in this release.