AI Tool Fixes Annoying Log Glitches on Windows

If you've ever used a command-line AI tool, you know how frustrating it is when logs get jumbled or commands go missing. That's exactly what happened with llama.cpp, a popular open-source project that lets you run AI models locally. A recent bug caused color reset codes to land on the wrong line, creating empty log lines and causing the router to miss important commands. This made debugging harder and could even break multi-step AI tasks.

The fix is technical but simple in effect: each log line is now self-contained, so colors don't bleed into the next line. The router now passes its color settings to child processes, and on Windows, the logger turns on virtual terminal mode so colors show up properly instead of raw escape codes. This means Windows users finally get the same colorful, readable logs as Mac and Linux users.

Even more importantly, the way commands and logs are separated has been improved. Previously, the child process sent both state commands and logs on the same pipe, forcing the router to guess which was which. Now, commands go through stdout and logs through stderr, so the router can reliably handle both. This eliminates a whole class of bugs and makes the system more robust.

For anyone using AI tools that run locally—whether for coding, writing, or data analysis—this means fewer headaches. Logs are cleaner, commands are executed reliably, and the overall experience is smoother. It's a behind-the-scenes fix, but it makes AI tools more dependable for everyday use.