I put a local AI model on a bootable USB, so I can troubleshoot a dead PC with a chatbot that doesn't need the internet
My rescue USB now also helps me make sense of disk errors.
A bootable Linux USB drive can be used to troubleshoot a non-starting PC, allowing one to examine internal components and utilize recovery tools without the need for the broken operating system. The reporter sought to enhance this setup by incorporating a local AI model on the same USB drive, enabling the use of a chatbot for diagnosing issues even without an internet connection.
A 64GB USB 3 drive was chosen as the platform, onto which Ventoy was installed to host bootable ISO files. The remaining space was designated for AI runtime and model files. SystemRescue, a Linux distribution with built-in disk diagnostics, partitioning tools, file recovery utilities, and Firefox, was selected for its utility in investigating Windows installations.
The AI model utilized in this arrangement was llama.cpp with Qwen3-4B-Instruct-2507, specifically the Q4_K_M version, which consumes around 2.5GB of memory and operates with a context size of 2,048 tokens on an 8GB Lenovo laptop. CPU inference was employed to ensure the chatbot's functionality without relying on GPU drivers. After booting SystemRescue, the USB data partition was mounted, and the llama-server was launched, providing a chat interface accessible via Firefox.
SystemRescue was used to analyze the computer's detection of connected drives and partitions, providing the chatbot with an lsblk output and a description of the problem. This information, combined with the environment details (Windows operating system running on SystemRescue), enabled the chatbot to provide a more informed diagnosis.
The chatbot could generate a drive-health report using smartctl or analyze dmesg messages for storage-related errors, helping to identify potential hardware issues. The chatbot could also assist in identifying the Windows drive, explaining the significance of the lsblk output, and suggesting the next diagnostic steps without altering the disk.
This approach allowed for a more informed and targeted troubleshooting process, preserving the integrity of the problematic Windows installation while leveraging the capabilities of local AI models.
Written by urgent.news from XDA Developers's reporting — not their text. Machine-written — may contain errors; check the original before relying on it.
