Running a Plex Media Server on a Raspberry Pi 5 has been a reliable way to stream personal media libraries for years, but even the most stable setups can hide subtle hardware problems. When a recent AI-powered diagnostic uncovered an issue I’d completely overlooked—a lingering SD card causing I/O errors—it raised questions about how often home servers should undergo automated checkups.
Unlike traditional troubleshooting methods that rely on manual log parsing or trial-and-error fixes, Google’s Gemini AI, running in the experimental Antigravity IDE, performed a four-stage diagnostic sweep. It examined system vitals, Plex logs, transcoding performance, and storage health—all while requesting explicit approval for each terminal command executed on the Pi. The result? A B+ grade for the server, with a single critical warning: despite upgrading to an NVMe drive, the old SD card was still being accessed, flooding logs with errors.
- Hardware: Raspberry Pi 5 with NVMe storage (intended boot drive) + residual SD card (unintended access)
- Software: Plex Media Server, Google Antigravity IDE, Gemini 3 Flash model
- Diagnostic Process: Four-phase check (vitals → logs → transcoding → storage)
- Key Finding: I/O errors from SD card, misdiagnosed as a boot failure by Gemini 3 Flash
- Correction: Disconnected SD card; cleared Plex transcoder cache
- Grade: B+ (vitals strong, but lingering hardware issue)
The misdiagnosis—initially flagging the SD card as a boot drive rather than a leftover storage device—highlighted both the power and the limitations of AI in hardware diagnostics. A second opinion from Claude Opus 4.5 confirmed the Pi was correctly booting from NVMe but was still polling the SD card, creating false alarms. The fix was straightforward: unplug the old card and clear Plex’s transcoder cache, but the incident underscored a broader lesson for home server owners.
For most users, Plex runs smoothly until a file fails to transcode or a library update stalls. But automated diagnostics—especially those leveraging AI’s pattern-recognition strengths—can spot issues like degraded storage or inefficient transcoding before they become critical. The Raspberry Pi 5’s combination of low power draw and robust performance makes it a popular Plex host, but even high-end hardware benefits from periodic checkups. In this case, Gemini’s four-phase approach—vitals, logs, performance, and storage—mirrored what a seasoned sysadmin might do manually, but with the added advantage of speed and consistency.
While Gemini 3 Flash’s initial error was corrected by a more precise model, the experience revealed how AI can serve as both a preventive tool and a teaching one. For users without deep technical knowledge, platforms like Antigravity lower the barrier to diagnosing complex systems. The tradeoff? Full access to system files and logs, which requires careful consideration of privacy and permissions. For now, the lesson is clear: even a well-maintained Plex server can hide surprises—and an AI checkup might just be the oil change your media library needs.
