Offline Camera Handling and ROCm GPU Updates

Transcript

Good morning, this is your Frigate NVR update for January 23rd, 2026.

Yesterday brought three merged pull requests focused on reliability and AI enhancements. Hawkeye217 merged offline preview image support, addressing a long-standing user pain point when cameras go offline or lose their detect stream. The change adds 297 lines of code across the media API, preview output system, and live player component, ensuring users see meaningful feedback instead of blank screens during camera outages. The implementation includes comprehensive test coverage for both the HTTP API and preview loader functionality.

NickM-27 contributed two significant updates. The first adds live context tools to Frigate's LLM integration, expanding the AI chat system with 188 new lines of code. This enhancement allows the language model to access real-time system information, making conversations more contextually aware and useful for troubleshooting and system monitoring.

The second update from NickM-27 brings ROCm support to version 7.2.0, specifically benefiting users with consumer AMD GPUs. The update touches Docker configuration files and detection runners, streamlining the codebase while expanding hardware compatibility. This change represents AMD's continued investment in AI acceleration for consumer hardware.

All three pull requests received single approvals and were merged within hours of submission, indicating solid code review processes and contributor familiarity with the codebase.

What's next: The offline preview feature should immediately improve user experience during camera maintenance windows. The ROCm update positions Frigate users to leverage newer AMD GPU capabilities for object detection workloads.

That's your Frigate development update. Back tomorrow with more changes from the community.