Show HN: ModelSignature – Get end-user feedback for your OS AI models (modelsignature.com)

ModelSignature is a lightweight system that embeds an actionable feedback URL directly into open-source AI models by using LoRA fine-tuning. Developers register a ModelSignature feedback page, fine-tune the model (~40–50 minutes on a free GPU using LoRA), and deploy the resulting model anywhere (Hugging Face, Replicate, custom endpoints). End users can ask the model where to report issues and receive the URL, enabling direct bug reports, feature requests, and access to verified model specs and authenticated provider information. This approach is significant because it creates a persistent, low-friction feedback loop for models running in the wild, improving accountability, provenance, and iterative maintenance for OS models. Technically, using LoRA means the change is small, fast, and compatible with common deployment workflows; the method doesn’t require full re-training or heavy compute. By surfacing a verified contact and spec page inside the model itself, ModelSignature reduces discovery friction for reporting problems and requesting features—helpful for safety monitoring, rapid bug triage, and community-driven model improvement—while also raising considerations around verification and abuse-resilience for the embedded reporting channels.