I gave my Windows 11 clipboard a local LLM brain, and it made trivial copy-paste operations a lot more productive
Since everything remains local, I can run inference operations on private clipboard data freely
Locally-hosted models may not match the computational power of their cloud-based counterparts, but for someone utilizing outdated hardware, they can still meet the majority of productivity needs. The author has built a suite of productivity apps and lightweight agent harnesses that rely on locally run AI models. Among these additions is a clipboard assistant that performs analysis tasks, translations, and OCR workloads without compromising privacy by sending data to external servers.
After trying various community tools, the author settled on the Sirius AI Tray Assistant, powered by the Gemma 4 E4B (Q4_K_M) LLM, which runs efficiently on the author's GPU. The assistant is easily configured to use llama.cpp as the server and connects to the Gemma 4 E4B instance via the http://localhost:8080/v1 link. Custom prompts allow for tailored analysis, such as removing formatting from PowerShell scripts, Linux commands, and code snippets.
The tool also supports translating text to English and OCR workloads on screenshots, making it versatile for various tasks. The assistant even integrates with screenshots created using the Windows Snipping Tool and can generate OCR reports when text is not copyable via Ctrl + C. Despite the author's outdated systems, the localized setup is efficient and runs well on various GPUs, making the Sirius Assistant a valuable addition to their Windows machines.
Written by urgent.news from XDA Developers's reporting — not their text. Machine-written — may contain errors; check the original before relying on it.
