So while talking with the Station PC AI chat, it shared that the Pocket Cloud hardware is Rockchip RK3588. I researched and it seems there are 3 NPUs on that device with 4GB RAM? If so, it would be killer you could enable an image of Ollama or LocalAI to leverage and run smaller models on NPUs. How awesome for a docker app on the Pocket Cloud that we could configure to support interactive chat, or agents like OpenClaw or Dev Harnesses like Pi. If this is possible, please do it… would be so awesome!
See here for demo and bits and pieces that might help make it happen:
Videos of it working on NPUs:
https://www.reddit.com/r/RockchipNPU/comments/1p4t33u/i_created_a_llamacpp_fork_with_the_rockchip_npu/
That’s such a great insight! You’re absolutely correct — this is indeed our next-generation 4-bay NAS product.
Just to clarify: the currently available PocketCloud is equipped with the RK3568 chip, which fully meets daily NAS storage needs and lightweight network service usage.
Our team has already included your idea in our in-depth development roadmap. We’d also love to hear more from you: what real-world scenarios do you mainly use for running local AI in daily life? Such as private AI chat, local document parsing, AI automation agents, local code development, or other personalized demands. Your actual usage scenarios will help us better prioritize features and optimize product adaptation.