BPi-f3 for energy monitor/home automation

for those who like to work on the cutting edge . you may want to consider the bpi-f3 . since I like to see how much one can squeeze into a device.

currently on this $90 that has 8 gigs of ram and 8 cores riscv64 cpu. ( you pay that just for the coralAI )

i am running as openwrt router platform. so acts as the home router with vpn with policy routing control( you can control what ips go through a vpn or directly), NAS. but also have home automation. zigbee2mqtt hub. and my version of openwrtAI, and on top of this that new with riscv build via docker I am running frigate NVR, immich . kodbox and openwebui ( running deepseek and llama. qwen models) all running at minimal loading of 14%. unless frigate is actively tracking and doing facial and vehicle recognition and capture and recording , or if immich received a new batch of media and AI processiing it into categories or running LLM. then average load is about 60%.

benefits of the bpi-f3 over the rk3588 (ie orangpi 5+ and similar variants) 1/2 the price and better internal hardware support. thought software support is abit lacking since it riscv64 but workable as there lots of accessible documentation.

but I would say the nicest benefit of the bpi-f3 for me is the option of running frigate NVR on the openwrt router platform with built in NPU support. sure you could do it on the rk3588, intel or RPI4 easier as well. but the rk3588/intel will cost you twice as much and rpi4 with added npu will cost you up to 4X as much. plus the added benefit of very very low power consumption

if building your own docker images for bpi-f3 use this as the base image

harbor.spacemit.com/bianbu/bianbu:latest  ( it ubuntu:noble based)

and up date source and porting

sed -i '/^Types:/ s/$/ deb-src/' /etc/apt/sources.list.d/bianbu.sources
sed -i '/^Suites:/ s/$/ noble-porting\/snapshots\/v2.1 /'  /etc/apt/sources.list.d/bianbu.sources

this ill give you access to the hardware with out having to compile for other base images to have support
what you are looking for : ffmpeg and gstreamer for hardware encoding decoding,
mpp far codec acceleration
python3-spacemit-ort for npu access s for llm and other model