ipro_ai_link¶

Board firmware for IPRO AI Studio "Run on board": loads .inm models and inputs from microSD (sd:/), runs them on the NPU and prints the raw outputs; also usable by hand from the console

Item	Details
Source	`apps/test/ipro_ai_link`
Boards and configs	Reef (`defconfig_reef`); `.config` targets the IPRO7 EVB
Console	UART 115200 (also on a USB CDC port)
Build	Reef: `make -C apps/test/ipro_ai_link DEFCONFIG=defconfig_reef`

Run it on the Reef¶

This application has no README; the steps below follow its source. You need a FAT-formatted microSD card.

Build
```
make -C apps/test/ipro_ai_link DEFCONFIG=defconfig_reef
```
The image is apps/test/ipro_ai_link/build_reef/build_out/ipro_ai_link_IPRO7.bin.

Flash: enter download mode, then from the IntelProCube directory run

./ipro_iot_tool_macos --chipname ipro7 --chip-variant IPRO7AI \
    --port /dev/cu.usbserial-XXXX --baudrate 921600 \
    --firmware <SDK>/apps/test/ipro_ai_link/build_reef/build_out/ipro_ai_link_IPRO7.bin

Check for [All Success], then press SW2.

Run: open the console at 115200 baud; the board prints ipro_ai_link ready. On a computer's USB port the board shows up as two CDC serial ports (VID 0x3896, PID 0x0704): the first carries data, the second is a console.
Run a model by hand: copy an .inm model and a raw input file to the SD card, insert it, then type
```
ai_mount
ai_load sd:/model.inm
ai_input sd:/input.bin
ai_run
ai_out
```
ai_out prints the output tensor in hex, every byte stored XOR 0x80. Paths always start with sd:/.

The SDK ships models to try, for example apps/platform/ipro7_demo/demo/ai/ai_model/mbv2.inm. Download IPRO AI Studio from GitHub.