Run it
Docker with at least 8 GB of memory, on an Apple-silicon Mac or an x86 or arm64 Linux or Windows machine (Intel Macs are not supported). Inference runs on your CPU. The image is 0.74 GB to download and 2.1 GB on disk (arm64, measured on the build from main 006fb29), plus 1.1 GB of model files you download only for S08.
docker pull ghcr.io/action-state-group/devcon-workshop:latest
docker run -it --name devcon8 -v devcon8-home:/home/participant \
-p 127.0.0.1:3141:3132 -p 127.0.0.1:8080:8080 -p 127.0.0.1:9611-9614:9621-9624 \
ghcr.io/action-state-group/devcon-workshop:latest
Inside the container, devcon8 help lists every step, and each step is safe to rerun:
devcon8 doctor
devcon8 s02 # then open http://127.0.0.1:8080/settlement-record.html
devcon8 models # 1.1 GB, once, only for S08
devcon8 s08 start # then the steps in the walk-through, in order; devcon8 s08 stop when done
devcon8 go-seal
devcon8 brain claude # or codex; then devcon8 nanda <email>, then devcon8 trade
Joining the demo mesh as a serving node takes a few more commands and the mesh's join key; the walk-through has both. A pinned image digest, the USB-stick path and disk-space advice are in the README.