Edge image

MatterWave on a Raspberry Pi: the API serving the web app, Postgres, and a CPU llama-server decoding grt-nano, as three containers on one box. No cloud in the loop.

What you need

• A Raspberry Pi 5 with 8 GB (a Pi 4 with 4 GB runs grt-nano only), an SSD, and a 64-bit OS with podman 4.4+ (or docker compose).
• The deploy/edge/ bundle from the repository: quadlets, a compose file, first-boot.sh and edge-up.sh.
• The grt-nano GGUF at ~/.local/share/matterwave/models/grt-spark.gguf.
• The API image for linux/arm64, tagged matterwave-api:<sha>-arm64, from the table below.

Put the tags in ~/.config/matterwave/edge.env (MW_EDGE_API_TAG, MW_EDGE_DECODE_TAG), run first-boot.sh, then edge-up.sh. The full sequence, what does and does not run on a Pi, memory estimates and back-up: MatterWave on a Raspberry Pi.

The decode image (the GRT llama.cpp fork for aarch64) is not yet built by CI; the page above says where that stands.

Image tags

No release has been published yet. The table fills itself from the release job.