Go from a folder of tarballs to a stack that answers. The last step asks the running engine what it can see, before anything is spent.
Verify the download against the digests published beside it:
shasum -a 256 -c SHA256SUMS
You should see revetly-engine-0.8.0-linux-<ARCH>.tar.gz: OK (as shasum shows it)
Load the tarball for your architecture as a Docker image:
docker load < revetly-engine-0.8.0-linux-<ARCH>.tar.gz
Replace <ARCH> with arm64 on Apple Silicon and amd64 on Intel.
You should see Loaded image: revetly:0.8.0 (as Docker shows it)
Scaffold an instance into the folder your project lives in:
docker run --rm -v "$PWD:/project" -u "$(id -u):$(id -g)" revetly:0.8.0 revetly init --transport container --group-id your-product
You should see Instance 'your-product' at /project (container transport, engine revetly:0.8.0)
Start the two services init declared:
docker compose up -d
You should see Container revetly-neo4j-your-product Started (as Docker Compose shows it)
Ask the running engine what it can see:
docker compose exec revetly revetly doctor --offline
[PASS] project mount /project is bind-mounted from <YOUR_FOLDER>
Revetly reads this machine's own files and names the key a full instance needs — a lite instance needs none.
--offline and it names the address it tried. Correct NEO4J_URI in .env
at the project root.