Retrieving build snapshots or container images isn't something we're able to do, and that applies to both services regardless of the deployment IDs or digests identified. The one path is whatever survives in the running container itself, accessible via railway ssh or the service's Console tab. Your frontend service uses a multi-stage Dockerfile so only the final nginx stage is present, as you found. However, your backend service (agir-poc) is built with Railpack, not a multi-stage Dockerfile, so the full repo source should still be on disk inside that running container. You can shell in with railway ssh and use tar piped through base64 to pull the files out, since there is no SCP/SFTP.

To correct our earlier reply, we do support SCP and SFTP, so the tar-and-base64 workaround is unnecessary. The easiest extraction path for the backend service (which uses Railpack, not a multi-stage Dockerfile, so the full repo source should be on disk) is railway service files download /app ./local-dir from the CLI, which pulls a whole directory in one step. You can also use scp -r against ssh.railway.com with an SSH key registered on your account, or browse and download files from the service's Console tab in the dashboard. Retrieving the build snapshot itself isn't something we're able to do, regardless of the deployment or digest identified.