All three deployments failed at the BUILD_IMAGE step in about 2 seconds with the error: Root directory "backend" was not found in the deployed source. The builder is not stuck or unhealthy. The build was rejected before it reached the builder because the backend directory does not exist at the root of the deployed commit on the integration/mapcaller-accepted-baseline branch. Verify that the branch still contains a backend directory at its root, or update the root directory in your service settings to match the repository structure at that commit.

Thanks. I still need help because the diagnosis appears inconsistent with the actual Git repository state.

I verified the repository directly on GitHub.

The backend/ directory exists at the repository root in BOTH exact commits:

Candidate:

d96ba91b731ab2292fe484e4bb82c6292c84ead3

Canonical:

4aa5fa83574724a67d27b9a670adaeea419dad2c

Therefore Root directory "backend" was not found in the deployed source does not match the actual tree for either SHA.

There is also a second discrepancy:

The failed deployments were intended to deploy the M13 candidate:

feature/m13-legal-registry-acceptance

d96ba91b731ab2292fe484e4bb82c6292c84ead3

but your response says Railway evaluated the source from:

integration/mapcaller-accepted-baseline

Could you please check which exact repository, branch/ref and commit SHA Railway actually resolved for deployment IDs:

• 589cba1e-1478-4209-8dfb-3361c689b228
• f09529e0-b489-447b-ab60-efd5c6fa51e6
• 619a2a02-9a93-4a74-adec-302430b1f46a

Specifically, please confirm:

1. repository resolved by Railway;
2. branch/ref resolved;
3. exact commit SHA resolved;
4. repository root contents Railway saw;
5. why Railway reports backend/ missing when it exists in both relevant Git trees;
6. why the deployment appears to resolve integration/mapcaller-accepted-baseline instead of the intended M13 candidate branch.

At this point this looks like a deployment source/ref resolution issue rather than an application-code or repository-structure issue.

We have stopped further retries until the resolved source/ref is clarified.

The latest failed deployment with that error is a CLI deployment, and it doesn't have a backend folder. The error message is correct, but it's slightly misleading because it mentions a repository in a CLI deployment; we will fix the misleading copy.

Thanks — this resolved the issue.

Your diagnosis was correct: our Railway CLI deployment was using backend/ itself as the archive root while the service already had Root Directory = backend.

We corrected the deployment context and uploaded from the repository root instead, keeping the Railway service configuration unchanged.

The new deployment completed successfully:

0a993dde-42db-48b8-82e1-dde0b6e88ed5

The application started normally and /health now returns HTTP 200.

Thank you for identifying the CLI source-root issue. Your clarification that this was a CLI deployment, rather than a Git repository deployment, was especially helpful.

Also, thank you for noting the misleading repository wording in the original error message — improving that message should make this kind of issue much easier to diagnose in the future.

Great support, much appreciated.