Hi Community,
I sent the below message to Replit Support who have escalated to an engineer, and I am waiting…I was wondering if somebody may be able to help…
My production Autoscale deployment has been failing at the Build stage for the last several hours on a pnpm-workspace, artifact-mode repl that was deploying cleanly up until about 11 hours ago.
Last successful deployment revision: e060a73e (published ~11 hours ago at the time of writing)
Recent failed revisions (sample): 278a4713, 988b1b31, 4903d1ee, 2d046808, 2de594f2, plus the most recent failure ~1 minute ago.
The Build stage fails with these bullets (wording has varied slightly across attempts, which is itself notable):
- “No build command is configured for the deployment”
- “No run command is configured for the deployment”
- “The build process defaults to running ‘pnpm’ without arguments, which only displays help text and does not build the application”
Earlier failures in the same incident showed slightly different wording such as “No build command provided for the monorepo project” and “no entry point is configured,” which makes me think a pre-flight validator has been updated.
Repo configuration (nothing unusual, and essentially unchanged from the last-green deploy):
.replit registers [[artifacts]] id = “artifacts/api-server” and [[artifacts]] id = “artifacts/mockup-sandbox”, with [deployment] set to router = “application”, deploymentTarget = “autoscale”
no top-level build/run overrides (by design, so per-artifact [services.production] is used).
artifacts/api-server/.replit-artifact/artifact.toml has a complete [services.production] block: build = [“bash”, “-lc”, “pnpm install --frozen-lockfile && pnpm --filter db run push-force && pnpm --filter @workspaceworkspace/api-server run build”], run = [“node”, “–enable-source-maps”, “artifacts/api-server/dist/index.mjs”], health startup /api/healthz, PORT 8080.
Root package.json also has “build”: “pnpm --filter @workspace/api-server run build && pnpm --filter db run push-force” and “start”: “node --enable-source-maps artifacts/api-server/dist/index.mjs” (added by the Replit Agent in commits 6a370b4 and eb3d583 specifically to try to unblock the same failure, commit eb3d583’s message even diagnoses the bare-pnpm help-text output as a shell-glob expansion issue inside the deploy wrapper).
artifacts/ioss-portal and artifacts/salesvat-website are static frontends with their own .replit-artifact/artifact.toml containing correct [services.production] build/publicDir/serve = “static” blocks; they’re served statically behind the api-server artifact.
Git evidence that nothing human-authored broke this: between the last-green revision ~11h ago and now, only Agent commits touched deploy-relevant files — 0096622 and 8b4af7c (Task #203, schema sync non-interactive in api-server artifact.toml), 6a370b4 (add root start/build scripts), and eb3d583 (simplify root build). No edits to .replit since the last-green deploy.
I have not modified .replit myself (I made one exploratory change earlier in this session and then reverted it byte-for-byte; a backup .replit.bak-20260418-213308 is still in the workspace and diff .replit .replit.bak-… comes back empty).
Given that the same configuration deployed successfully 11h ago, that the Agent has already diagnosed and attempted to fix the bare-pnpm symptom at the root package.json level without success, and that the validator’s error wording has shifted between attempts in the same hour, this looks like a resolver/pre-flight regression for pnpm-workspace artifact-mode Autoscale deployments rather than a config issue on my end. Could you take a look at the Build stage traces for the recent failed revisions listed above and confirm whether the validator recently changed behaviour for this deploy pattern
Thanks,
Gurpreet