Choosing a platform · By Rivtor · Updated
Monorepo hosting: compare service roots and shared builds
Good monorepo hosting must distinguish service roots from the build context that contains shared packages. Compare dependency installation, build order, environment boundaries and independent releases using the same repository before choosing a platform.
Compare Rivtor plans for your workload →
Identify what must be built together
A repository containing apps/web, apps/api and packages/shared has three source boundaries but may have only two runtime services. A shared TypeScript package is usually a build dependency, not another public service. Draw the import graph before assigning service roots.
If the API installs dependencies inside its own directory while importing a workspace package outside that directory, the build can fail even though local development works. The working directory, workspace lockfile and files available to the builder must agree.
Compare build and release behaviour
| Requirement | Test in the representative repo |
|---|---|
| Workspace install | The committed package-manager lockfile is used |
| Shared package | A clean build can import and compile it |
| Build order | Dependencies build before their consumers |
| Separate runtime | Frontend and API start independently |
| Environment scope | Private values do not enter browser bundles |
Inspect how each platform handles change detection and whether shared-package changes rebuild every dependent service. Build caching and deployment are different systems; a cached result still needs a correct runtime artifact.
Design a clean-room acceptance check
Use a fresh checkout without local node_modules or generated output. Install from the root with the pinned package manager. Build the shared package and each consumer, then start each runtime using only the files its production image contains. A development symlink can hide missing output that the deployed runtime cannot access.
On Rivtor, review detected roots and supported deploy.toml overrides. When a custom Dockerfile is required, make its context and COPY paths explicit. Do not promise support for arbitrary workspace layouts without trying the repository.
Choose for your release pattern
A small frontend/API repo may benefit from one coordinated release. A larger codebase may require independent service promotion, staged compatibility and stronger operational controls. Compare those requirements alongside region and cost. Keep the application structure that serves the team; changing hosts is not a reason to reorganize the repository without evidence.