Documentation

Guide / GitHub deployment

Deploy an app from a GitHub repository

Use the Resilis dashboard to deploy an application from a GitHub repository. Resilis builds and runs the selected source with the settings you review.

Prepare the repository and deployment target

To deploy an app from a GitHub repository, sign in to Resilis with a GitHub account that can access the source. Fork a sample before changing it. Confirm that the selected branch contains the application you intend to run.

Choose the project, environment, and compute target. Keep development and production credentials separate. Set the application root to the directory containing the build inputs: a single dot (.) for a repository-root app, or the relevant subdirectory in a repository containing several apps.

Prepare the dependency manifest, lockfile, start command, required variables, and health endpoint. Test installation and startup locally so that you have a known result to compare with the remote build.

Connect a repository

In the Resilis dashboard, start an application deployment and connect the GitHub repository that contains your app. Select the branch you want to deploy.

Choose Railpack or a Dockerfile

Resilis uses Railpack for automatic builds and supports an explicit Dockerfile strategy. Railpack prepares a build plan from the selected source. Review the detected settings before deploying; detection is a starting point for configuration. A repository containing a Dockerfile can still use Railpack when you explicitly select it, as the Node.js walkthrough demonstrates.

With Railpack, leave command fields empty when the detected defaults fit your application. Use a start override when the service needs a specific command, such as npm start in the task board example. Check that the command starts the production server and that any compiled output exists after the build. An install step cannot substitute for a missing application build step.

With Dockerfile selected, review the Dockerfile path and files available to the build. Keep command overrides empty when you want the checked-in image instructions to control startup. The Docker guide shows this route with the same task board. Choose the strategy deliberately so that later troubleshooting follows the build path you actually used.

Review the application settings

Check the selected source, build settings, environment variables, and compute target before you run the app. Store sensitive configuration in the environment settings instead of committing secrets to the repository.

Configure the listener, variables, and secrets

For a web service, make the process listen on 0.0.0.0 and the port configured for the deployment. A listener bound only to localhost cannot accept requests from outside its container. If your application reads PORT, set it to the intended value and align the app and deployment settings. The recorded samples set PORT=3000; do not assume every target injects an arbitrary port automatically.

Enter the values for required environment variables before starting the app. Importing detected variable names does not supply their values. Mark passwords, API tokens, and connection strings as secrets. Keep them out of the repository and avoid printing the complete environment in build or runtime logs. A setting used during compilation and a setting read by the running process can affect different stages.

For a database-backed application, create and check the database before connecting the app. The PostgreSQL walkthrough places both resources in the same project, environment, and region and uses the internal connection URI as the DATABASE_URL secret. Keep public database access disabled for that example. Its persistence check verifies stored data after an app rebuild.

Deploy and inspect the result

Start the deployment and check deployment status. Use the available logs, metrics, preview, or domain to confirm that the application serves as expected.

Check builds, startup, and application health

Follow deployment status, then inspect build logs when dependency installation or compilation fails. Check the first actionable error rather than treating the final failed status as the cause. For a lockfile installation failure, regenerate and commit the lockfile using the package manager selected for the app. For a missing build or start script, correct the configured command or add the intended script to the manifest in the selected root.

After the build completes, use runtime logs to investigate startup failures, missing configuration, and connection errors. Open the live app and call its health endpoint. In the task board examples, /ready identifies both readiness and the storage mode. Then perform an application action, such as adding and moving a task. A Running status alone does not prove that the page, API, or database interaction works.

Update the app and plan recovery

When the repository webhook and branch deployment trigger are configured, a push to the selected branch can start another deployment. Check the configured branch before expecting a push elsewhere to update this environment. After each update, check the new deployment status, logs, and live result. Keep the source revision with your verification notes so that you can identify what changed.

For a manual rebuild, open Configuration, choose Build, and select Rebuild. Review saved settings first; rebuilding with an incorrect start command or secret repeats the same problem. The sample guides use this action to apply saved demo settings and to test what happens when a new process starts.

App rollback selects a previous deployment; it does not restore database data or reverse migrations. Before a change that writes new data or changes a schema, decide how the earlier app will read that data. Treat data recovery as a separate operation. For the memory-backed samples, added tasks disappear across restarts regardless of which application version you choose.

Common GitHub deployment questions

Why is the repository missing? Check that the connected GitHub account can access it and that you selected your fork. Repository access and branch selection are separate checks. Revisit both before changing build settings.

Why did installation fail? Check the package manager selected for the source, its lockfile, and the application root in the build logs. Commit a consistent manifest and lockfile, then rebuild the corrected source.

Why is the build successful but the URL does not work? Check runtime logs, the start command, required variable values, and the listener host and port. Then call the health endpoint to separate a startup problem from an application request failure.

Will pushing code always redeploy? Push-triggered updates depend on the repository webhook and deployment trigger for the selected branch. Use a manual rebuild when you need to request the build directly, and confirm the source and resulting deployment.

Can rollback recover deleted records? No. App rollback changes the deployed application version. Database restore and migration recovery require their own plan; keep that distinction explicit before a release.

Choose the walkthrough that matches your build and storage needs. The examples use different branches or repositories; keep the source and environment from the selected guide together.

Deploy it with your AI agent instead

A connected AI agent can deploy a local workspace through Resilis over MCP, the Model Context Protocol. This route needs Git and an agent with local file and command access; GitHub is unnecessary. Review the source, requested access, target, build settings, and secrets before authorizing the deployment.

Deployment limits and costs

Custom domains are unavailable on connected compute and S3-backed static sites. App rollback does not restore data or reverse database migrations. Review the target capabilities and usage rates before adapting this example.