All jobs

Move a Mongoose API from Travis CI to GitHub Actions

Yes. An agent replaces a Mongoose API's .travis.yml with a GitHub Actions workflow running the same lint and Jest commands against MongoDB, and opens a PR.

Run this on RushOpens the console with this prompt ready to send, in a Project that has your repository.

What the agent does

The agent reads the Travis file first, because the job is to keep what it does and move where it runs. In a Mongoose API like hagopj13/node-express-boilerplate, .travis.yml runs on Node 12 with the mongodb service, caches Yarn, builds only master, exports PORT, MONGODB_URL=mongodb://localhost:27017/node-boilerplate, JWT_SECRET, JWT_ACCESS_EXPIRATION_MINUTES and JWT_REFRESH_EXPIRATION_DAYS, runs yarn lint then yarn test, and sends coverage to Coveralls from after_success with yarn coverage:coveralls. package.json defines those scripts: lint is eslint ., test is jest -i --colors --verbose --detectOpenHandles, coverage:coveralls pipes jest -i --coverage --coverageReporters=text-lcov into the coveralls package, and engines asks for Node 12 or newer. yarn.lock is a v1 lockfile and prepare runs husky install.

Then it reads what the tests need, because that decides the service container. src/config/config.js validates MONGODB_URL with Joi and appends -test to it when NODE_ENV is test, which jest.config.js sets, and tests/utils/setupTestDB.js connects with Mongoose, empties every collection before each test and disconnects after all. The three integration suites under tests/integration and the four unit suites under tests/unit therefore need a reachable MongoDB and nothing else. docker-compose.yml already pins mongo:4.2.1-bionic for local development and docker-compose.test.yml runs yarn test against it, so the agent uses the same image tag in CI rather than introducing a second version.

The workflow it writes, .github/workflows/ci.yml, triggers on push to master and on pull_request, which is wider than Travis's branches: only: master and is said so in the pull request. One job runs on ubuntu-latest with a services block starting mongo:4.2.1-bionic and publishing 27017, so the Travis MONGODB_URL works unchanged. actions/setup-node installs Node with cache: yarn, which replaces Travis's cache: yarn, and a matrix keeps 12 from the Travis file and adds a current LTS so the pull request shows whether the suite still passes on a supported release. yarn install --frozen-lockfile with HUSKY=0 skips the Git hook install, then yarn lint and yarn test run with the same five environment variables. Coverage moves from after_success to a step that runs yarn coverage and uploads coverage/lcov.info with the Coveralls GitHub Action using the repository's GITHUB_TOKEN, since the coveralls package's Travis-only token flow does not apply.

Finally it removes .travis.yml and the traces of it: the Travis badge at the top of README.md becomes the workflow's badge, the CI: continuous integration with Travis CI line in the Features list is reworded, and travis leaves the keywords array in package.json. The agent runs yarn install --frozen-lockfile, yarn lint and, with a MongoDB it starts in the sandbox, yarn test, so it has seen the same commands pass before the workflow runs them, and it checks the workflow file with actionlint.

Start it with one API call

POST https://api.phoenix.so/o/{org}/sessions with this body:

{
  "agent": "claude",
  "input": "Move this project's CI from Travis to GitHub Actions. Read .travis.yml, package.json, yarn.lock, jest.config.js, src/config/config.js, tests/utils/setupTestDB.js, docker-compose.yml, docker-compose.test.yml, .eslintrc.json and README.md first. Write .github/workflows/ci.yml that runs on push to master and on pull_request, starts the same mongo image tag docker-compose.yml uses as a service container on 27017, installs Node with actions/setup-node and Yarn caching on a matrix of the Travis Node version and a current LTS, runs yarn install --frozen-lockfile with HUSKY=0, exports the five environment variables from .travis.yml, runs yarn lint and yarn test, and uploads coverage to Coveralls with the Coveralls GitHub Action instead of the coveralls package. Delete .travis.yml, swap the README badge for the workflow badge, reword the Travis line in the README features list and drop travis from package.json keywords. Start a MongoDB in the sandbox and run yarn lint and yarn test yourself, run actionlint on the workflow, and open a pull request that maps each Travis setting to its replacement and says what changed in trigger scope.",
  "resources": [
    {
      "url": "https://github.com/your-org/your-app"
    }
  ]
}
  1. Get an API key. In the console, open Settings, then API keys, and create a key for your org. Send it in the Authorization header as a Bearer token.
  2. Send the request. POST the body in request.json to https://api.phoenix.so/o/{org}/sessions, with your repository as the resources URL. The run starts in its own cloud sandbox.
  3. Read what came back. Follow the run on its live stream or wait for the run.completed webhook, then read the result and its trajectory.

What comes back

What comes back is a pull request that adds .github/workflows/ci.yml, deletes .travis.yml and touches README.md and package.json. The description is a table from each Travis key to its replacement: services: mongodb to the services block, cache: yarn to setup-node's cache, the env block to the job's env, script to two steps, and after_success to the Coveralls action. It says plainly that pull requests from any branch now run CI, where Travis only built master.

The run's record keeps the yarn lint and yarn test output from the sandbox, with the integration suites talking to the MongoDB the agent started there, and the actionlint result, so a reviewer knows the commands pass before the first workflow run.

Before merging, a reviewer confirms the Coveralls repository is enabled for GitHub Actions, decides whether to keep Node 12 in the matrix now that it is out of support, and checks the required status check on master is switched from the Travis check to the new workflow's job name, which the agent names in the description.

Run it on every repository still carrying a .travis.yml

A team moving many repositories off Travis starts one run per repository from its own server: one POST /o/{org}/sessions each, with that repository in resources and a request_id that is a UUID v5 of the repository's full name, so a retried call returns the run already started instead of opening a second pull request. Each run reads its own .travis.yml, starts its own database in its own sandbox, and the pull requests arrive independently.

The same prompt works unchanged across them because it names no Node version or image tag; each run takes those from the repository it is in. A key's webhook_url reports when each run finishes, with the pull request in the payload.

The use case this job follows, with its automation and trigger: Hand off a long task.

Why run it on Rush?

A CI migration is judged by one thing, whether the new workflow passes on the first try, and that depends on details like the test database's name and the lockfile flag that only show up when the commands run. The agent runs them in its own sandbox, with a MongoDB it starts for the purpose, before writing the workflow that runs them again.

Every command the agent ran stays with the run, so the pull request can cite the passing suite and the lint output. Migrating a dozen repositories this way at once costs no more attention than one, because each run has its own machine and its own record.

Frequently asked questions

Will the Jest suite find MongoDB in GitHub Actions?

Yes. A services container publishes 27017 on the runner's localhost, which is the host in the Travis MONGODB_URL. src/config/config.js appends -test to the database name under NODE_ENV=test, so the suite gets its own database inside that container, as it does on a developer's machine.

Why not keep sending coverage with the coveralls package?

On Travis the coveralls package found its token from the Travis environment. On GitHub Actions it needs a repository token the workflow would have to store as a secret. The Coveralls GitHub Action uses GITHUB_TOKEN instead and reads the lcov.info that yarn coverage writes, so no new secret is needed.

Why set HUSKY=0 during install?

package.json runs husky install from the prepare script on every yarn install. It succeeds on a runner but installs Git hooks nobody will trigger. Husky 7 reads HUSKY=0 to skip it, which keeps the install step to what CI needs.

Does the trigger change anything about who runs CI?

Travis built only master. The new workflow also runs on every pull request, including ones from forks, where GITHUB_TOKEN is read-only and the Coveralls upload may fail; the agent marks that step continue-on-error so a fork's pull request still shows lint and test results.

More Mongoose jobs

The same job on other stacks

More jobs like this: Hand off a long task

Keep reading