A red CI job tells you a collection failed. To find out why, you might need to know which request failed, which data row it used, or whether setup failed before the first check. Noodle 0.10.4 can save those results as a JUnit report for a test viewer or as JSON for a script.

Take a collection from checkout to artifact

For a repository with a collection in ./api, these job steps install Noodle, run the collection, and make the XML report available to download:

- uses: actions/checkout@v6
- uses: noodle-rest/setup-noodle@v1
  with:
    version: v0.10.4
- run: noodle collection run ./api --reporter junit --output results.xml
- uses: actions/upload-artifact@v7
  if: ${{ always() && hashFiles('results.xml') != '' }}
  with:
    name: api-results
    path: results.xml

The two versions do different jobs. @v1 selects the setup action. version: v0.10.4 selects the CLI binary, which the action verifies against the release checksum. The collection command writes its usual summary to the job log. If an assertion fails, the job fails and the upload step still saves the report. The CI guide includes a complete workflow with a staging credential.

Read the failure at the request that caused it

Each selected execution gets one JUnit case. If a setup request fails, you can see the failed setup, the skipped main requests, and the outcome of teardown. If the collection uses a dataset, the case name includes the iteration so you can tell two executions of the same request apart. Each case includes the time spent on the full request lifecycle.

When a downstream script needs response data, use JSON:

noodle collection run ./api --json --output results.json

The file contains exactly the JSON envelope printed to stdout. A graceful interruption preserves partial results in either format and exits with status 130. Choose a fresh report path each time. Noodle checks the destination before the first request and writes only a complete file.

Noodle also masks known secrets when they appear as JSON-escaped strings. JSON reports can still include other server response data. JUnit keeps the request outcomes and diagnostics and omits response bodies.

See the setup action, report formats, and full changelog.