Running tests

The Tests sub-tab runs the configuration against sample data and lists past runs.

Sample input files

A test needs an input file of sample telemetry. When you start a run, choose the source:

• My Files. Files you have uploaded.
• Templates. Sample telemetry provided by Telflo.
• Paste Text. Paste a sample directly. Enter a Name and select Use Sample; the paste is saved to My Files as a reusable file. For a logs pipeline fed by a filelog or syslog receiver, paste raw log lines; otherwise paste OTLP/JSON matching the pipeline's signal type.
• Upload. Add a new file from the My Files view.

The Paste Text panel can also generate a sample: describe it in the Generate with AI field, such as the signal type and the shape of the data, and select Generate. A test pipeline must be selected first, because generation is signal-specific.

Run a test

1. Open the Testing tab and select Tests.
2. Select Configure and run to open the Select Test File dialog.
3. Select an input file.
4. Select the Test Pipeline to run it through. The run exercises only this pipeline; other pipelines in the configuration do not run. The pipeline also selects which validation rules are evaluated.
5. Set the Run duration (seconds): 1 to 60, default 5. Increase the duration for pipelines with longer windows, such as batch with a long timeout or tail_sampling, which would otherwise produce no output in time.
6. Select Run.

To repeat a past run, select the re-run action on its row in the history. It replays that run's input file, pipeline, and duration against the configuration as it currently appears in the editor.

If the shape of the sample file does not match the selected pipeline's signal type, Telflo warns you before running: a mismatched file decodes to zero records, so the run would likely produce no output. You can run anyway or cancel.

Run status

Each run is recorded in the history and moves through states as it executes, such as queued, running, and a final status. A run that cannot start, because the testing service is unavailable, rate-limited, or offline, is marked Error rather than left pending. Select a run to view its captured output. Runs started by the AI assistant appear in the history like your own and can be re-run the same way.

A run that captured collector diagnostics shows a Logs link. Open it to read the collector's stderr, which is useful when a run produced no output or fewer records than expected. The input, output, and log viewers each have a Download action.

A run's details also show the configuration the run executed:

• Original YAML. The configuration exactly as it appeared in the editor when the run was triggered.
• Test YAML. The YAML the sandbox collector executed. Receivers and exporters are swapped for sandbox stand-ins so the run is isolated from your real sources and destinations.
• Mutations. The list of sandbox rewrites that were applied, each with a stable identifier, a severity, and an explanation.

Requirements

• A configuration version must be loaded. The run tests the configuration as it currently appears in the editor.
• The run uses the collector version selected in the toolbar.
• $orgvar: references are replaced with their org-default values for the run; variable-group overrides do not apply to test runs. See Vault.

Next steps

• Define pass/fail checks for output. See Validating output.
• Group reusable tests. See Test suites.