Tutorial: Viewing and Exporting Results

Inspect a completed pipeline run, find and download its outputs, and keep enough context to reuse the results in analysis or writing.

Before you start

You need access to the project and a pipeline run that has produced artifacts. If you have not run one yet, follow Running a Workflow.

Review the run

1. Open Jobs. Use the Completed tab, Search jobs, or the Job type filter (Pipeline run) to find the run.
2. The Job column shows the run's name and ID, such as Pipeline run · #42. Status reads Completed when the worker has finished; Duration shows how long it took.
3. Select the row, or View details and log in its ⋮ menu. The details show when the run was scheduled, started, and finished, its Inputs and settings, and the Processing log. Use Copy log to share messages or errors.
4. Select View artifacts to open the run's Artifacts page.

Find the outputs

The Artifacts table lists every file the run produced, with its Type, Format, Status, and creation time. For each artifact you can View artifact details, Preview artifact content (images and text), or Download artifact. An artifact is downloadable once its status reads ready.

Runs that process many plots can produce hundreds of files. Narrow them with the controls above the table:

• Filter field and Filter value keep artifacts with one exact value, such as Associations · plot_id or a metadata field a module recorded.
• Group by collects artifacts into collapsible groups by a field, such as Metadata · survey, showing how many artifacts each group holds.

The available fields come from the run's associations (project, study, dataset, pipeline run, and plot) and from the metadata each module saved with its outputs. Module authors choose that metadata; see Describe the outputs.

What a run produces

Every run writes its definition, context, and results:

Type	File	Contents
Pipeline setup	pipeline.json	The submitted modules, versions, and parameters
Pipeline Context	context.json	The runtime context the pipeline executed in
Results	pipeline_results.xlsx	The measurements the modules returned, one row per processed plot or file. Results that are not tables are saved as pipeline_results.json.

Depending on the modules, a run can also produce:

Type	File	Contents
Run details	run_metadata.json	Values the modules recorded in the metadata store
Analysis output, or the module's own type	Images, masks, tables, or other files	Files saved by a module, often one per plot. A module can set its own type, such as plot_hsv_image, shown as Plot Hsv Image.
Results	results/*.xlsx	Nested tables a module returned

A batched run writes one results file per batch and a results.json manifest instead of a single workbook.

Keeping pipeline.json alongside the results is what makes a run reproducible: it records the module versions and parameter values, so you can rerun or audit the analysis later without relying on memory. Rerun pipeline in the run's ⋮ menu on Jobs submits the same definition again.

Keep the context

Record the dataset, modules, versions, and parameters with your downloaded files. Before comparing two tables, check units, plot IDs, survey dates, and whether both runs used comparable methods.

The Jobs table shows the project but not the dataset. Read the dataset and modules from Inputs and settings in the run's details, or from pipeline.json. Runs started from Home are named after their dataset; runs started from the Editor are named after the time they started.

Continue with the Agent or MCP

Ask the Agent to retrieve the run and explain a specific output:

Find the results from pipeline run 42 and help me understand the plot measurements.

The run ID is the number shown in the Job column. When the Agent downloads an artifact, a button named after the file appears in the chat; select it to save the file through your browser.

An MCP client uses list_artifacts (filtered by the run ID) and download_artifact. Files downloaded by an MCP client are stored on the MCP host, which may differ from your laptop. The Python and R clients download artifacts with client.artifacts.download(...).

Use the reviewed evidence for downstream analysis or drafting a research section.