Skip to content

Artifacts

When a step writes an output artifact, like a report, a plot or a CSV, Dagweave lists it on the run and can open it in the page. It works with AWS S3 and any S3-compatible store.

Files stay in your store. Dagweave keeps only each artifact’s name, bucket and key. Delete a file from the bucket and the run still lists it, but it won’t open.

This is on by default in the connector’s Helm values:

features:
artifacts: true

What you add is the Secret with your store’s credentials, named in artifactRepoRef on the ArgoConnection:

argoConnection:
extraSpec:
artifactRepoRef:
name: artifact-repo
namespace: argo

Without it, runs list artifact names but you can’t open them.

The Secret needs an endpoint key, even for AWS S3. The connector also reads bucket, region, accessKeyID and secretAccessKey, plus the optional sessionToken, insecure and addressingStyle (path or virtual-hosted). The Secret stays in your cluster, and Dagweave never sees the keys.

Open a run. The Artifacts table, below the steps, lists each artifact’s name, bucket and key, size and content type.

Click a name to download the file straight from your bucket. Click View to open it in the page.

View works for:

  • HTML
  • images: PNG, JPEG, GIF, WebP and SVG
  • PDF
  • text: .txt, .log, .json, .csv, .tsv, .yaml, .yml and .md

Dagweave goes by the file’s contents, not its name. Anything it can’t render, or anything over 20 MiB, you download instead.

The file comes through the connector, so the connector has to be online. Dagweave doesn’t store it.

Every file opens in a sandbox on a separate origin. A script in an HTML report can’t read your Dagweave session, call the API as you or make network requests.

Download links last an hour. After that the table shows link expired. The bucket and key are still right, so you can fetch the file from your store, and View still works.

Anyone on the team who can see a run can view and download its artifacts. Anyone you send a download link to can use it until it expires.

Each View goes in the team’s audit log: who opened which file from which run.

When you run a workflow that declares input artifacts, Dagweave opens Attach input files, with a row per input. Pick a file for each one you want to supply. Leave a row empty to use the workflow’s own source. The dialog shows the size limit. A test run from Test with sample data asks for the same files.

A small text file goes inside the run itself. A larger or binary file goes to the cluster’s artifact store under uploads/{namespace}/{uuid}/{filename}. Dagweave never deletes these, so set a lifecycle rule on the bucket to clear old ones.

Uploads are off by default. To turn them on:

  1. Set features.artifactUploads to true in the connector’s Helm values.
  2. Give the credential in the artifactRepoRef Secret PutObject on uploads/*. Without it the store refuses the upload, and the connector passes on the store’s error.