For data teams
Publish notebook exports, dashboards, and recurring analyses as links that update in place, so stakeholders stop asking whether they are looking at the latest run.
The situation
Your pipelines and notebooks already produce HTML. The hard part is delivery: stakeholders want a link, not a notebook; the analysis re-runs on a schedule; and every re-run currently creates a new file, a new URL, or a new attachment that competes with the last one.
What you are publishing
- Notebook exports
- A rendered notebook with charts and narrative, exported to standalone HTML and published without stripping the interactivity.
- Scheduled reports
- A weekly or monthly analysis that should live at one address, not a new address each cycle.
- Model and eval scorecards
- Comparison tables, metric summaries, and run-over-run diffs that reviewers need to open quickly.
What matters most here
One URL across runs
Publish versions against the same artifact so the link in the team wiki always resolves to the current run.
Version history for free
Each run is a version, so you can point at what a metric looked like three cycles ago without keeping a folder of dated files.
Published from the pipeline
One REST call at the end of the job. No manual export, no upload, no separate publishing step for someone to forget.
Interactive output survives
Charts with inline JavaScript render in the sandboxed viewer, so an interactive analysis does not have to be flattened to images.
How the workflow runs
- 01Export the analysis to a single self-contained HTML file at the end of the job.
- 02Publish it on the first run and store the returned artifact id alongside your pipeline state.
- 03On every later run, post a new version against that id.
- 04Link the stable share URL from the wiki, the dashboard, or the scheduled message.
- 05Group related analyses under one project when a stakeholder needs the whole set.
The questions this team asks first
Our notebook exports are large. Will they fit?
Builder accepts up to 10 MB per document and Pro up to 50 MB. Embedded base64 images are usually what pushes an export over the line; downsampling plots or switching to vector output normally solves it.
Can the report query live data?
No. An artifact is a static document rendered in a sandbox. Anything dynamic has to be baked in at publish time, which for a scheduled report is usually what you want anyway.
How do we stop stakeholders seeing sensitive rows?
Aggregate or redact before publishing. Unlisted links are shareable by anyone who has them, and private visibility restricts access to your own account rather than to a named group.