Feat: last pipeline run section in dashboard

[anuunchin]

This PR introduces a new pipeline run section to the dashboard app which uses a simple html snippet.

[cloudflare-workers-and-pages]

Deploying with    Cloudflare Workers

The latest updates on your project. Learn more about integrating Git with Workers.

Status	Name	Latest Commit	Preview URL	Updated (UTC)
✅ Deployment successful! View logs	docs	e1a163c	Commit Preview URL Branch Preview URL	Nov 14 2025, 06:25 PM

[rudolfix]

terminology (from https://www.notion.so/dlthub/dlt-Entities-their-Lifecycles-and-Relations-2679fb8e23cf80169921c4ea069bfff8?source=copy_link#2679fb8e23cf8109b9c3eee928a2a377). definitions are pretty clear and you can easily refer to code

• pipeline run succeeded or failed (correct your status)

Idea display status of load package(s) as boxes like the "COMPLETED"

• pipeline created load packages: collect packages from all steps to get load id and package
• display color coded states of all packages (with load ids on it): pending | loaded | aborted
• package can be "partially loaded" (I think in pending and aborted states). there's a method on package info. I hope we have it in trace. if you see it - it should be a red color ie: "load_id: partially loaded in aborted"

schema indicators

• look for "applied schema update" if it is found in the trace - display "migration" box

for later

• detect state sync (loaded job with state table name)
• detect dev_mode (I do not think we have it in trace but we could use dataset name)
• detect full refresh type (I do not think we hace it in trace)

feedback to current UI
experiment: if we have trace just put it on top, without "Pipeline Run" and switch. Also you do not need to repeat pipeline name, it is already on top

[rudolfix]

LGTM. all the comments are minor. visually is really good.

you could add example pipeline with partially loaded package:

@dlt.destination
    def wait_forever(item, schema):
        # never exit
        from time import sleep

        # this one does not wake up on signal
        while not _done:
            sleep(1)

        # make the job fail if it gets here
        raise KeyboardInterrupt()

    pipeline = dlt.pipeline(
        "signal_waiter", destination=wait_forever(), dataset_name="_data", progress=DictCollector()
    )

you can modify this destination so it passes for one table but raises for another triggering "partially loaded" state.

[rudolfix]

we have humanize library. it should be able to generate friendly strings. ie humanize.precisedelta pls take a look

[anuunchin]

humanize doesn't seem to be generating friendly enough strings - I think it's better to leave it the way it is currently 👀

[rudolfix]

LGTM!