Lineage of computed datasets

A published result should answer “what produced this?” without a Slack thread.

Local scripts

cs.set_project("cable-qual")
cs.track_script(__file__)   # opt-in
cs.publish(traces=[result], name="sdd21-serial-6")

cs.track_script() may attach the script file and a short run record (what was read/written, SDK version) to the published dataset. It is off until you call it. Use it for analysis you want to reproduce; skip it for throwaway REPL math.

Project script runs

When you Run a saved script from the Scripts tab (or the assistant’s Run), the run record is the lineage: script name, version, who started it, log, input datasets, output datasets. The result dataset should show as computed and link back to that run.

Workflows

If a workflow produced the dataset, lineage should read like a sentence: produced by run N of extract_s22 v3 via workflow “Post-process snapshots,” triggered by a snapshot from instrument 182. Walk the chain from the dataset; do not keep the only copy of that sentence in a notebook note.

What lineage is not

It is not a substitute for parameters. Still set lot, sn, mixed_mode on the result so filters work. Copy the source serial onto the published mixed-mode dataset so SDD21 sits next to its four S2Ps.