Reproducibility

Every artifact Promenade produces — a discovered model, a conformance check, a filtered log — records its provenance: which algorithm produced it, with which parameters, from which input artifact, and when. This is visible in the inspector panel for any artifact, not tucked away in a log file.

Provenance matters for research in two directions. Forward, it means a colleague opening your workspace can see exactly how a figure in a paper was produced, without you writing that down separately. Backward, it means when a result looks surprising, you can trace it to a specific parameter choice rather than re-running the analysis from scratch to remember what you did.

Because computation runs client-side against data that stays local, reproducing a result means sharing the input artifact and the recorded parameters — not a server-side pipeline someone else has to stand up.

What’s recorded

• The algorithm identifier and version (for plugin-provided algorithms, the plugin’s own version).
• Every parameter exposed in the inspector at the time the artifact was created.
• A reference to the direct input artifact(s), forming a provenance chain back to the original imported event log.

What’s not automated yet

Promenade does not currently export a single machine-readable “workspace manifest” describing an entire analysis pipeline end to end; provenance is per-artifact. Exporting a full pipeline description is tracked as future work — see the architecture notes for the current state of artifact storage.