Tunny Icon
TunnyDocs

The next-gen Grasshopper optimization tool.

Artifact Gallery

The Artifact Gallery displays artifacts belonging to trials in the current study, associating each file with its trial and with analysis results. Use the Artifacts toolbar button to select an artifact root directory, or set it with -a / --artifact <directory>. For a local Journal, an adjacent artifacts/ directory is discovered automatically when no explicit root has been selected. Artifact scans are shared across studies in the open storage; switching studies does not clear them. The gallery displays only trials in the current study's view.

View modes

  • All: trials with an artifact at the selected index, sorted by creation order and shown as a paginated thumbnail grid. An active selection filters these trials.
  • By Cluster: runs k-means clustering with your settings (k / target space / selection mode / init) and groups artifacts into collapsible cluster sections, with pagination.
  • By MCDM Rank: runs the selected MCDM method and lists artifacts in ranking order (Top-N), with the rank and score shown on each card.

Artifact selection

The shared Artifact # selector chooses the zero-based artifact entry index shown per trial in all three modes. It is not an artifact count or a count of unique MIME types. Each trial contributes at most one card; trials without an entry at the selected index are skipped.

The range is derived only from trials in the current study's view, not from other studies in the storage or from the active selection filter. The of up to label shows the maximum selectable index, exactly matching the selector's upper bound. For example, a study with at most two entries per trial offers indices 0 and 1 and displays of up to 1.

When the study has at most one entry per trial (including no artifacts), the selector is hidden and index 0 is used. Switching studies preserves a valid index; an out-of-range index is clamped to the new study's maximum valid index.

Cards

Card labels use the trial's zero-based creation-order number within the current Study, matching Trial Table and trial details. Each Study starts at trial 0; failed or pruned trials can leave gaps in the displayed numbers. Filtering, grouping, and pagination do not renumber trials. Journal-wide trial IDs are used internally to associate files and interactions, not as card labels. If a trial number is unavailable, the label uses its row index in the Study view.

Image files (PNG / JPEG / GIF / WEBP) are shown as thumbnails. Clicking an image opens the trial detail modal, including its artifacts. Other files (CSV, etc.) show an icon and an Open button that opens the file with the system default application. Clicking a card's trial label highlights the corresponding trial across other widgets. Each card also shows the trial's objective values. Size selects Small, Medium, or Large thumbnails.

Comparing settings (A vs B)

Clustering and MCDM results are cached per setting. To compare, for example, k = 3 against k = 5, place two Artifact Gallery widgets on the canvas and give each its own setting. Each gallery reads its result from the shared cache, so both configurations can be viewed side by side. A setting already computed by another Clustering / MCDM widget is reused; an uncomputed setting can be run from the gallery itself with Run.