L’interface est localisée ; les présentations et documents des Skills sont traduits automatiquement par IA dans votre langue. Le code, les commandes, chemins et termes techniques restent en version originale, et le document de l’auteur fait foi.
Explorer les Skills
Design et création · SKILL.md

visualization-utilities

visualization-utilities

Guides agents using datamol visualization, molecule image rendering, lasso highlighting, dataframe display, parallel utilities, filesystem helpers, RDKit log controls, and environment diagnostics.

Independent GitHub snapshot · Automated classification · Original documentation governs

Documentation de la Skill

GitHub

name: visualization-utilities description: "Guides agents using datamol visualization, molecule image rendering, lasso highlighting, dataframe display, parallel utilities, filesystem helpers, RDKit log controls, and environment diagnostics." disable-model-invocation: true metadata: disco-role: operating license: Apache 2.0

Visualization Utilities

Use this sub-skill when the task asks to render molecules, save SVG/PNG depictions, highlight substructures, display molecule dataframes, visualize existing conformers, run small datamol utility jobs, inspect fsspec paths, silence RDKit logs, or diagnose datamol/RDKit runtime behavior.

Route First

  • For molecule parsing, cleanup, dataframe construction, SDF/CSV/XLSX IO, or property preservation before rendering, use ../molecule-io-prep/ first.
  • For conformer generation, 3D coordinate creation, SASA, alignment, scaffolds, fragments, reactions, tautomers, or isomers, use ../structure-generation/ first.
  • For fingerprints, clustering, distance matrices, MCS, graph comparisons, or diversity selection before visualization, use ../fingerprints-similarity/ first.

What This Covers

  • 2D grids with dm.viz.to_image() or top-level dm.to_image() from SMILES or RDKit molecules.
  • SMARTS, query-molecule, and explicit atom-index highlighting with dm.viz.lasso_highlight_image() and dm.viz.match_substructure().
  • File and notebook output choices, including SVG strings, Pillow PNG objects, and outfile paths handled through fsspec.
  • Existing-conformer display with dm.viz.conformers() when optional notebook widgets are installed and conformers already exist.
  • Utility helpers: dm.parallelized(), dm.parallelized_with_batches(), dm.JobRunner, dm.utils.fs, dm.utils.perf.watch_duration, dm.utils.decorators.disable_on_os, and RDKit log/version helpers.

Reference Map

  • Use references/api-reference.md for signatures, parameters, return types, utility helper notes, and ownership boundaries.
  • Use references/workflows.md for SVG/PNG rendering, lasso highlights, dataframe rendering, parallel jobs, fsspec checks, RDKit log control, and diagnostics recipes.
  • Use references/troubleshooting.md for headless rendering, invalid highlight indices, SVG/PNG/Pillow issues, notebook assumptions, fsspec URI errors, and joblib pickling failures.
  • Run scripts/visualization_utility_smoke.py --help for a safe executable smoke test, or run it with --output-dir <dir> to render a tiny SVG and exercise utility helpers.

Decision Checklist

  1. Normalize input molecules with the root skill or ../molecule-io-prep/ if the task starts from mixed files, tables, or potentially invalid SMILES.
  2. Choose use_svg=True for deterministic text artifacts and use_svg=False only when the caller explicitly needs PNG/Pillow output.
  3. Prefer lasso_highlight_image() for publication-style outline highlights and to_image(highlight_atom=..., highlight_bond=...) or match_substructure() for simple RDKit grid highlights.
  4. Keep parallel helper callables top-level or otherwise picklable when using process workers; use scheduler="threads" or n_jobs=0/1 for closures and notebook workflows.
  5. Wrap noisy parsing or rendering probes in dm.without_rdkit_log() and restore global logs with dm.enable_rdkit_log() after temporary global muting.
Skills enregistrées