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

hpc-paraview

hpc-paraview

Build, review, debug, and automate ParaView post-processing and visualization workflows. Use when working with ParaView readers, filters, color maps, screenshots, state files, pvpython or pvbatch scripts, remote pvserver sessions, or cluster-side batch visualization.

Independent GitHub snapshot · Automated classification · Original documentation governs

Documentation de la Skill

GitHub

name: hpc-paraview description: Build, review, debug, and automate ParaView post-processing and visualization workflows. Use when working with ParaView readers, filters, color maps, screenshots, state files, pvpython or pvbatch scripts, remote pvserver sessions, or cluster-side batch visualization.

HPC ParaView

Treat ParaView as a pipeline-driven post-processing and visualization stack with both GUI and scripted execution paths.

Start

  1. Read references/workflow-manual.md before creating or repairing a ParaView workflow.
  2. Read references/readers-filters-and-pipeline.md when mapping datasets to readers, filters, and pipeline objects.
  3. Read references/color-maps-layout-and-views.md when tuning representations, coloring, views, screenshots, or layout behavior.
  4. Read references/pvpython-pvbatch-and-traces.md when choosing between GUI tracing, pvpython, and pvbatch.
  5. Read references/state-files-animation-and-extracts.md when using .pvsm, Python state, screenshots, or saved data products.
  6. Read references/remote-and-parallel-visualization.md when the workflow needs pvserver, reverse connections, SSH tunnels, or distributed rendering.
  7. Read references/cluster-execution-playbook.md when staging a ParaView workflow for scheduler-backed cluster post-processing.
  8. Read references/error-recovery.md when readers, filters, scripts, state files, or remote connections fail.

Reusable Templates

Use assets/templates/ when a concrete starting scaffold is needed, especially:

  • pvpython_screenshot.py
  • pvbatch_slice_extract.py
  • paraview-pvbatch-slurm.sh
  • paraview-pvserver-slurm.sh

Guardrails

  • Do not guess reader or filter semantics from file extensions alone when metadata is available.
  • Do not use pvbatch and remote Connect() logic in the same script.
  • Do not save brittle state files with absolute data paths unless that is intentional.
  • Do not run heavy rendering or batch extraction on login nodes.

Outputs

Summarize:

  • input datasets and readers
  • key filters and representations
  • chosen execution path such as GUI, pvpython, pvbatch, or pvserver
  • expected screenshots, extracts, animations, or saved data products
Skills enregistrées