Weld QC & Pipe Fabrication Skills

_{Developed by the QCDatabase.AI team.}

AI-agent skills (Claude Code / Claude Agent SDK) for the welding, boiler-repair, and pipe-fabrication quality-control world. Each skill is self-contained — bundled Python library, scripts, reference specs, and worked examples in a single .skill file — with no PyPI publishing or network access required for the core tooling. Ready-to-load packages are in dist/.

Works great with QCDatabase.AI — no membership required. Every skill integrates with QCDatabase.AI (via its MCP server) for weld tracking, drawing management, and QC workflows, but each is fully functional on its own: files in, drawings/workbooks/CSVs out.

The skills

1. weldb — Boiler Weld Map Database

Create, edit, render, and export boiler-repair weld maps stored as .weldb YAML files. One file is both the 2D weld-map drawing and the authoritative weld record: properties, append-only revision history, and grid-based views of tubes, membranes, and welds. Describe a panel in plain language and get a validated weld map, a color engineering-drawing PDF, and weld CSVs with on-drawing coordinates — kept in sync automatically ("always render on save"). Panels are archived, never deleted. Converts weld positions to canvas pixels for placement on QCDatabase.AI drawings. Includes 18 worked panel examples plus HTML and Tkinter visual editors.

• GitHub: https://github.com/ProcessQualitySolutions/weldb
• Package: dist/weldb.skill · Deps: fpdf2 (PDF output only) · License: MIT
• Key scripts: create_panel_extended_membrane.py, create_panels.py, save_panel.py, archive_panel.py, render_pdf.py, build_weld_csvs.py, query_welds.py, validate_welds.py, weld_positions_to_canvas.py

2. pcf2isometric — Pipe Spool Manipulation & Isometric Rendering

Parse, inspect, split, merge, and render Piping Component Files (PCF) — the spool format exported by Plant 3D, Revit tools, and Isogen-class systems — into single-line isometric PDFs with BOM, cut list, dimensional callouts, weld dots, and captured weld positions. A rules-based replacement for the abandoned Isogen pipeline: cut lengths are END-POINT arithmetic with explicit weld-gap accounting, spools split at any weld or coordinate and merge with adjacency checks, and every weld's type, joint, NPS, and page position lands in a CSV ready for weld-map upload — with revision reconciliation so weld identity survives re-renders. The parser round-trips byte-identically (verified against 752 production exports). Dimensional callouts persist to an editable .dims.json sidecar — override, suppress, or add, then re-render.

• GitHub: https://github.com/ProcessQualitySolutions/pcf2iso
• Package: dist/pcf2isometric.skill · Deps: fpdf2 (PDF output only) · License: MIT
• Key scripts: init_project.py, pcf_info.py, cut_lengths.py, split_pcf.py, merge_pcf.py, render_iso.py, validate_pcf.py, weld_positions_to_canvas.py, record_tracking_welds.py, reconcile_welds.py

3. welder-lots — Statistical NDE Lot-Acceptance

Group a welder's production welds into statistical NDE lots (RT lot-acceptance per ASME B31.1 / B31.3), evaluate each lot as open or closed, and render an XLSX workbook (cover sheet of open lots plus one tab per welder). Fully deterministic — the same welds, NDE results, and filters always produce the same lots and verdicts — with failure escalation to 100% examination and aging warnings handled by the engine. The random weld draw uses a cryptographic RNG and writes an audit-log line for every selection. Pure-stdlib lot engine backed by a complete behavior spec (welder_lots_system_spec.md).

• GitHub: https://github.com/ProcessQualitySolutions/welder_lots_skill
• Package: dist/welder_lots.skill · Deps: openpyxl (XLSX output only) · License: MIT
• Key scripts: welder_lots.py, build_workbook.py, random_weld_select.py

Installing a skill

Download the .skill file from dist/ (or each project's GitHub releases) and load it into your agent environment (e.g. Claude Code). A .skill file is a zip containing the manifest (SKILL.md), bundled Python library, scripts, references, and examples.

Rebuilding the packages

python package_all.py --dist dist    # rebuild every skill and collect into ./dist
python package_all.py --list         # list the skills that would build

Regenerating this overview

See regenerate_overview.md for instructions on how to regenerate this README when skills are added, removed, or changed.