pattern-forge

AI-driven garment patterns on top of Seamly2D. AI configures validated drafting recipes — it never invents geometry freehand.

Quick start

uv sync                                      # install (Python managed by uv)
uv run pytest                                # verify: all green
uv run python examples/export_trousers.py    # generate your first pattern

Open this folder in Claude Code → enable the pattern-forge MCP server → say:

"Client: waist 84, hips 100, waist-to-floor 107, crotch 83, knee 50 — make classic trousers and show me."

MCP tools

tool	what it does
draft_and_show	one-shot: draft + validate + preview + open in Seamly2D
list_recipes / describe_recipe	discover recipes, measurements + options
draft_pattern	measurements (cm) → validated .sm2d
save_client / get_client / list_clients	reusable client profiles
create_measurements_file	client data → .smis
render_preview	pattern → PNG pages
export_pattern_file	pattern → PDF / SVG / DXF-AAMA
open_in_seamly2d	open or refresh the pattern in the real app

Outputs (patterns, exports, client profiles) go to ~/.pattern-forge/out by default; set PATTERN_FORGE_WORKSPACE to choose another folder.

Recipes

recipe	pieces	status
trousers	front + back + waistband, darts, matched inseams	✅ validated on 8 body types
skirt	A-line front + back + waistband, darts, on fold	✅ validated
aline_skirt	single-piece demo	✅ (Phase-0 demo)

All pieces print with labels, cutting instructions, and grainlines.

Documentation

doc	content
docs/architecture.md	how the engine works, design decisions, layout
docs/trousers-blueprint.md	the trouser drafting method, decoded
docs/skirt-blueprint.md	the A-line skirt drafting method
docs/roadmap.md	phases: done / next

License

Engine code: proprietary. src/pattern_forge/schemas/ contains unmodified XSD files from the Seamly2D project (GPLv3). The Seamly2D application is used as an external binary, unmodified.