Skip to content

Latest commit

 

History

History
66 lines (46 loc) · 1.96 KB

File metadata and controls

66 lines (46 loc) · 1.96 KB

EnvArt CADMCP Upgrade

DESIGNOSFORGE v1.6.1 upgrades EnvArtBoardOS into a CAD-aware environmental-art workflow.

The goal is simple: CAD truth first, visual expression second.

Capability Model

EnvArtCADMCPBridge routes environmental-art work through CADMCP channels before board styling, image2 prompts, or delivery.

Supported channels:

  • dxf_deterministic: cad_health, dxf_inspect, dxf_audit
  • dwg_core_console: cad_health, scr_write, accoreconsole_run
  • autocad_com_live: cad_health, autocad_state, autocad_open, autocad_send_command, autocad_lisp
  • tianzheng_architecture: cad_health, tianzheng_launch

Geometry Locks

Before environmental-art styling, lock:

  • units, scale, north arrow, site boundary, structural axes
  • outer envelope, walls, columns, openings, stairs and cores
  • circulation paths, room or zone names, dimensions, title block
  • semantic layers and editable text

Output Modes

  • CAD source audit
  • semantic layer report
  • plan-fidelity prompt
  • analysis diagram prompt pack
  • competition board prompt pack
  • construction drawing QA
  • DWG/DXF processing plan

Quality Gates

Do not change CAD topology to improve aesthetics.

Avoid:

  • wall additions, deletions, drift, or deformation
  • door/window symbols pasted over unbroken walls
  • copied source layer 0 habits
  • fake dimensions, north arrows, roads, POI, or room names
  • construction notes used as decorative texture
  • renders replacing missing plan logic

CLI

PYTHONPATH=. python -m app.cli envart-cad plan "用 CADMCP 审核环艺 DWG 平面并生成展板分析图提示词"
PYTHONPATH=. python -m app.cli lora recommend --domain environmental-art --context spatial-cad-production

LoRA Memory

The batch envart-cadmcp-foundation-2026 reserves workflow memory for:

  • CADMCP channel selection
  • source-fidelity and topology locks
  • semantic layer classification
  • construction drawing QA
  • plan-locked image2 prompts
  • board evidence-chain composition