Open Source// CTQ-EXTRACTION
CTQ-Extraction
AI agent skill extracting Critical-To-Quality parameters, GD&T feature control frames, and verbatim tolerances from engineering drawings into validated JSON.
PythonMultimodal LLMsJSON SchemaClaude CodeAntigravity
ASME / ISO
Standards
14
GD&T Symbols
The Problem & Operational Context
Manufacturing quality inspection requires manual transcription of Critical-To-Quality (CTQ) dimensions, datums, and geometric tolerances from complex CAD blueprints, creating an error-prone bottleneck.
Key Architectural Decisions
Designed a cross-platform agent skill using multimodal vision models constrained by strict JSON Schema validation to parse ASME Y14.5 and ISO 1101 geometric tolerances verbatim.
Interactive Architecture Topology
Multimodal AI Extraction & ASME Y14.5 Schema Pipeline
Standards: ASME Y14.5 / ISO 1101
ASME Y14.5 / ISO 1101 Verified Symbols14 Standard Symbols
⌖Position⏥Flatness⌯Cylindricity⏊Perpendicularity∠Angularity◎Concentricity↗Circular Runout⌭Total Runout
Multimodal LLM Vision
(Spatial Symbol Reasoning)Quality Gate:14 GD&T Symbols Detected
Segments and crops high-resolution drawing regions; recognizes geometric symbols (position, flatness, runout), diameter symbols (⌀), and modifier letters (Ⓜ, Ⓛ).
Pipeline Step:Claude Vision • Spatial Bounding • OCR•Function:Visual identification of GD&T feature control frames
Key Architecture & Capabilities
- Automated verbatim extraction of 14 ASME Y14.5 / ISO 1101 GD&T symbols, datums, and tolerance zones into structured JSON.
- Zero hallucination on tolerance bounds via strict post-extraction schema and range validation.
- Cross-platform compatibility across Claude Code, Antigravity, and Cursor agent environments.