Skip to content

Drawing Analysis

Status: Canonical product, validation, detector-serving, and labeling map Last verified: 2026-07-16 Primary use: Start here before changing any system called DraftLint

Overview

This section separates the product workflow, full validation service, YOLO-serving infrastructure, model-training operations, and historical experiments that have all been called DraftLint.

Use DraftLint alone only for the system family. Use a complete name for a concrete repository, service, or workflow.

What Lives Here

Current Architecture In One Sentence

RapidDraft and the full Railway Validator call one stable IONOS detector gateway; CVAT keeps its native Nuclio AI Tools path; local document processing, OCR, parsing, rules, and report generation run in the Railway Validator; and optional LLM validation calls the configured LiteLLM/local-AI route.

Labeling Working Model

The current v1 workflow is intentionally conservative:

  1. Detect visual objects and regions first.
  2. Run OCR and parsing after those objects are stable.
  3. Store error labels separately from the first-pass detector classes because most drawing errors are relational, not purely visual.

This keeps the first model grounded in observable structure instead of asking it to infer engineering correctness too early.

Scope Decisions For V1

  • Start with mechanical detail and assembly drawings only.
  • Use the 18 object classes from the workbook as the first-pass detector vocabulary.
  • Keep the 5 error types as review tags or relation outputs, not as the primary box classes.
  • Split train/validation data by original sheet, never by crop or tile.
  • Use the PDF examples as reference snippets for consistency, not as exhaustive definitions.

Sources