Documentation is tedious — Engineers spend hours documenting CAD models manually
Knowledge lives in heads — Verbal explanations during reviews aren't captured
CDR prep is painful — Gathering images, writing descriptions, creating BOMs
FEA setup requires context — Atomizer needs model understanding that's often verbal

The Solution

Record yourself explaining your CAD model. CAD-Documenter:

Extracts key frames at scene changes (smart, not fixed intervals)
Transcribes your explanation via Whisper
Analyzes visually using GPT-4o or Claude vision
Correlates visual + verbal to identify components
Generates documentation with images, BOM, and Atomizer hints

Output

📄 Markdown documentation — Structured, version-controlled
📊 Bill of Materials — CSV with components, materials, functions
🔧 Component registry — Detailed specs per component
🎯 Atomizer hints — FEA objectives, constraints, parameters
📑 PDF — Professional output via Atomaste Report Standard

Installation

# Clone
git clone http://100.80.199.40:3000/Antoine/CAD-Documenter.git
cd CAD-Documenter

# Install with uv
uv sync

# Or with pip
pip install -e .

Requirements

Python 3.12+
ffmpeg (for video/audio processing)
OpenAI or Anthropic API key (for vision analysis)

# macOS
brew install ffmpeg

# Ubuntu/Debian
sudo apt install ffmpeg

# Windows (with chocolatey)
choco install ffmpeg

Quick Start

# Set API key
export OPENAI_API_KEY="sk-your-key"

# Run
uv run cad-doc walkthrough.mp4

# With all features
uv run cad-doc walkthrough.mp4 --bom --atomizer-hints --pdf

Usage

# Basic
cad-doc video.mp4

# Custom output directory
cad-doc video.mp4 --output ./my_docs/

# Full pipeline
cad-doc video.mp4 --bom --atomizer-hints --pdf

# Just extract frames
cad-doc video.mp4 --frames-only

# Use Anthropic instead of OpenAI
cad-doc video.mp4 --api-provider anthropic

# Better transcription (slower)
cad-doc video.mp4 --whisper-model medium

Configuration

Create a config file:

cad-doc --init-config
# Creates ~/.cad-documenter.toml

Or set environment variables:

export OPENAI_API_KEY="sk-..."      # Required for OpenAI
export ANTHROPIC_API_KEY="sk-..."   # Required for Anthropic
export CAD_DOC_PROVIDER="anthropic" # Override default provider

See docs/USAGE.md for full configuration options.

Recording Tips

For best results when recording:

Spin slowly — Give the AI time to see each angle
Name components — "This is the main bracket..."
Mention materials — "Made of 6061 aluminum"
Describe functions — "This holds the motor"
Note constraints — "Must fit within 200mm"
Point out features — "These fillets reduce stress"

Architecture

┌─────────────────────────────────────────────────────────────────────┐
│                         CAD-DOCUMENTER                              │
│                                                                     │
│  ┌──────────┐    ┌──────────┐    ┌──────────┐    ┌──────────┐     │
│  │  Video   │───►│  Frame   │───►│  Vision  │───►│  Struct  │     │
│  │  Input   │    │  Extract │    │ Analysis │    │  Output  │     │
│  └──────────┘    └──────────┘    └──────────┘    └──────────┘     │
│       │              │                │               │            │
│       ▼              ▼                ▼               ▼            │
│  ┌──────────┐    ┌──────────┐    ┌──────────┐    ┌──────────┐     │
│  │  Audio   │───►│ Whisper  │───►│ Correlate│───►│ Generate │     │
│  │  Track   │    │Transcribe│    │ Timeline │    │   Docs   │     │
│  └──────────┘    └──────────┘    └──────────┘    └──────────┘     │
└─────────────────────────────────────────────────────────────────────┘

Atomizer Integration

CAD-Documenter generates FEA optimization hints for Atomizer:

cad-doc walkthrough.mp4 --atomizer-hints

Output atomizer_hints.json:

{
  "objectives": [{"name": "mass", "direction": "minimize"}],
  "constraints": [{"type": "frequency", "value": ">100 Hz"}],
  "parameters": ["thickness", "fillet_radius"],
  "critical_regions": [{"feature": "fillet", "concern": "stress_concentration"}]
}

See docs/ATOMIZER_INTEGRATION.md for details.

Python API

from cad_documenter.pipeline import DocumentationPipeline

pipeline = DocumentationPipeline(
    video_path="walkthrough.mp4",
    output_dir="./docs"
)

results = pipeline.run_full_pipeline(
    atomizer_hints=True,
    bom=True,
    pdf=True
)

Project Structure

CAD-Documenter/
├── src/cad_documenter/
│   ├── cli.py              # Command-line interface
│   ├── pipeline.py         # Main orchestrator
│   ├── config.py           # Configuration management
│   ├── video_processor.py  # Frame extraction
│   ├── audio_analyzer.py   # Whisper transcription
│   ├── vision_analyzer.py  # AI vision analysis
│   └── doc_generator.py    # Output generation
├── prompts/                # AI prompts
├── templates/              # Jinja2 templates
├── tests/                  # Test suite
├── docs/                   # Documentation
└── examples/               # Example configs

Roadmap

Core pipeline (frames, transcription, vision)
Configuration system
Atomizer hints extraction
BOM generation
Part Manager integration (P/N lookup)
Interactive review mode
Gitea auto-publish
SolidWorks add-in

License

MIT

Credits

Built by Atomaste for the engineering community.