CADventory is a 3D CAD file, categorical tagging, and associated metadata inventory management system.
It is tested on Mac, Linux, and Windows, includes code coverage testing, and code security testing.
- Install: CMake 3.25+, SQLite3, BRLCAD, and Qt6
- Clone/Install BRLCAD from https://github.com/BRL-CAD/brlcad.git
- Clone/Download CADventory from source
- Install Ollama and LLaMA 3 (required for AI tagging):
- Download Ollama from: https://ollama.com
- Once installed, run the following command to install the model:
ollama pull llama3
- Compile (see .gitlab-ci.yml for variations): mkdir .build cd .build cmake .. -DCMAKE_INSTALL_PREFIX=/path -DCMAKE_BUILD_TYPE=Release -DQt6_DIR=/path/to/qt6 -DBRLCAD_ROOT=/path/to/BRLCAD cmake --build . --config Release
Run (Mac and Linux): ./bin/cadventory Run (Windows): .\Release\bin\cadventory.exe
Click + in GUI and navigate to a folder to index.
⚠️ AI-based tagging requires Ollama to be running and thellama3model to be available.
To clear all settings, run with --no-gui command-line option.
- Create build directory from root of the project (mkdir build)
- Navigate to build directory (cd build)
- cmake -DCMAKE_BUILD_TYPE=Debug -DBRLCAD_DIR=/path/to/brlcad/build ..
- make
- ctest --output-on-failure
- lcov --capture --directory . --output-file coverage.info
- lcov --remove coverage.info '/usr/*' --output-file coverage_filtered.info
- genhtml coverage_filtered.info --output-directory coverage_report
- view coverage_report/index.html
- Dependencies:
Install CMake: On macOS: brew install cmake On Ubuntu: sudo apt install cmake On Windows: Install from the CMake website. Install LCOV and GenHTML: On macOS: brew install lcov On Ubuntu: sudo apt install lcov On Windows: Install LCOV and GenHTML through Cygwin or equivalent. Ensure make or mingw32-make is available.
- Environment Variable for BRLCAD_ROOT:
You can specify the BRLCAD_ROOT path interactively when the script runs.
- Cross-Platform Commands:
make: Uses mingw32-make on Windows. File Paths: Adapts to Linux/macOS vs. Windows paths (/usr/* vs. C:/Program Files/*). Opening HTML Reports: Uses start, open, or xdg-open based on the OS.
- Testing Workflow:
Compiles the application. Runs unit tests with CTest. Generates and filters a coverage report. Opens the report in the default browser.
- Running the Script Ensure Python is installed (version 3.6+). Run it in your project directory (cadventory): python run_tests.py or python3 run_tests.py
Here's an architecture diagram:
┌──────────────────────────────────────────────────────────────────────┐
│ CADventory │
│ │
│ ┌──────────────────────────────────────────────────────────────┐ │
│ │ User Interface (Qt) │ │
│ └─────────────┬───────────────────────┬─────────────────┬──────┘ │
│ │ │ │ │
│ ┌─────────────▼───────────┐ ┌────────▼─────────┐ ┌────▼──────┐ │
│ │ Filesystem Processor │ │ SQLite or JSON │◄─│ Report │ │
│ └─────────────┬───────────┘ │ Storage Manager │ │ Generator │ │
│ │ └──────────────────┘ └────┬──────┘ │
│ │ │ │
│ ┌─────────────▼───────────┐ │ │
│ │ Geometry/Image/Document │ │ │
│ │ Handler │ │ │
│ └─────────────┬───────────┘ │ │
│ │ │ │
│ ┌─────────────▼────────┐ │ │
│ │ CAD Libraries │◄───────────────────────────────┘ │
│ └──────────────────────┘ │
└──────────────────────────────────────────────────────────────────────┘
- AI Powered Model Tagging (ModelTagging.cpp/.h)
- Tag based filtering/searching
- Logging System for AI
- Window compatible command execution
- Major UI/UX Improvements
- Major bug fixes (e.g report generation and image/thumbnail issues)
/src
├── main.cpp # Entry point for GUI
├── ModelTagging.* # AI tagging system using Ollama + LLaMA 3
├── ModelParser.* # BRL-CAD model parser for title + object names
├── executeCommand.* # Platform-specific command execution
├── UI files (.ui, etc) # Qt user interface logic
/tests # Unit tests
/scripts # Automated testing tools
- Checks for Ollama + model
- Builds prompt and runs AI tagging
- Uses QProcess to write/read tag results
- Signals:
tagsGenerated,tagGenerationCanceled
- Uses
mgedCLI to extract title and object paths - Converts file paths to WSL-compatible if needed
- Windows-safe hidden execution of commands
- Also supports redirection for prompt generation
- Powered by LLaMA 3 (via Ollama)
- Parses .g BRL-CAD files for object paths and titles
- Sends structured prompt to local LLaMA model via command-line interface
- Returns exactly 10 one-word tags for classification/search
- Ensure Ollama is installed and added to your
PATH - Call ModelTagging::generateTags(filepath) to begin tag generation
- Automatically emits a tagsGenerated(tags) signal when complete
- Implement automatic PDF report generation (1-page per library model).
- Integrate metadata support into GUI for tagging primary models.
- Graphically depict models and iconography during GUI browsing.
- Implement support for creating & modifying collections manually.
- Implement support for identifying between AI generated and manually added tags
- Multi-threading support for tag generation
- Refine accuracy of the AI Tagging
- AI tagging logic: Written in-house using prompts run via Ollama
- BRL-CAD Integration: Leveraging mged CLI from BRL-CAD
- AI backend: Powered by LLaMA 3 via Ollama
- Qt GUI: Built using Qt6
MIT