compLete: AI-Driven SysML Modeling Assistant

compLete is an integrated AI-powered framework that accelerates and automates SysML model development, validation, and completion for automotive systems-engineering projects. It combines:

• Python middleware (Flask + RAG + OpenAI) that exposes a secure REST API to interpret partial SysML models and user requests, retrieve domain knowledge, and generate structured model-edit suggestions.
• Java MagicDraw / Cameo plugin (via the Model Context Protocol) that invokes the middleware, parses AI-generated suggestions, and applies them directly to your SysML model—blocks, ports, connectors, diagrams—in a single undoable session.

---

🚀 Key Features

Capability	Description
Model Validation	Planned: automatically check requirements coverage (e.g., ISO 26262, ASPICE) and flag unsatisfied requirements or missing elements.
Model Completion	AI-driven addition of missing Blocks, Ports, Connectors, and Requirement allocations from partial models and textual queries.
Domain Grounding	Retrieval-Augmented Generation (RAG) using FAISS + SentenceTransformer over a curated corpus of standards, past models, and best practices.
Secure & Containerized	Token-based authentication, TLS on every API call, Docker/Docker-Compose for reproducible deployment.
Seamless Tool-chain Integration	From MagicDraw/Cameo: Tools → AI ▶ Complete Model, then review & apply suggestions with full undo/redo support.

---

📦 Repository Layout

compLete/                     ← Git repo root
├── .gitignore                ← ignores build artifacts, venv, IDE files
├── README.md                 ← this document
│
├── python_middleware/        ← AI API service
│   ├── docker-compose.yml    ← middleware container stack
│   ├── app/
│   │   ├── Dockerfile                ← middleware image
│   │   ├── middleware.py             ← Flask REST API (/complete_model, /health)
│   │   ├── rag_engine.py             ← FAISS + SentenceTransformer retrieval
│   │   ├── utils.py                  ← prompt builder, JSON parser, Java-CLI invoker
│   │   ├── auth.py                   ← token-based HTTPBearer auth
│   │   └── config.py                 ← env-based configuration
│   ├── tests/                        ← pytest units
│   │   ├── test_middleware.py
│   │   └── test_rag.py
│   ├── requirements.txt
│   └── data/                         ← FAISS index & documents
│
└── java_plugin/              ← MagicDraw/Cameo plugin
    ├── build.gradle
    ├── settings.gradle
    ├── gradlew*               ← Gradle wrapper scripts
    ├── gradle/                ← wrapper JAR + properties
    ├── src/
    │   ├── main/java/com/compLete/plugin/
    │   │   ├── CompLetePlugin.java        ← plugin entry point
    │   │   ├── AppConfig.java
    │   │   ├── ConfigLoader.java
    │   │   ├── SysMLModelService.java     ← model I/O, HTTP calls
    │   │   ├── HttpClientHelper.java
    │   │   └── ModelDiagramHelper.java
    │   └── resources/application.properties
    └── src/test/java/com/compLete/plugin/
        └── ConfigLoaderTest.java

Note: the middleware's Dockerfile resides in python_middleware/app/.

---

🔧 Prerequisites

• Python 3.10 +
• Docker & Docker-Compose (for containerised middleware)
• OpenAI API key (GPT-4 or equivalent access)
• Java JDK 17 + (for Gradle wrapper & plugin)
• MagicDraw/Cameo Systems Modeler (with MCP / Teamwork Cloud SDK)

---

⚙️ Setup & Installation

1 · Launch the Python middleware

cd python_middleware

# Copy your FAISS index & docs into ./data/
# Create cert.pem + key.pem for TLS

cat <<EOF > .env
OPENAI_API_KEY=sk-...
FLASK_API_KEY=<secure_token>
EOF

docker-compose up --build

Service starts at https://localhost:5000

curl -k https://localhost:5000/health
# {"status":"ok"}

---

2 · Build & install the Java plugin

cd java_plugin
./gradlew clean build          # downloads Gradle, compiles & tests

# Copy plugin JAR into MagicDraw's plugins folder
cp build/libs/compLete-1.0.0.jar \
   "/Applications/Cameo Systems Modeler/plugins/"

Adjust application.properties inside the JAR (or re-package) to set the middleware URL and API key.

The Python middleware's execute_java_command expects the built plugin JAR at java_plugin/build/libs/compLete-1.0.0.jar, so ensure the Gradle build step has been run before invoking that function.

---

3 · Use compLete inside MagicDraw

1. Open your SysML model.
2. Navigate to Tools → AI ▶ Complete Model.
3. Review AI suggestions, click Apply to insert or Cancel to discard.
4. Save the model.

---

🧪 Testing

# Python tests
cd python_middleware
pytest

# Java tests
cd ../java_plugin
./gradlew test

---

🚢 Deployment

Environment	Notes
On-Premises	Host the Python service behind corporate TLS termination and firewalls.
Cloud / K8s	Deploy the container to Kubernetes or Docker-Compose; secure via ingress TLS.

Configure the Java plugin’s application.properties to point to the correct middleware endpoint.

---

🛠️ Configuration

Variable	Purpose	Default
OPENAI_API_KEY	OpenAI key for LLM calls	—
FLASK_API_KEY	Bearer token for middleware auth	—
OPENAI_MODEL	Model name (e.g., gpt-4)	gpt-4
RAG_INDEX_PATH	FAISS index file	data/faiss.index
RAG_DOCS_PATH	Pickled docs list	data/docs.pkl
OPENAI_TIMEOUT	LLM request timeout (s)	30
SSL_CERT / SSL_KEY	TLS cert/key paths	cert.pem / key.pem
middleware.url	Python service URL (plugin)	—
middleware.apiKey	Token plugin → middleware	—
middleware.connectTimeout	HTTP connect timeout (ms)	5000
middleware.readTimeout	HTTP read timeout (ms)	30000

---

🤝 Contributing

1. Fork the repo & clone.
2. git checkout -b feat/my-feature
3. Implement & test.
4. git commit -m "feat: add …"
5. git push and open a Pull Request.

---

📜 License

Released under the MIT License. See LICENSE for details.

---

📚 References

• ISO 26262 – Functional safety for automotive E/E systems
• SysML v1.x / v2 – OMG systems-modeling languages
• FAISS & SentenceTransformers – semantic retrieval stack
• OpenAI GPT-4 API – LLM engine
• MagicDraw/Cameo OpenAPI & MCP – tool integration layer

---

compLete — Your AI co-engineer for robust, complete, and compliant SysML models.