https://github.com/rb248/model-based-se-assistant
AI-powered Model-Based Software Engineering Assistant using LangGraph multi-agent orchestration. Transforms UML diagrams into production-ready code with SOLID analysis, design pattern detection, and comprehensive test generation.
https://github.com/rb248/model-based-se-assistant
ai-agents code-generation design-patterns faiss langchain langgraph mbse rag refactoring solid-principles test-generation uml
Last synced: about 2 months ago
JSON representation
AI-powered Model-Based Software Engineering Assistant using LangGraph multi-agent orchestration. Transforms UML diagrams into production-ready code with SOLID analysis, design pattern detection, and comprehensive test generation.
- Host: GitHub
- URL: https://github.com/rb248/model-based-se-assistant
- Owner: rb248
- Created: 2025-11-16T18:47:10.000Z (about 2 months ago)
- Default Branch: main
- Last Pushed: 2025-11-16T18:52:55.000Z (about 2 months ago)
- Last Synced: 2025-11-16T20:24:31.702Z (about 2 months ago)
- Topics: ai-agents, code-generation, design-patterns, faiss, langchain, langgraph, mbse, rag, refactoring, solid-principles, test-generation, uml
- Language: Python
- Size: 30.4 MB
- Stars: 1
- Watchers: 0
- Forks: 0
- Open Issues: 0
-
Metadata Files:
- Readme: README.md
Awesome Lists containing this project
README
Model-Based Software Engineering Assistant (MBSE Assistant)
=========================================================
A compact multi-agent system that converts UML/PlantUML diagrams into refactored code, tests, and analysis reports. The workflow is designed for clarity and easy extension.
# Overview
## Screenshot
...existing code...
Model-Based Software Engineering Assistant
This repository provides a multi-agent system that turns UML/PlantUML diagrams into refactored code, tests, and architectural guidance.
## Architecture and Workflow
--------
This project uses a LangGraph workflow with multiple agents to:
- parse UML (PlantUML) into an intermediate representation (IR)
- analyze the model for SOLID and design-pattern-based issues
- generate refactored code (Python) with interfaces and dependency injection
- generate tests (pytest) and run them in a sandbox
- attempt to self-correct syntax/import errors
- produce a critique report that includes issues, recommendations, and detected patterns
Key components
--------------
- backend: FastAPI server and multi-agent orchestration
- codegen: code generation and refactoring agent
- testgen: automated test generation agent
- parser: PlantUML to JSON IR converter
- analysis: deterministic detectors + LLM-RAG for context-aware suggestions
Quick start
-----------
Prerequisites
- Python 3.12+
- Node.js for the UI
- Optional: Google Gemini/OpenAI keys for LLMs
Install
```bash
git clone https://github.com/yourusername/model-based-se-assistant.git
cd model-based-se-assistant
python -m venv ag
source ag/bin/activate
pip install -r requirements.txt
cp .env.example .env
# Edit .env to add LLM keys if you have them
```
Run backend + UI
-----------------
```bash
# Terminal 1 - Backend
source ag/bin/activate
uvicorn backend.api:app --reload --port 8000
# Terminal 2 - Frontend (dev)
cd ui
npm install
npm run dev
# Open http://localhost:5173
```
Build & serve UI in production
-------------------------------
```bash
cd ui
npm run build
# optionally copy ui/dist/ to backend/static/ and configure the backend to serve static files
```
Features
--------
- Analysis:
- SOLID violation detection (SRP, OCP, DIP, ISP, LSP)
- Coupling/cohesion measures
- Deterministic detectors + LLM-enhanced recommendations
- `patterns_detected`: heuristics and LLM output (Repository, Service, Router/Controller, Strategy, Factory, Facade, Adapter, Observer)
- `strengths`: things done well in the model (clear naming, repository usage, DI)
- Code generation:
- Splits god classes, extracts interfaces
- Produces multi-file Python project skeleton with services and repositories
- Test generation and execution:
- Generates pytest suites and attempts to run them in a sandbox with timeout
- If tests fail due to syntax/import errors, a self-correction loop tries to fix errors (up to a retry limit)
Configuration
-------------
Edit `backend/config.py` or `.env`:
```python
PROJECTS_DIR=./projects
LLM_PROVIDER=gemini
LLM_FALLBACK_MODEL=gpt-4o-mini
SANDBOX_TIMEOUT=60
```
Development notes
-----------------
- Avoid storing generated artifacts in the repo while the dev server runs (set `PROJECTS_DIR` to an external folder to avoid auto-reloads).
- The `analysis_report` includes `patterns_detected` and `strengths` which you can render in the UI to show "what's good" along with problems.
Tests
-----
```bash
# Backend unit tests
pytest -m "not integration"
# Integration tests (LLM keys needed)
pytest -m integration
```
Roadmap / TODOs
---------------
- Frontend: full dashboard, interactive UML editor, WebSocket progress
- RAG: ingest PDFs and docs, multi-modal RAG
- Multi-language codegen: TypeScript, Java
- DevOps: docker-compose, helm charts
- Security: user auth, sandbox hardening
Contributing
------------
Please fork, add tests, and open a PR.
License
-------
MIT