CmdGen – AI-Powered Infrastructure Automation Platform

A polyglot full-stack application that translates natural language into executable shell commands using offline Vector Embeddings.

---

Overview

CmdGen is a tool designed to bridge the gap between human intent and system execution. Unlike simple API wrappers, CmdGen utilizes a local semantic search engine powered by Sentence-Transformers (BERT) to map natural language queries (e.g., "Find all files larger than 100MB") to precise Linux commands.

This project demonstrates a Microservices Architecture where a Java Spring Boot gateway orchestrates communication between a high-performance React UI and a Python Inference Service.

Architecture

The system consists of three decoupled microservices:

1. Frontend (React + Vite): A reactive UI for user interaction and real-time command visualization.
2. Backend (Java Spring Boot): Acts as the API Gateway and Business Logic layer, handling validation and routing.
3. Inference Engine (Python FastAPI): A dedicated AI service that uses Vector Embeddings (Cosine Similarity) to match queries against a dataset of 500+ optimized shell commands.

---

🛠 Tech Stack

Component	Technology
Frontend	React.js, Vite, Tailwind CSS (optional)
Backend	Java 17, Spring Boot 3, Maven
AI / ML	Python, FastAPI, Pandas, Scikit-Learn, Sentence-Transformers (HuggingFace)
Data	CSV (Flat file), Vector Embeddings (In-memory)

---

Getting Started

Follow these steps to run the complete system locally.

1. Prerequisites

• Python 3.9+
• Java JDK 17+
• Node.js & npm

2. Setup AI Service (Python)

The AI service must be running first to handle inference requests.

cd model_service

# Create and activate virtual environment
python -m venv .venv
# Windows:
.venv\Scripts\activate
# Mac/Linux:
source .venv/bin/activate

# Install dependencies
pip install -r requirements.txt

# Run the service (Port 8000)
python ai_service.py

Note: On the first run, it will automatically download the all-MiniLM-L6-v2 model.

3. Setup Backend (Spring Boot)

Open a new terminal.

cd backend

# Run using Maven Wrapper (Port 8080)
# Windows:
./mvnw.cmd spring-boot:run
# Mac/Linux:
./mvnw spring-boot:run

4. Setup Frontend (React)

Open a new terminal.

cd ui/cmdgen-frontend

# Install dependencies
npm install

# Run the development server
npm run dev

---

How It Works (The "Offline AI" Engine)

Instead of relying on slow and expensive external APIs (like OpenAI), CmdGen uses Semantic Search:

1. Initialization: On startup, the Python service loads linux_commands.csv and converts all descriptions into high-dimensional vectors using all-MiniLM-L6-v2.
2. Querying: When a user types a request, the input is converted into a vector.
3. Matching: The system calculates the Cosine Similarity between the input vector and the dataset vectors.
4. Result: The command with the highest similarity score is returned with <200ms latency.

---

📂 Project Structure

CmdGen/
├── backend/                 # Spring Boot Application
├── model_service/           # Python FastAPI AI Service
│   ├── ai_service.py        # Inference Logic
│   └── requirements.txt     # Python dependencies
├── ui/cmdgen-frontend/      # React Application
├── linux_commands.csv       # Dataset for command mapping
└── README.md                # Documentation

👤 Author

Pranay Kelotra