Bridging the Gap Between Theoretical Knowledge and Practical System Design.
System design is fundamentally an exercise in navigating trade-offs. Traditional preparation methods rely heavily on static reading materials and passive video consumption, leaving a critical gap in active, applied learning. SystemCraft was built to address this void by providing a high-fidelity, interactive environment where engineers can design, iterate, and receive objective feedback on complex system architectures.
By combining an intuitive architectural canvas with a sophisticated AI reasoning engine, SystemCraft simulates the pressures and constraints of real-world technical interviews, delivering deep insights into scalability, structural integrity, and design trade-offs.
Modern software engineering requires systems that are resilient, scalable, and highly available. Evaluating a candidate's or an engineer's ability to design such systems has traditionally required the time of senior staff. SystemCraft democratizes this process.
Our engine evaluates architectures not just by checking boxes, but by understanding the why behind design decisions. It analyzes the physical connections between nodes to ensure structural soundness, while simultaneously evaluating the qualitative reasoning against industry best practices.
- Interactive Architectural Canvas: A robust workspace engineered for building complex distributed systems. Users drag and drop industry-standard components (load balancers, databases, message queues) to construct their designs visually.
- Dynamic AI Interviewer: A simulated environment where an AI agent assumes the role of a senior engineering interviewer. It introduces live constraint changes, asks probing questions, and forces the user to adapt their architecture in real-time.
- Hybrid Evaluation Engine:
- Structural Linter: Deterministically scans the canvas for physical anti-patterns, such as disconnected subnets, missing redundancy, or single points of failure.
- Qualitative Analysis: Powered by Google Gemini, the engine reads the architecture and the interview transcript to score the design out of 100, outlining explicit strengths, critical weaknesses, and actionable improvement steps.
- Enterprise-Grade Observability: A fully instrumented Prometheus and Grafana stack tracks API latency, error rates, Node.js runtime metrics, and LLM token utilization.
SystemCraft is built on a modern, serverless-first stack designed for speed and reliability.
- Frontend Application: Next.js 15 (App Router), React 19, and Tailwind CSS 4.
- Backend Infrastructure: Next.js API Routes operating in a serverless environment.
- Data Persistence: MongoDB Atlas managed via Mongoose ODM.
- Identity & Access Management: Firebase Authentication supporting OAuth providers.
- Artificial Intelligence: Google Gemini accessed via the Google AI SDK and OpenRouter.
- Telemetry & Monitoring: Prometheus for time-series metric collection and Grafana for dashboard visualization.
SystemCraft is engineered to handle significant concurrent loads. We validated our architecture using k6 load testing, simulating 500 concurrent virtual users hitting the API.
Our load testing revealed the following scaling characteristics when moving from a single container to a dynamically scaled Kubernetes cluster utilizing the Horizontal Pod Autoscaler (HPA).
| Metric | Single Container | Kubernetes HPA (3 Replicas) | Improvement |
|---|---|---|---|
| Total Requests | 23,381 | 61,026 | +161% |
| Throughput (req/s) | ~155 req/s | ~351 req/s | +126% |
| Latency p(95) | 3.33s | 861ms | 74% Faster |
Note: The single Node.js instance experienced event-loop saturation at 155 req/s. By enabling Kubernetes HPA, traffic was distributed across multiple pods, drastically reducing the 95th percentile latency while more than doubling the system's throughput.
xychart-beta
title "System Throughput (Requests per Second)"
x-axis ["Single Container", "Kubernetes HPA (3 Pods)"]
y-axis "Requests / Sec" 0 --> 400
bar [155, 351]
xychart-beta
title "P(95) Latency in ms (Lower is Better)"
x-axis ["Single Container", "Kubernetes HPA (3 Pods)"]
y-axis "Latency (ms)" 0 --> 4000
bar [3330, 861]
Ensure the following dependencies are available in your environment:
- Node.js version 20 or higher
- A provisioned MongoDB Atlas cluster
- A Firebase project with Authentication configured
- An OpenRouter API key (or direct Google AI Studio key)
-
Clone the Repository
git clone https://github.com/Shashank0701-byte/System-Craft.git cd System-Craft -
Install Dependencies
npm install
-
Configure the Environment Duplicate the
.env.examplefile to a new file named.envand populate it with your specific credentials:MONGODB_URL=your_mongodb_connection_string NEXT_PUBLIC_FIREBASE_API_KEY=your_firebase_key # Refer to the .env.example file for the complete list of required variables
-
Initialize the Application
npm run dev
For environments requiring isolation or for production deployments, SystemCraft provides a complete Docker Compose configuration. This includes the core web application alongside the Prometheus and Grafana observability stack.
Please refer to the Docker Setup Guide for detailed instructions on containerized execution.
This software is released under the MIT License. Please refer to the LICENSE file for complete details.
We welcome contributions from the community. To ensure a smooth integration process, please create a dedicated feature branch and submit a Pull Request for review.