How to make a System Architecture Diagram
Step-by-step guide on making a System Architecture Diagram
- Write a prompt describing the system architecture. A one-liner can work, but a clearer diagram usually comes from a 3–5 sentence prompt. Use this approach:
- Start by listing the main components – services, data stores, APIs, ML models, pipelines – and group them into layers (e.g. frontend, backend, data, infrastructure).
- Then describe how these components interact. Explain the runtime flows, such as request handling, background processing, or data movement.
- Mention any deployment context – e.g. containerized services, cloud regions, or availability zones – if relevant to the structure.
- Make a diagram with the completed prompt. Use your AI diagram tool to create the diagram based on your description.
- Refine the diagram with follow-up prompts (requires signing in to Eraser). For example: “Group all data components into a ‘Data Layer’” or “Add an API Gateway in front of the backend services.”
- Tweak the layout using GUI controls (requires signing in to Eraser). Adjust spacing, reroute lines, or rename components for clarity.
Tips on making a System Architecture Diagram
- Instead of writing from scratch, paste in snippets from docs, README files, Docker Compose, Kubernetes manifests, or Terraform code.
- Use an LLM to stress test your design – ask it to point out missing components, potential bottlenecks, or failure points based on your initial prompt.
- Call out grouping or layout preferences in your prompt: e.g. “Put inference and monitoring services in separate zones.”
- For complex systems, split into multiple diagrams – e.g. training pipeline vs. serving infrastructure.
Example diagrams