Prerequisites
Before installing EVA, ensure that the system requirements and environment setup are met. EVA operates on a Kubernetes-based environment and requires a GPU to leverage various AI models such as ML, LLM, and VLM.
Server Requirements
The recommended specifications are for reference, and the server performance for the App and Agent can be adjusted based on the frequency, importance, and processing speed of detection scenarios.
EVA App
The EVA App uses ML models to detect objects in camera footage and visualize analysis results. The detected results are sent to the EVA Agent for further situational judgment and notification processing.
| Item | Recommended Specification |
|---|---|
| GPU | NVIDIA L4 |
| RAM | 32GB |
| OS | Ubuntu 22.04 |
| CUDA | 12.8 |
Higher specifications allow for connecting more cameras and handling complex detection scenarios quickly and reliably.
💡 The EVA App is responsible for real-time object detection, camera status monitoring, and event log management, rapidly processing and visualizing data generated in the field.
EVA Agent
The EVA Agent is a component that runs an AI engine based on LLM and VLM. It receives event or image data detected by the EVA App, interprets situations, makes scenario-based judgments, and provides alerts using natural language.
| Item | Recommended Specification |
|---|---|
| GPU | NVIDIA L40s |
| OS | Ubuntu 22.04 |
| CUDA | 12.8 |
Beyond simply receiving detection results, the EVA Agent uses VLM (Vision Language Model) to understand visual context and LLM (Large Language Model) to generate natural language descriptions and alert scenarios tailored to the situation.
💡 The EVA Agent is an intelligent module that interprets analysis results from the EVA App and provides natural language-based insights and alerts.
Kubernetes Environment
Since EVA operates in a Kubernetes environment, the following setup is required for cloud or on-premises environments:
- Kubernetes cluster
- GPU-accessible configuration (e.g.,
nvidia-device-plugin) - Ubuntu 22.04-based nodes
- CUDA 12 or higher installed
- Helm CLI (v3.0 or higher)
💡 In a cloud environment, deploying the EVA App and Agent on separate servers can enhance scalability and stability. 🧩 In an on-premises environment, configuring the App and Agent to share GPU resources can create a cost-efficient Edge AI infrastructure.