OmniAPI

English | Tiếng Việt

OmniAPI is a high-performance REST API Server designed to make Edge AI effortless for developers. Whether you're building a Mobile App, an intelligent AI PC, or seamlessly integrating advanced analytics into a customer's Video Management System (VMS) or Enterprise Server – OmniAPI lets you execute state-of-the-art AI workloads in seconds using simple HTTP & JSON.

Zero AI Knowledge Required. No need to learn about tensor shapes, ONNX model conversions, PyTorch configurations, or hardware dependencies. If you know how to make an API call, you know how to deploy Vision AI.

---

⚡ What is OmniAPI?

Built on top of the robust CVEDIX EdgeOS engine, OmniAPI abstracts complex GStreamer media pipelines and lower-level inference engines into straightforward, universal REST API calls.

By acting as a universal translation layer, OmniAPI empowers application teams to integrate complex Vision AI features (Face Recognition, License Plate Reading, Behavior Analysis) directly into their applications in a flash, drastically reducing integration time from weeks to hours.

🌟 Key Developer Benefits

• Effortless Integration: Call our REST API from any platform: iOS, Android, Web UI, Desktop software, or Backend Servers.
• Zero AI Knowledge Required: You send the video stream URL (RTSP/RTMP/File) and tell OmniAPI what you want to detect. OmniAPI handles the AI pipelines, object tracking, scaling, and bounding boxes internally.
• Production-Ready Accuracy: Ships with pre-trained, high-accuracy models that are robust and ready-to-use out-of-the-box in real-world environments.
• VMS-Ready: Seamless integration with existing Video Management Systems (Milestone, Exacq, Nx Witness) via standard streams and HTTP Webhooks.
• Hardware Agnostic (Write once, run anywhere): OmniAPI automatically optimizes processing for the underlying NPUs/GPUs (NVIDIA, Rockchip, Intel, Hailo, Qualcomm). No custom code for different hardware!

---

🔌 Integrating in a Flash

Everything in OmniAPI is designed with an API-first mindset. Start an advanced Vision pipeline instantly using simple JSON payloads.

🟢 Tier 1 — Zero Knowledge (Just pick a category)

Turn any camera into a smart AI node — no model configuration needed.

curl -X POST http://localhost:8080/v1/securt/instance \
  -H "Content-Type: application/json" \
  -d '{
    "name": "Lobby Camera",
    "category": "security",
    "input": { "url": "rtsp://admin:pass@192.168.1.100/stream", "type": "rtsp" },
    "output": { "url": "rtmp://localhost:1935/live/lobby", "type": "rtmp" },
    "autoStart": true
  }'

🔵 Tier 2 — Choose a Specific Feature

Select a targeted AI feature within a category for precise analytics.

# Traffic jam detection — just pick category + feature
curl -X POST http://localhost:8080/v1/securt/instance \
  -H "Content-Type: application/json" \
  -d '{
    "name": "Intersection Monitor",
    "category": "its",
    "feature": "jam",
    "input": { "url": "rtsp://10.0.0.50/stream", "type": "rtsp" },
    "autoStart": true
  }'

🟣 Tier 3 — Custom Model (Expert)

Bring your own trained model — full control for AI researchers and engineers.

curl -X POST http://localhost:8080/v1/securt/instance \
  -H "Content-Type: application/json" \
  -d '{
    "name": "Custom Fire Detection v2",
    "category": "custom",
    "solution": "fire_smoke_detection",
    "input": { "url": "rtsp://10.0.0.100/stream", "type": "rtsp" },
    "additionalParams": {
      "MODEL_PATH": "/opt/models/fire_v2.weights",
      "CONFIG_PATH": "/opt/models/fire_v2.cfg",
      "LABELS_PATH": "/opt/models/fire_labels.txt"
    },
    "detectionSensitivity": 0.35
  }'

Python / Mobile / Backend Example

Build security analytics natively into your Backend or App in 5 lines of code.

import requests

API_BASE = "http://localhost:8080/v1/securt"

# 1. Start a security instance (zero AI knowledge needed!)
response = requests.post(f"{API_BASE}/instance", json={
    "name": "Perimeter Camera",
    "category": "security",
    "input": {"url": "rtsp://192.168.1.100:554/live", "type": "rtsp"},
    "autoStart": True
})
instance_id = response.json().get("id")

# 2. Add an intrusion line dynamically
requests.post(f"{API_BASE}/instance/{instance_id}/lines", json={
    "name": "Fence Zone",
    "coordinates": [{"x": 100, "y": 500}, {"x": 1800, "y": 500}],
    "direction": "Up",
    "classes": ["Person", "Vehicle"]
})
print("Intrusion line armed successfully!")

---

🤝 Supported Hardware Acceleration

OmniAPI automatically leverages available hardware accelerators to maximize inference FPS while minimizing power consumption.

Vendor	Specific SOC / Family	Acceleration Backend
✅ NVIDIA	Jetson AGX Orin, Orin Nano, RTX GPUs	TensorRT
✅ Rockchip	RK3588 (OPI5-Plus)	RKNN
✅ Hailo	Hailo-8 (1200 / 3300)	HailoRT
✅ Qualcomm	QCS6490 (DK2721)	SNPE / QNN
✅ Intel	Core Ultra (R360)	OpenVINO
✅ AMD	Ryzen 8000 (2210)	Vitis AI

---

🚀 Quick Start

1. Installation

Recommended: Debian ALL-IN-ONE Package
The easiest way to get started is by utilizing our pre-built standalone .deb package containing the API server and all required edge dependencies.

# Install the downloaded package
sudo dpkg -i omniapi-all-in-one-*.deb
sudo apt-get install -f

# Start the OmniAPI daemon
sudo systemctl start omniapi

(For manual build instructions, please refer to INSTALLATION.md)

2. Verify Server Status

# Check if the API is running correctly
curl -s http://localhost:8080/v1/securt/health | jq

---

🎯 Solution Categories

OmniAPI organizes 43+ optimized edge processing nodes into solution categories. Just pick a category — no need to manage models:

Category	Default Feature	Available Features	Use Case
🛡️ security	SecuRT (full pipeline)	crossline, intrusion, loitering, crowding, face	Enterprise security, perimeter defense
🚗 its	Crossline counting	line_counting, jam, stop, wrong_way, obstacle	Smart traffic, intelligent transportation
🔥 firefighting	Fire/Smoke detection	fire, smoke	Fire safety, industrial monitoring
🎯 armed	SecuRT (full pipeline)	—	Military / defense applications
🔧 custom	(user-defined)	(requires explicit solution + model paths)	Research, custom-trained models

---

📚 Ecosystem & Documentation

• Swagger / OpenAPI Spec: View the interactive API sandbox at http://localhost:8080/swagger
• API Reference: Complete documentation on every endpoint.
• Architecture Guide: Deep dive into how OmniAPI communicates with the underlying AI Runtime and EdgeOS SDK.
• Development Guide: How to structure code, write new controllers, and build custom models.
• Environment Variables: Configure ports, thread limits, logging, and database paths.

---

📝 License

Proprietary - CVEDIX