# YOLOv8 Inference Greengrass component
---
## com.aws.yolov8.inference
- This component is used to run YOLOv8 on an edge device. Whichever model and camera needs to be run has to be identified in the `recipe.json` file. The following shows an example of the `camera_id` and `model_loc` inside the `recipe.json` file:
```
"Configuration":
{
"event_topic": "inference/input",
"output_topic": "inference/output",
"camera_id": "0", OR "samples/video.mp4",
"model_loc": "{edge/device/path/to/models}/yolov8n.trt" OR "{edge/device/path/to/models}/yolov8n.onnx" OR "{edge/device/path/to/models}/yolov8n.pt"
}
```
- Once the component is built, published and a deployment is run on the edge device, the ML model can be managed using MQTT Test Client.
- Subscribe to `inference/output` topic on the MQTT Test Client.
- Publish the status message on `inference/input` to either `start`, `pause` or `stop` inference processes.
- Once the inference is started, the inference results will start showing messages on the published topics.
The MQTT Messages Subscription/Publishing are as follows:
