> For the complete documentation index, see [llms.txt](https://nx.docs.scailable.net/llms.txt). Markdown versions of documentation pages are available by appending `.md` to page URLs; this page is available as [Markdown](https://nx.docs.scailable.net/ai-manager-v6.1.4/support-and-troubleshooting/troubleshooting/plugin-checks.md).

# Plugin checks

Things to check if the plugin is running correctly, and how to check it.

Confirm the [System checks](https://nx.docs.scailable.net/ai-manager-v6.1.4/support-and-troubleshooting/troubleshooting/system-checks.md) pass first.

## Check the Integration Installation <a href="#check-the-plugin-installation" id="check-the-plugin-installation"></a>

Check the device settings page for the "AI Manager" plugin in the Integrations tab. Download the plugin if the tab or the AI Manager entry is missing.

{% hint style="warning" %}
Ensure you download Nx Meta version 6.0 or later to be able to use the AI Plugin.
{% endhint %}

## Is the Plugin Available <a href="#is-the-plugin-available" id="is-the-plugin-available"></a>

When opening the camera settings the Integrations tab is available, and on the Integrations tab the AI Manager is present

<figure><img src="https://2982163321-files.gitbook.io/~/files/v0/b/gitbook-x-prod.appspot.com/o/spaces%2FKGfC0TFSmuMDjbpLErO3%2Fuploads%2Fgit-blob-7aabb66783daa6ef400d1e62b3af5042a7ac84d6%2Fclient-integrations-tab-empty.png?alt=media" alt="Camera Settings dialog with the Integrations tab selected, showing AI Manager listed in the plugin list"><figcaption><p>AI Manager present in the Integrations tab</p></figcaption></figure>

## Is the Plugin Installed <a href="#is-the-plugin-installed" id="is-the-plugin-installed"></a>

If the plugin is not available, verify whether it is installed and check its version.

* Nx Meta version 6.0 or newer is required.
* The plugin requires version 4.0 or higher. This document describes the file layout of plugin version 5.x.

The plugin might not be properly installed. Follow the instructions at [Starting the plugin](https://nx.docs.scailable.net/ai-manager-v6.1.4/ai-manager-plugin/quick-start/starting-the-plugin.md), and check that the plugin file is installed at:`<server-install-dir>/mediaserver/bin/plugins/nx_ai_manager_plugin/libnx_ai_manager_plugin.so`

If the plugin file is not there, or it exists but the server still cannot detect it, reinstall the plugin by following the [Starting the plugin](https://nx.docs.scailable.net/ai-manager-v6.1.4/ai-manager-plugin/quick-start/starting-the-plugin.md) again.

## Are the Needed Drivers Installed: NVIDIA Orin Accelerated <a href="#nvidia-orin-accelerated" id="nvidia-orin-accelerated"></a>

If you're running on NVIDIA Orin - is a current version of [NVIDIA Jetpack](https://nx.docs.scailable.net/ai-manager-v6.1.4/ai-accelerators-support/nvidia-support.md) installed?

## Are the Needed Drivers Installed: Hailo Accelerated Devices <a href="#hailo-accelerated-devices" id="hailo-accelerated-devices"></a>

The runtime using Hailo accelerators requires the correct [Hailo Runtime](https://nx.docs.scailable.net/ai-manager-v6.1.4/ai-accelerators-support/advanced/hailo-support.md) (**hailort**). Currently, the AI Runtime supports Hailo driver versions **4.17.0** to **4.20.0**. Check the installed HailoRT version with `hailortcli -v`.

## Is the AI Runtime Running <a href="#is-the-nx-ai-runtime-running" id="is-the-nx-ai-runtime-running"></a>

The AI Runtime is the hardware-specific inference engine of the AI Manager plugin. It runs inside the Mediaserver, not as a separate process, and starts once a device is set active.

## Check the Plugin Interface <a href="#check-the-plugin-interface" id="check-the-plugin-interface"></a>

If the plugin is not running, the Plugin tab shows a deactivated toggle for the device.

<figure><img src="https://2982163321-files.gitbook.io/~/files/v0/b/gitbook-x-prod.appspot.com/o/spaces%2FKGfC0TFSmuMDjbpLErO3%2Fuploads%2Fgit-blob-ff399a6588a4f048dfcbaf6cdcfd696c0946aebe%2Fclient-ai-manager-device-inactive-intel-runtime.png?alt=media" alt="Camera Settings Integrations tab showing the AI Manager toggle in the off/disabled position"><figcaption><p>Device disabled - toggle is off, device is not sending frames for inference</p></figcaption></figure>

Enable the AI Runtime by clicking the **Device Active** switch.

<figure><img src="https://2982163321-files.gitbook.io/~/files/v0/b/gitbook-x-prod.appspot.com/o/spaces%2FKGfC0TFSmuMDjbpLErO3%2Fuploads%2Fgit-blob-27f177f0f97289600cc8e517769870339f10dbcf%2Fclient-ai-manager-device-active-intel-runtime.png?alt=media" alt="Camera Settings Integrations tab showing the plugin active with Device Active toggle on and active pipeline summary in Device Overview"><figcaption><p>Plugin active with Device Active enabled - the AI Runtime is running</p></figcaption></figure>

## Check the Plugin Logs

The AI Manager does not start separate processes or sockets that can be checked from the command line. The runtime runs inside the Mediaserver, so check the plugin logs instead:

| Log file         | Location                                                                   |
| ---------------- | -------------------------------------------------------------------------- |
| Plugin log       | `<server-install-dir>/mediaserver/var/nx_ai_manager/nxai_plugin.log`       |
| Plugin start log | `<server-install-dir>/mediaserver/var/nx_ai_manager/nxai_plugin_start.log` |
| Runtime log      | `<server-install-dir>/mediaserver/bin/runtime.log`                         |

```bash
tail -f <server-install-dir>/mediaserver/var/nx_ai_manager/nxai_plugin.log
```

{% hint style="info" %}
The pipeline chain is defined in the AI Manager Cloud (the pipelines of the device) and fetched by the plugin. No `chain_*.json` file is written on the server, even when chained pipelines, such as feature extraction, are running.
{% endhint %}

## Is It a Model Problem?

Confirm that the model has something to detect. For example if you're using a model that is expecting vehicles, that will not be detected on a camera that is viewing an empty warehouse.

## Is the Model Working

Check if the model is started by looking into the runtime log at `<server-install-dir>/mediaserver/bin/runtime.log`.

```bash
tail -f <server-install-dir>/mediaserver/bin/runtime.log
```

If there is no output, or the only output is an error message, the runtime is not running the model.

When a model is not supported by the selected accelerator, the runtime falls back to the CPU without showing an error in the client. In that case `runtime.log` contains a warning such as `Compile on … failed …, falling back to CPU`.

## Is the Model Downloaded

Find the model ID on the model detail page in the AI Manager Cloud:

<figure><img src="https://2982163321-files.gitbook.io/~/files/v0/b/gitbook-x-prod.appspot.com/o/spaces%2FKGfC0TFSmuMDjbpLErO3%2Fuploads%2Fgit-blob-63ef3f8f3a580f7230e450138df7246c7870ccdd%2Fmodel-detail-page-id.png?alt=media" alt="AI Manager Cloud model detail page showing the model UUID highlighted in the URL and on the page"><figcaption><p>The model ID (UUID) is shown on the model detail page: use it to locate the cached file.</p></figcaption></figure>

Models are stored in the local filesystem at:

```
<server-install-dir>/mediaserver/var/nx_ai_manager/cache/model_files/
```

Open an SSH session to the server and use `ls` to confirm the file exists:

```bash
sudo ls --full-time -h <server-install-dir>/mediaserver/var/nx_ai_manager/cache/model_files/459d2273-1514-431c-9d34-f5b72f3bfe20.onnx
-rw-r--r-- 1 <server-user> <server-user> 24M 2025-10-21 10:12:52.853826476 +0200 <server-install-dir>/mediaserver/var/nx_ai_manager/cache/model_files/459d2273-1514-431c-9d34-f5b72f3bfe20.onnx
```

The model file is stored encrypted, so its size will be similar to but not exactly match the size shown on the model detail page in the cloud.
