> 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/ai-manager-plugin/advanced-configuration/ini-file-contents.md).

# .ini file contents

Description of the .ini configuration file options.

## Nx SDK Logging

If enabled, the NX SDK logging will be logged to console. Default: false

`#enableOutput=false`

## Console Logging

If enabled, the AI Plugin, AI Manager and Inference Engine logs will be logged to console. Default: false

`#logToConsole=false`

## File Logging

If enabled, the AI Plugin and AI Manager logs will be logged to files. Default: true

`#logToFile=true`

## Override Inactive AI Manager

If enabled, the AI Plugin will send frames to the AI Manager even when inactive. Default: false

`#sendFramesOverride=false`

## Automatically Restart on Crash

If enabled, the AI Plugin automatically restart the AI Manager if it crashes. Default: true

`#restartAIManagerOnCrash=true`

## Cache Runtimes

If enabled, the AI Plugin will cache runtime archives. To force re-download, delete the relevant runtime archive. Default: false

`#cacheRuntimeArchives=false`

## Ambarella Integration

If enabled, the AI Plugin will enable Ambarella integration, ie. offloading model inference to the Ambarella CV75 chip. Default: false

`#enableAmbarellaIntegration=false`

## Nvidia TensorRT Integration

If enabled, the NxAI Plugin will enable the NVIDIA TensorRT runtime. Default: false

`#enableNvidiaTRTIntegration=false`

## Profiling

If enabled, the NxAI Plugin log profiling details. Default: false

`#enableProfiling=false`

## Always Show Recovery Buttons

If enabled, recovery buttons (reselect runtime, clear cache, etc.) are always shown in the device settings, even when there are no errors to recover from. For debug purposes. Default: false

`#alwaysShowRecoveryButtons=false`

## Override API Cloud URL

The URL of the NXAI cloud. Default: "<https://api.sclbl.nxvms.com/dev>"

`#CloudAPI="https://api.sclbl.nxvms.com/dev"`

## Override Admin URL

The base URL of the device admin. Affects popup browser window. Default: ""

`#DeviceAdminOverride=""`

## Override Runtime Storage URL

URL of the storage containing the runtimes. Default: ""

`#RuntimesStorageURL=""`

## Override AI Manager Download URL

URL of where to download the AI Manager. This link should only be to the directory; the filename will be programatically determined. Default: ""

`#AIManagerStorageURL=""`

## Override Working Directory

Directory the plugin should use as its working directory. This includes logging files, AI Manager and model file downloads. Default: ""

`#WorkingDirectory=""`

## Loglevel

1. No logging.
2. Basic logging.
3. Verbose logging.

Default: 1

`#logVerbosityLevel=1`

## Download Timeout

This sets the timeout for the request when downloading a model. Increase this in environments with slow or unstable networking. Default: 5

`#modelDownloadSeconds=5`

## HTTP Timeout

This sets the timeout for HTTP requests to the NxAI cloud API in seconds. Default: 3

`#HTTPRequestTimeout=3`

## Parameters Refresh Timeout

This sets the timeout in seconds to wait for the runtime manager loop to process an immediate devices parameters refresh request. Default: 5

`#DeviceParametersRefreshTimeout=5`

## Parameters Initial Timeout

This sets the timeout in seconds for settingsReceived() to wait for a device's initial cloud parameters to be processed before applying GUI settings on top of them. Default: 1

`#SettingsAwaitParametersTimeout=1`

## Concurrent Input Threads

This controls how many independent threads will pre/postprocess inferences. On powerful systems higher values might result in higher throughput, but on low-powered systems this might increase latency. (Max 1024) Default: 8

`#MaxNumberConcurrentInputOutputThreads=8`

## Concurrent Inferences

This controls how many inferences can be queued in the runtime. Higher values might lead to higher throughput on systems with performance headroom, but might lead to higher latency on lower powered systems. Default: 8

`#MaxNumberConcurrentInferences=8`

## Number of Samples for Rolling Average

The number of samples used to calculate the engine throughput rolling average. Default: 100

`#EngineThroughputWindowSize=100`

## Maximum Retries

The amount of times the AIM will try and get a ready signal from the external processor. After this limit has been reached, the AIM will assume the process is not running or crashed, and will skip it. Default: 5

`#ExternalProcessorReadyCheckMaxRetries=5`