On-Premise ML Client Setup
Revision and Version History
Date |
Comments |
09-24-2020 |
Initial release to the public |
09-25-2020 |
Added additional logging statements |
Setup and Installation
NOTE: The user installing this service MUST be an adminstrator on the machine where it is being installed
The On-Premise client is created as a Windows service. Only machine administrators have the access rights to modify Windows services.
Steps to install
-
Navigate to the 'On-Premise' download page at https://ml.convergenceresearch.com/OnPrem/DownloadClient
Fig 1 - ML Client Download
-
Enter your API Key and select 'Verify Key|Download'
-
When the download is complete, locate the downloaded zip file in your browser's 'Downloads' folder.
-
Open the Windows File Explorer (Windows Key + E) and browse to the drive you wish to install the service
Fig 2 - Choose Install Drive
-
Create a new folder on the selected drive titled 'MLOnPremise'
-
Go back to downloaded zip file from Step #3 and right-click it select 'Properties'
-
NOTE: If you have an option to 'Unblock' the zip file (see fig 3), check the box titled 'Unblock' and click 'Apply' followed by 'OK'
Fig 3 - Unblock downloaded zip file
-
Extract the *.zip file to the folder created in Step #5. You should have the following files:
Fig 4 - Extracted files in the target folder
-
Right-click the file titled Install.bat and select 'Run as Administrator'. Select 'Yes' when prompted to continue.
A black command window should appear.
-
Follow the prompts in the command window. If you are asked to input your API key, do so but most installations should already have it embedded in the installer.
Be sure to choose a valid local Binding Address (Host Address). Valid addresses will take the format http://<hostname or ip address>:<port>
NOTE: Both HOST and PORT are required!
For example, if you are installing the service on a machine with the machine name mlserver to port 34500, the binding address would look like this:
http://mlserver:34500.
You can also bind to an ip address. EX: http://192.168.1.120:34500
-
Once you have completed the installer, the browser should open up the page at the host address entered in Step #10
Fig 5 - On-Premise service management page
When the service is done syncing with the Convergence ML API, it will be available and ready to classify. This should take no more than a few minutes.
Fig 6 - Service is ready to classify
-
Use the test form to ensure the service is working. NOTE: The first classification will take 3-4 seconds to complete but
subsequent classifications should average around 9 milliseconds each.
About the ML Client
The Convergence On-Premise ML client is installed as a Windows Service titled 'Convergence OnPrem ML'. Once installed, it can be managed from the Windows 'Services.mmc'
We recommend setting the service start up to 'Automatic' and using the 'Local System' account.
Convergence OnPrem ML Service Properties
If you start and stop the service, this will force a sync with the Convergence ML models. This is equivalent to opening the host address and clicking 'Resync Model'.
Uninstalling the Service
To uninstall the service, you need to run the 'uninstall.bat' script as an administrator. The script can be found in the same folder as the 'install.bat' script
and should be a part of the original downloaded zip file.
Steps to uninstall
-
Go to folder you created (see Setup and Installation - Step #5)
-
Right-click the .bat file titled 'Uninstall.bat' and choose 'Run as Administrator'
- Follow the prompt to remove the service.
NOTE: This will not delete the logs or remove the source from the file system. If you wish to remove these files associated with the service, you can do so through the file system.
Support and Assistance
Should you require help with your installation, please contact it@convergenceresearch.com.
If you have a log file available, please include it as an attachment when contacting support.
For questions regarding the format of requests/responses, you can use the documentation resources available at
https://ml.convergenceresearch.com/Docs/classifierdocs.
In addition, you can also use the test form's request and response to see examples in both JSON and XML.