AI Domain Categorization
AI categorization module can be configured to categorize unknown domains with the help of OpenAI.
Categorization task runs once per hour during usual update task and selects the list of uncategorized domain names from the traffic history. Each domain is categorized using the artificial OpenAI model. Results of the categorization are saved into the folder C:\ProgramData\Diladele\WebProxy\N.N\var\spool\categories_ai and are applied in the same way as the normal domain categories from the default database.
OpenAI Integration
To access settings for OpenAI integration, click Admin UI / Proxy Settings / Site Categorization / AI Categorization as shown on the following screenshot. Adjust the settings as desired.
Important
Note, in order to be able to call OpenAI API for site categorization using AI you would need to request a unique OpenAI key as explained in this OpenAI support article. Keep this API key secret because if it is leaked, anybody can use it on your behalf. OpenAI API usage is not free.
You can also adjust the following aspects of the categorization.
| Setting | Description |
|---|---|
| Model | OpenAI model to use for categorization. Currently GPT 5.6 Terra model is recommended. |
| Confidence | A numerical value between 0 and 1 which indicates how certain an AI model is that detected categories of the given domain are correct. |
| Holding | If a domain cannot be categorized by AI, try recategorization again after given number of days. Default value is 30. |
| Previous Days | Number of previos days in the traffic statistics to load the uncategorized from. |
The Ignore Domains field allows you to specify which domains are to be ignored during AI categorization. Separate the entries with comma.
Categorized Domains
Categorized domains are stored in the C:\ProgramData\Diladele\WebProxy\N.N\var\spool\categories_ai folder and used automatically during site categorization. If a site is categorized as belonging to the category which is blocked, then access to that site will be blocked automatically after the categorization is finished.
An intermediate storage for recategorizer module is located at C:\ProgramData\Diladele\WebProxy\N.N\var\recats folder. Logs of recategorization is part of the log for the normal definition update scheduled task.
2026/10/26 14:44:05 AI Domain Recategorizer 4.8.0.821D (amd64, windows, windows) is starting...
2026/10/26 14:44:05 Command line arguments:
2026/10/26 14:44:05 Product : wfp
2026/10/26 14:44:05 Action : categorize
2026/10/26 14:44:05 ConfigFile : C:\ProgramData\Diladele\WebProxy\4.8\etc\recats.json
2026/10/26 14:44:05 Telemetry configuration:
2026/10/26 14:44:05 Enable : false
2026/10/26 14:44:05 CategoriesDir : C:\ProgramData\Diladele\WebProxy\4.8\var\spool\categories
2026/10/26 14:44:05 StatsDir : C:\ProgramData\Diladele\WebProxy\4.8\var\stats
2026/10/26 14:44:05 RecatsDir : C:\ProgramData\Diladele\WebProxy\4.8\var\recats
2026/10/26 14:44:05 Categorize Limit : 100 unknown domains
2026/10/26 14:44:05 Categorize Unknowns of Last : 3 days
2026/10/26 14:44:05 Recategorize Unknowns After : 30 days
2026/10/26 14:44:05 Confidence : 0.900000
2026/10/26 14:44:05 Ignore domains visited less than: 100 times
2026/10/26 14:44:05 Ignored Categories:
2026/10/26 14:44:05 content_delivery_networks
2026/10/26 14:44:05 generic_non_categorized
2026/10/26 14:44:05 network_infrastructure
2026/10/26 14:44:05 Ignored Domains:
2026/10/26 14:44:05 .amazonaws.com
2026/10/26 14:44:05 .azure.com
2026/10/26 14:44:05 .azurewebsites.net
2026/10/26 14:44:05 OpenAI configuration:
2026/10/26 14:44:05 ApiKey : *** masked ***
2026/10/26 14:44:05 Model : gpt-5.6-terra
2026/10/26 14:44:05 Running categorize action...
2026/10/26 14:44:05 Categorize is disabled, nothing to do.
2026/10/26 14:44:05 AI Domain Recategorizer is completed successfully, no domains were categorized, doing nothing