How It Works

Web search can be enabled for a conversation using the Web Search icon below the input area. When enabled, AfferLab allows the conversation to retrieve information from the internet. AfferLab supports two different ways to perform web search.

Model Native Search

Some models provide built-in web search capabilities. In the model select panel, these models display a web search icon. When such a model is selected, AfferLab will use the model's native search mode directly during the conversation.

AfferLab Web Search Tool

If the selected model does not support native search, AfferLab can perform web search through its built-in search tools. This requires the model to support tool usage. In this mode, the model calls AfferLab's search tools to retrieve web results and use them in the response.

When AfferLab performs a tool-based web search, the process follows two steps.

First, AfferLab sends the search query to the search engine and retrieves the top N relevant webpages. The value of N can be configured in Settings → Web Search → Limit, with a default of 5.

After the results are returned, the model selects the webpages it considers relevant and retrieves their content for further reading before generating the final response.

When Search Is Not Available

If the selected model supports neither native search nor tool usage, web search cannot be used in the conversation.

Search Provider

AfferLab currently performs tool-based searches using a local Bing browser session. This approach is free to use but may not always guarantee the accuracy of search results.