This plugin enables the Hypothetical Document Embedding (HyDE) technique.
HyDE consists in asking the Language Model to simulate an answer (hypothetical answer) to a question. Such answer is embedded and used to recall relevant context from the vector memories.
HyDE is beneficial in Question-Answering tasks. The underlying idea is that using the embedded hypothetical answer in a vector similarity search would lead to better results rather than using the embedded question itself.
Warning When this plugin is enabled, the Cat makes and additional call to the language model API.