RAG Ollama specific token context length guidance missing

[jasonculligan]

In the RAG documentation it states:

warning
If you're using Ollama, note that it defaults to a 2048-token context length. This means that retrieved data may not be used at all because it doesn’t fit within the available context window. To improve Retrieval-Augmented Generation (RAG) performance, you should increase the context length to 8192+ tokens in your Ollama model settings.

But it doesn't do anything to explain how to do this. Where should one go to find the proper guidance?