This is considered the template or prompt format for communicating with the particular model that you're using. You can find more information and the specific templates on https://ollama.com/library/mixtral:latest. Since mistral is based on llama, it uses the [INST]prompt[/INST] template. If using gemma then the template would look something like <start_of_turn>model {{ .Response }}<end_of_turn>. It's basically formatting to yield better results from the model that is being used.
I didn't realize that langchainrb had included ollama. That's really cool.
Personally, I prefer to wrap my own classes around something like ollama since it does provide an API. And a lot of knowledge from those kinds of gems could be taken. I am planning an episode of vector searching with meilisearch, but am waiting for it to be released/stable so that could take a bit.
My main thing against something like langchainrb is if it is going to be a core part of my application. If it is, then I probably wouldn't want to rely on a gem to provide core functionality. If it is a nice-to-have feature, then I would add it to my list of considerations when making feature.
