Why Nushell is so BRILLIANT at working with AI

Fundamentally LLMs are all about relating text to text.

An LLM can complete, enrich or predict text by relating input text from the user to all the text relations already in the models interior.

The person working with an LLM will want to benefit from its power and vast “knowledge” of language.

Nushell is the best way I have found to do this in a way that is both systematic and, once you have learned the basics of Nushell, easy and straight forward.

Let us have a look at an example. We have this table in Nushell:

We want to create a small fiction about each of these persons.

In Nushell then simply do:

$table | insert fiction { $in | aichat provide very short fiction based on input }

As you may have noticed I also use the terrific CLI tool AIChat

Back to our example, we now have this:

In the above example we asked Nushell to relate each whole row to the LLM with the $in variable.

If we instead want to relate just one cell from each row, this can easily be done like so:

$table | insert description { $in.hobbies | aichat provide very short one paragraph description of the hobby }

Notice the $in.hobbies - only feeding each hobby cell in the table to the LLM this time around.

Let us “save” the result in a new table, though:

let hobby = $table | insert description { $in.hobbies | aichat provide very short one paragraph description of the hobby }

Giving us this:

In case we are only interested in the hobby descriptions, Nushell makes it very easy to select only that column like so:

$hobby | select description

… and so on.

I need to 🏊 now. But remember, I’ve only just scratched the surface here.

The possibilities for relating text, specific parts of texts and text-based structured data to the immense power of LLMs with nushell are almost endless.

You can read more about Nushell here: Nushell · Kiils

Also make sure to visit the Nushell homepage here: Nushell