sqlite-utils-llm 0.3

Run llm from sqlite

sqlite-utils-llm

This is a plugin for sqlite-utils that adds a custom SQL function llm(model, prompt) that can be used to call out to an LLM model using the llm command line tool.

Installation

Have sqlite-utils and llm already installed, then install this package with pip:

sqlite-utils install sqlite-utils-llm

Example

Given the following combos.csv file with ice cream flavor combinations we want to complete the suggestion column with a flavor suggestion based on the other two flavors using an LLM.

flavor1,flavor2,suggestion
vanilla,chocolate,
strawberry,vanilla,
chocolate,strawberry,
mint,lemon,

Let's load the data into a SQLite database:

% sqlite-utils insert icecream.db combos combos.csv --csv --empty-null

Now we can use the llm function to update the suggestion column with a flavor suggestion:

% sqlite-utils query icecream.db "UPDATE combos set suggestion=llm('gpt-4o-mini', 'What other flavor would I like if I have choosen ' || flavor1 || ' and ' || flavor2 || '? Return just the name of the flavor.')"

Now we can query the table to see the suggestions:

% sqlite-utils query icecream.db "SELECT * FROM combos" --table
flavor1     flavor2     suggestion
----------  ----------  ------------
vanilla     chocolate   Strawberry
strawberry  vanilla     Chocolate
chocolate   strawberry  Vanilla
mint        lemon       Basil