The idea is to be able to plot the mapping between a query, a document, and the proximity of the response given to the actual "correct" set of responses. For extra step we could even point at a specific document as the "correct" one for the given input query. 

Should setup a vector DB, ideally using pg_vector.

The Poc should consist on getting a vector db up and running, populate it with data, and query the model so that the responses can be plotted as the described mapping.

Review Ollama and Tangerine as part of the process, use Nomic as the embedding model (https://ollama.com/library/nomic-embed-text)