Using LLMs to find movies, TV, documentaries (etc)

alt text

Movie, TV and documentary preferences

If you're looking for something good to watch on Netflix tonight (or elsewhere) try the following prompt-and-context template.

This is (another) great use-case for the context-snippet idea which you can use as a low-tech surrogate for RAG (or try what I'm attempting to set up: have a context repository that serves as a pipeline to a RAG setup!).

The idea is as follows:

You jot down your movie preferences once and save that as a template file like daniels-movie-preferences.md.
Then, you reuse that tiny text snippet to prompt to get targeted movie recommendations, even if the LLM is totally context-"naive"

A model context snippet is below. Because where we are based in the world affects the choice of entertainment available to us, I've included a little note of where the user is located.

Prompt model

You could prompt something like this:

Any movies I might enjoy from this year?

Here are my preferences

(insert context snippet)

How you add the context snippet

If you're using a web UI youc an simply drag and drop the file.

Manual method: copy and paste.

Example context snippet recording your general entertainment preferences

Interests:

- Space Exploration  
- Nature and Wildlife  
- Crime Thrillers  
- Sports and Adventure  
- Political Dramas  

## What I don't like

- Reality TV shows  
- Romantic comedies  
- Movies that are overly long  

Topics, types, genres:

- Space exploration stories  
- Fictional accounts of historical events  
- Documentaries about nature  
- Political intrigue  

## Limitations

I live in Canada. You can recommend content that isn't in Netflix's Canada library, but if you do, flag it.