The system should support two types of users.

For a theater we want to record an id and a maximum occupancy (available seats). Furthermore, only some theatres are 3D theatres and some theaters have a fancy sound system.

The schedule for the cinema records for which date and time which movie is shown in which theater. The cinema has regular opening hours and movies can only be shown during the opening hours. Obviously, at each time only a single movie can be shown in a particular theater.

For each movie screening (timeslot at a particular theater where the movie is screening), we record ticket sales. For each screening we cannot sell more tickets than the maximum occupancy of the theater in which the screening takes place. The price of a ticket is determined as follows:

• There is a base price for tickets of $15
• If the screening is in a 3D theater then the price is increased by $5
• If the screening is in a theater with a fancy sound system than the price is increased by $3
• If the movie is by a major studio then the price is increased by $3

• If the movie is old then the price is decreased:

  • if the release date more than 2 months in the past, then the final price is reduced by 20%
  • if the release date more than 2 years in the past, then the final price is reduced by 40%

A single ticket sale can be for multiple tickets for the same screening. Tickets can either be brought by registered clients (in which case we need to record which registered client brought the ticket for reward purposes) or by an anonymous client (in this case we have no information about who brought the ticket). When a ticket is bought by a registered client then we need to record which of the client's payment method was used.

For movies we would like to record several pieces of information:

• Title
• Major studio: was the movie produced by a major studio or not
• Release date
• Length
• Original language

• Available languages

  • one or more languages
• Directors

• Actors

  • We need to store which actor played which character (name) in the movie. An actor may play multiple characters in one movie.
• Writers
• Producers

The system should allow users (clients and admins) to register using their email address, setting a password for login. Registered users can log into the system using their password. Based on the type of user, they have different functionality available to them.

The cinema wants to recommend movies to its clients using RAG (retrieval-augmented generation). The plan is to embed movie information (title, descriptions, …) using an embedding model. Clients registered with the cinema can submit a description of their interests to the cinema which will also be embedded. The embeddings should be stored in Postgres using the pgvector extension which allows you to store vectors (resulting from employing an embedding model) as values in postgres and to build indices (HSNW) for approximate nearest neighbor (ANN) search. To produce a recommendation for a client, you need to embed the clients description of interests and then find the 3 most similar movies in the embedding space as a recommendation.

The movie cinema has to schedule movies to theaters. An optimal schedule would assign movies to theaters such that maximum occupancy is achieved (ideally all theaters are as full as possible) without running out of capacity for theater rooms (not all clients who want to see a movie are able to get a seat). Of course, optimal planning would require knowing the future which is not possible. Instead the theater wants to use a machine learning regression model to predict future seat requirements for movies and then model optimal scheduling as an optimization problem.