Let's Build a Natural Language Movie Recommendation App using the Parallel Voting Agent Workflow in Swift
Learn how to build more accurate and reliable AI-powered apps using a Parallel Voting Agent Workflow, demonstrated through a Swift-based Movie Recommender app that combines LLMs and Google Search.
In my last blog post Let's build a PDF Query App using AI Agents Parallelization Workflow in Swift we went over the Agents Parallelization Workflow - having multiple AI agents run in parallel on smaller tasks and then synthesizing the outputs to provide the final answer to the user.
One additional use-case for the Agents Parallelization Workflow that was not mentioned is using it for voting for the best (most common) answer across LLM calls. To demonstrate how this works, image an Movie Recommender app where a user can enter any natural language query for the movie they want to watch, such as:
Romantic comedies available on Netflix or Amazon Prime
Bollywood movies with best dance moves
Documentaries about fast food
Murder mystery based on Agatha Christie
Horror movies based on real-life events
Running
Underrated indie films that deserve more attention
Movies featuring epic car chases
The app would then recommend 5 best movies that fit the topic as follows:
So how do we build this? Letβs dive in π