What is PartyRock?

It has been two weeks since Amazon announced PartyRock, an Amazon Bedrock Playground. It comes with the tagline “Everyone can build AI apps”. According to Amazon President and CEO, Andy Jassy, it was just an internal tool created by AWS developers to experiment with Foundation Models from Amazon Bedrock. The name PartyRock was in reference to it being a fun and collaborative way to experience Amazon Bedrock.

I joined the party and played around with PartyRock to see what all the fuss was about, and it literally took me less than five minutes to create my first app! PartyRock is essentially a code-free application of GenerativeAI. You start building an app by describing what you would like your app to do. Here, Prompt Engineering comes into play as you use appropriate words, sequenced effectively to be able to get your desired output. It utilizes Amazon Bedrock on multiple stages of building. 

With the utility of Amazon Bedrock, you can select from different models (Claude, Claude Instant, Jurassic-2 Mid, Jurassic-2 Ultra, or Classic)  to create the widgets to your application. Widgets can either be AI Powered Text Generation, Image Generation, or Chatbot that change values based on an input (User Input or Static Tex). 

The secret sauce to PartyRock is its simplicity. It is an easy way to learn about GenerativeAI. A user puts an input, and an AI-powered widget can generate results from a predefined prompt via a reference (using the @ sign) to the input widget. This widget can then be used by another widget to further provide more details about the generated result.

If you are stuck on the conceptualization of the app you want to initially build, you can also explore apps built by other PartyRock users and use the remix option to add your own flair to the app. Just recently, AWS added the Discover tab on PartyRock where you can explore based on different categories (App caviar, Mix it up, Oh snapped, and different Staff picks). 

5 Interesting Apps Built with PartyRock 

I was exploring what fun ideas other PartyRock users came up with and listed a few of my favorites:

PantryToRecipe

A recipe suggestion app, PantryToRecipe will take your pantry ingredients and suggest a recipe you can make. As someone who doesn’t have a reliable chef around, I do most meal preparations and cooking myself. Toward the end of the week, I find myself left with a few ingredients in my refrigerator that are not really bought together to create a dish. With this app, you can list those ingredients down, add a food genre (type of food), and other requirements such as time limitations and resource constraints on your cookware, and the app will build a recipe you can make. 

Resume Optimizer

No, I am not actively updating my resume! But ResumeOptimizer is an interesting app that I see myself recommending to those who want to utilize GenerativeAI to improve how their resume is written. It takes the job description into consideration and takes an active voice to writing. It also provides a sanity check to what is listed and identifies weaknesses or relevance to the role.

Python Assistant for Gaming Conversations

This app is really interesting to me as it was able to create Python code based on how I described the rules of a game. What made it more amazing is that it was able to identify that I was describing a tic-tac-toe game based only on the rules.

Discover Unique Insights and Stories of your Destination

Who needs a tour guide when you have GenerativeAI? This app helps provide insight into your tour destination that you have no idea of. It gives a 200-word list of five facts about your destination categorized as weird, unique, random, unexpected, and unbelievable (as specific as it can get). If that is not enough, there is also a chatbot that can provide more details on the topic.

Family Gift Finder

How many times did you catch yourself out of ideas when thinking of a gift for your family? This app makes it easy as you describe what the family member is fond of, and it can generate different gift suggestions for you. Quite handy in the holiday season!

Security Incident Response Scenario Generator

I lied. I can’t limit my favorite list to just five. This app is a security incident scenario simulator that can aid in training resources and providing incident response plans. Now that security is becoming a more popular topic, especially with the accelerated growth of AI, apps like this will also become more relevant.

Using PartyRock like a Real Developer

For a limited time, AWS is offering PartyRock users a free trial. If you have been exposed to GenerativeAI, an important concept to remember is being specific in the way you word your prompts to be able to generate better outputs. As usage is limited to a certain amount of credits, it is important to add restrictions when writing prompts in order to limit the output tokens to a few characters long and save on credits. Backstage provides an overall view of your apps as well as the overall consumption of credits in PartyRock.

Advanced Settings can also be added to the widgets to influence the response generated by the models. Temperature (creativity parameter) and Top P (or nucleus sampling) both change the response of the model in terms of randomness or unpredictability. Adding these parameters in the widget and changing the default values can significantly change the way the models select the response.

Review of PartyRock

PartyRock is built around Amazon Bedrock, which in itself is a great technology that makes interaction with Large Language Models accessible with an API. The main benefit of this is the accessibility of AI technology to generate text like a human and carry conversations in a creative way, in the same language as the user (inclusive).

What makes PartyRock awesome is its intuitiveness. It boasts that any user, whether a seasoned developer or someone with a no-tech background, can create AI apps in minutes. It is not overwhelming to learn and is a fun way to start your journey of exploring GenerativeAI.

References:

https://partyrock.aws/guide/faq

https://aws.amazon.com/bedrock/faqs/