Products

Here is PartyRock, the AWS platform for learning Ai

We tested the generative AI app creation tool, based on Amazon Bedrock, to create apps in minutes.

by Marco Trabucchi

3 July 2024

3' min read

Key points

How it works
AWS PartyRock, how to get started

3' min read

It is called Partyrock and has been designed by Amazon Web Service with the aim of making learning generative AI accessible to all by experimenting with the use of prompts, i.e. the instructions you give to artificial intelligence to get relevant answers to the question. The aim is to create an open and stimulating playground to test oneself with generative AI, learning the techniques needed to take full advantage of various LLMs, concatenating different prompts to get more targeted answers.

How it works

PartyRock provides developers with access to foundation models (FM) from leading AI companies including AI21 Labs, Anthropic, Cohere, Meta, Stability AI and Amazon via Amazon Bedrock in an 'intuitive, code-free app creation playground'. In practice, you can quickly and easily create mini-apps ¬for a wide variety of needs and where the limit is only creativity, delivering results in seconds and allowing you to focus on creation instead of the complexity of development. For example, it is possible to recommend what to cook based on the ingredients in the larder, analyse and optimise the budget of a party or create a storyteller to organise a fantasy role-playing campaign. All this while experimenting with different advanced language models (Claude 3, Llama or Titan, for instance) and sharing your creations online if you wish.

AWS PartyRock, how to get started

Access to PartyRock is simple thanks to an instant access user interface, which is separate from the AWS Management Console and does not require a specific account. You can experiment with prompts of your own creation or get started quickly using predefined templates.

We tried to make a demo and, given the approaching summer holidays, the choice fell on an app capable of giving travel advice. The app's interface is in English, but fortunately the instructions can be entered in Italian, as the various LLM (Large Language Models) are multilingual.

After creating a free account through Google, the first thing we did was to click on the button at the top of the left-hand column 'Generate App'. This opens the "What do you want to build?" window, where we write down what we want the app to do.

If ideas are in short supply, below Partyrock suggests some apps, dividing them by functionality: productivity, fun stuff and education. These are preset apps such as the 'social media post generator', to create content on social media, or the interview preparation app, and again, the seasonal recipe generator.

After writing the instruction to generate the app, the first screen opens up, which Partyrock has called 'Globetrotter's Guidance'. Already pre-set are two widgets: the 'desired destination' and the 'interests and preferences', the latter of which are crucial for guiding Ai's suggestions on the desired holiday. Underneath is the 'Travel Tips' text box, the prompt where the Ai will make its suggestions.

By default PartyRock has included 'destination image', an image generator and 'chat with travel expert'. Clicking on edit reveals that PartyRock uses Stable Diffusion XL and Titan for image generation.

To refine the search, we want to add more targeted instructions. We do this by clicking on the top-right button "add widget". This generates a window at the bottom where we have to specify that we want to generate a new "user imput".

Clicking on 'user imput' opens a new window on the right where you will have to titrate the widget, in this case we call it 'Trip Cost'. We click on the Save button and the generated widget is placed at the bottom by default. We drag it to the top for convenience.

The last step will be to concatenate all our inputs by refining the text generation prompt. On the "travel tips" window highlighted in green, we click on "edit" and add the instructions using the @ so that the Ai takes into account all the widgets created. In this case we add the cost of the trip. We do the same by editing the chat with the expert.