Auto GPT is a highly intelligent AI assistant, utilizing advanced language models like GPT-4. It provides wide-ranging capabilities such as performing Google searches, processing external documents, and more. In this tutorial, we will learn how to build such an AI assistant using Relevance AI, a no-code platform.
Relevance AI is a platform enabling users to create powerful Auto GPT applications without coding. With unique features like the ability to upload PDFs, send HTTP requests, and a visually engaging building experience, it offers an easy entry into the world of AI.
Before we start, it's crucial to note that the following example is just one of the many use cases for Auto GPT and Relevance AI. We're using it as a practical demonstration of how to utilize the platform. Once you understand the process, you'll be equipped to adapt it for a multitude of other tasks and scenarios.
After signing into Relevance AI, navigate to the main page. Here, you'll find several ready-to-use examples. Select the option to "Build AI chain" and add a new chain for your project.
For this example, we want to create an AI assistant to help write a personalized email for angel investing outreach. We might want the AI to research Mark Cuban's background on LinkedIn and Google his recent news. This will provide context and inform our email content.
Relevance AI's interface comprises two sections. The right side is where you construct a web app interface for users to input data, and the left side is where you develop the AI chain.
Start by adding a 'Make API Request' block to perform a Google search on Mark Cuban and retrieve the latest news. We'll use an API service called SERP, which offers high-quality Google search results.
To tailor SERP, copy the necessary parameters into your project's body. These parameters will help focus the search results. For example, fill the query field with "Mark Cuban" and replace the header field with your API key from SERP. Run the notebook to see if you get the desired results.
If your results aren't quite right, refine your search. Add a 'topic' field to filter for news about Mark Cuban relevant to AI. Adjust this parameter and run the search again to check the improved results.
Next, extract the news list for GPT to select the most pertinent article. Insert a JS code transformation block to run JS code that returns the news while removing irrelevant information.
With your news list, pass it to GPT, giving it a prompt to choose the best news for understanding Mark Cuban's perspective about AI. Remember to specify that you only need the news link, not the title or source.
Lastly, to scrape data from the chosen website, add another 'Make API Request' and utilize a service like Browserless to extract the required information.
By following these steps, you can build your own Auto GPT personal assistant with Relevance AI. Remember that this is just one example. With your newfound knowledge, you'll be prepared to adapt the process for a wide variety of use