No-code generative AI enhanced search
There has been a lot of talk recently about combining the natural language fluency of LLMs with various approaches to grounding, that is, generating responses that are precise and relevant to a particular use case. Vertex AI Search makes it possible to use natural language to ask questions and have the answers grounded in a data source that you specify. The data source can be a set of websites, a table in Big Query, or data in Cloud Storage. The data in Cloud Storage can be structured (a JSON file) or unstructured (PDF, HTML, or text files). Best of all, Vertex AI Search lets you combine the natural language flexibility of LLMs with grounding without having to do any coding. In this article, I’ll show how easy it is to set up Vertex AI Search to ask natural language queries about a PDF file.
Selecting a PDF file as a data source
Earlier this year, I wrote an article that included a description of how to use Flowise to ask questions about a PDF file. For that exercise, I used a PDF version of my book , Deep Learning with Structured Data, as a data source. I chose my book as the grounding document because I wanted to be able to ask very specific questions that could only be answered correctly using the book as a source.
