Vertex AI Search: Leverage the Power of Google Search and Gemini for Your Information Needs
One of the significant challenges organizations face is the discoverability of information. Data often resides in disparate sources — ranging from documents and databases to web pages — making it difficult for employees and customers to find what they need, whether it’s internal data or public information.
Vertex AI Search, part of Vertex AI Agent Builder, addresses this challenge. Vertex AI Search is a fully managed Google Cloud service that lets you build and deploy customized search engines across your own data, enhanced by the Gemini Generative AI model for natural language responses. You can index and query websites, structured data, documents, databases, Google Drive, Cloud Storage, and third-party sources with minimal setup. This enables developers, even those with limited AI expertise, to leverage cutting-edge innovations in Generative AI, Semantic Search, Embeddings, and Retrieval Augmented Generation (RAG) to create robust search experiences.
Use Cases
Vertex AI Search can be applied across various scenarios, including:
- E-commerce and Product Search: Facilitate detailed product searches, such as electronics, fashion, or books, with advanced filters and comparison features.
- Company Intranet: Enhance internal document, database, and communication searches to help employees find relevant information more efficiently.
- Industry-Specific Searches: Create specialized search engines, like one for medical research that indexes only reputable journals and websites, ensuring access to peer-reviewed content.
- Customer Support and Knowledge Bases: Assist users in finding relevant articles, FAQs, and support documents within a company’s knowledge base.
Getting Started: Out of the Box Experience
The core experience with Vertex AI Search starts with its Out of the Box setup. By following the Quickstart guide, you can build a basic search engine for various data types using the Google Cloud Console. You can then embed a production-ready, prebuilt widget on your website by simply copying and pasting a few lines of HTML code. This widget not only delivers search results but also provides natural language answers generated by Gemini based on those results.
Here’s an example of what the widget looks like for a search engine indexed on the Google Cloud Vertex AI Search documentation:
cloud.google.com/generative-ai-app-builder/docs/*To see it in action, you can visit this demo page. The source code for various widgets is available in the GoogleCloudPlatform/generative-ai GitHub repository.
If you want to get something up and running quickly, the prebuilt widget is a solid approach to get started. You can find a website containing widgets for a variety of search engines in this open source sample application. https://vertex-ai-search.web.app/. The source code is available in the GoogleCloudPlatform/generative-ai GitHub Repository.
For those looking to develop custom UIs, Vertex AI Search offers a REST API and client libraries in seven programming languages. You can refer to the code samples for more detailed guidance.
Advanced Features: Grounding and Custom RAG
Grounding is the capability of generative AI to link outputs to verifiable information sources. In Vertex AI, Gemini leverages this feature with data from Google Search and Vertex AI Search. You can use Vertex AI Studio or the Vertex AI SDK to generate accurate, contextually grounded responses based on the indexed data.
For a deep dive into grounding with Gemini on Vertex AI, check out this video of a workshop from Google I/O 2024 and the accompanying Jupyter notebook, Getting Started with Grounding with Gemini in Vertex AI.
For developers interested in building customized Retrieval Augmented Generation (RAG) systems, Vertex AI Search integrates with LangChain and LlamaIndex, allowing it to function as a vector store. There are also standalone APIs available to simplify many complex aspects of building a RAG system while maintaining high levels of customization. Explore these possibilities through the following resources:
- Question Answering Over Documents with Vertex AI Search and LangChain
- LlamaIndex — Vertex AI Search Retriever
- Build Your Own Grounded RAG Application Using Vertex AI Search Standalone APIs and LangChain
Learn More
For comprehensive details about Vertex AI Search and Generative AI on Google Cloud, refer to the official documentation:
Additionally, visit the Google Cloud Generative AI GitHub repository for more sample code, Jupyter notebooks, and demo applications showcasing the use of Vertex AI Search and other generative AI tools on Google Cloud.
By harnessing the capabilities of Vertex AI Search, organizations can significantly enhance the accessibility and accuracy of information retrieval, making it easier for users to find the correct answer, right when they need it.
