April 2024 Summaries

Retrieval Augmented Generation (RAG) is a technique that uses information retrieval methods to provide contextual information for generative AI. However, RAG based on vector similarity has some weaknesses, such as difficulty in answering questions involving multiple topics and limitations on the number of chunks retrieved. Knowledge Graphs can be used as an alternative or supplement to vector-based chunk retrieval. In a knowledge graph, nodes correspond to specific entities, and edges indicate relationships between the entities. This approach has several benefits over the similarity-based approach, including better handling of multiple topics and nuances from different sources. Knowledge Graphs can be created using LLMs (Large Language Models) and stored in databases like DataStax Astra DB for efficient retrieval. The use of knowledge graphs for RAG does not require graph databases or specialized query languages, making it easier to apply using a typical data store.

Winweb, an ERP software provider for the food and wholesale sectors, has integrated AI into its customer service infrastructure by developing an AI Assistant. The intelligent assistant is trained to answer questions about Winweb using information from the WinwebWiki and winweb-food help platforms. Through a collaboration with DataStax, Winweb introduced a chatbot powered by retrieval-augmented generation (RAG), enabling precise responses to complex queries by performing targeted searches across diverse information sources. The AI Assistant significantly enhances customer interaction by providing highly relevant and accurate information. Winweb uses Azure OpenAI's GPT 4 as the LLM, text-embedding-ada-002 as the embedding model, and LangChain for efficient data retrieval from DataStax Astra DB. Prioritizing data protection standards, customer data is secured using encryption protocols and hosting within the European Union. Winweb's AI Assistant has reduced the time required to find answers by 20X and aims to optimize support workflows and improve overall operational efficiency. The company plans to develop custom chatbots tailored to each customer's specific domain data, utilizing Langflow for rapid integration and experimentation.

Vercel has released v3 of their AI SDK, marking a significant step forward in the evolution of generative AI. This development provides web developers with the ability to build generative UI experiences powered by React Server Components. DataStax is excited to announce its Astra DB integration for Vercel, allowing developers to create Astra DB accounts and provision databases directly within the Vercel experience. The integration simplifies the process of getting started with building and reduces common errors associated with copying and pasting API keys. By connecting the AI SDK v3 demo to a database, users can now retrieve real-time stock price information from their Astra DB account. This demonstrates how developers can create more intelligent applications using generative UI capabilities.

Google NEXT '24 introduces an integration between DataStax Astra DB and Vertex AI Search and Conversation. This collaboration allows users to leverage their Astra DB data for building advanced search and conversational AI applications. Vertex AI Search and Conversation, a suite of tools from Google Cloud, simplifies the deployment of intelligent search and chatbot experiences with low-code or no-code approaches. The integration enables seamless connection between Astra DB data and Vertex AI Search and Conversation, unlocking new possibilities for users. Detailed instructions on configuring Astra DB as a data source are available in comprehensive documentation.

At Google Cloud Next, DataStax is demonstrating its integration with Gemini Code Assist for automatic generation of Apache Cassandra compliant CQL. This collaboration with Google Cloud aims to speed up the development of applications using Astra DB. Developers often work with numerous tables and require assistance in creating correct Cassandra queries. Gemini Code Assist not only generates accurate CQL statements but also provides context on how to use them in applications. The integration is now generally available, allowing developers to access it from the Google Cloud console.

DataStax has released an upgrade to their Astra DB Data API and clients that significantly increases compatibility with MongoDB, making it easier for developers familiar with document databases to transition to Astra DB. The update includes refinements to Python, TypeScript, and Java clients, aligning method signatures with those found in MongoDB clients, enhancing documentation, and streamlining overall functionality. Additionally, the Data API now supports ObjectId, a frequently used concept in MongoDB, allowing developers to reuse existing code without significant changes. The integration of Mongoose.JS also enables users to use their Mongoose models right in Astra DB, further simplifying the transition from MongoDB.

The DataStax Vertex AI Extension is a powerful tool designed to enhance interactions with Google Gemini and other Google Cloud-hosted LLMs, bridging the gap between your data and your LLM. It enables a more intelligent and data-driven approach to AI applications where the LLM can directly make decisions about which data sources or APIs to search for data. The extension integrates LLMs such as Google Gemini with your Astra databases, enabling direct retrieval and manipulation of data stored within Astra. This empowers your LLM to become a more active participant in the logical flow of your application as well as acting on responses. Vertex AI Extensions are currently under private preview, but their potential can be explored further on Google Cloud.

Physics Wallah (PW), an educational YouTube channel turned app, has experienced significant growth since its inception in 2016. The platform provides comprehensive courses catering to various domains and has amassed over 15 million downloads by January 2024. PW achieved unicorn valuation status and secured $100 million in funding. To enhance user experience, the company developed AI Guru, a personalized AI tutor application that uses DataStax Astra DB as its vector store, Azure OpenAI (GPT 3.5) as the LLM, and text-ada-32 as the embedding model. PW's architecture supports intelligent search capabilities with enhanced context understanding, delivering precise responses to user inquiries. The platform aims to scale its chatbot to accommodate 10 million active users and develop a recommendation engine for personalized course recommendations. Through its cost-effective approach, Physics Wallah ensures that every student has the chance to realize their limitless potential.

DataStax has acquired Langflow, a platform that simplifies the process of building AI applications. The integration of Langflow with DataStax Astra DB enables developers to create AI-powered apps quickly and easily. This acquisition aims to make all developers AI developers by providing an open source solution with a growing community of contributors and users. With Langflow's intuitive interface, developers can build on top of LangChain primitives and tap into the new AI programming stack. The platform is designed for rapid experimentation, testing, and iteration, making it easier than ever to get started with building AI applications.