November 2023 Summaries

DataStax Astra DB is a real-time vector database that can scale to billions of vectors, while Amazon Bedrock is a managed service offering foundation models via a single API. Together, they facilitate the creation of accurate, secure, and compliant generative AI applications. This post demonstrates how to create a simple question and answer system using Astra DB vector database, embeddings, and foundation models from Amazon Bedrock. The example uses Amazon Titan Embeddings and Anthropic Claude 2 model for the LLM. By combining Astra DB with Amazon Bedrock, developers can build applications that use retrieval augmented generation (RAG), FLARE, and model fine-tuning.

The Cassandra Summit, along with the AI.dev Open Source GenAI & ML Summit, will take place in San Jose, California on December 12th and 13th. This event brings together experts from the Cassandra community to discuss topics such as data partitions, building large clusters, and the latest developments in NoSQL databases. The summit will feature presentations by Hugging Face, LlamaIndex, Nvidia, AWS, and DataStax. Attendees can expect insights on how Cassandra supports popular applications like Netflix and Bloomberg Terminal, as well as discussions on performance tuning and e-commerce use cases.

Restworld is a unique platform that functions as a hybrid job board and recruitment agency within the restaurant sector. It aims to streamline the matchmaking process between employers and potential employees by leveraging cutting-edge technologies, including AI tools like automatic CV parsing and matching algorithms. The platform emphasizes personalized and efficient experiences for users in the hospitality industry. Restworld has recently delved into generative AI, utilizing embeddings and vector search for semantic interpretation of data, which facilitates knowledge representation and efficient data traversal. Collaborating with DataStax to implement GenAI and vector search capabilities in Astra DB has been pivotal for the company's growth. Restworld is not just revolutionizing recruitment; it's reshaping the future of hiring in the hospitality sector by offering a competitive advantage through personalized experiences for job seekers.

The recent ousting of OpenAI CEO Sam Altman has caused shock and uncertainty within the generative AI market, as OpenAI holds a significant 80% market share for large language models (LLMs). Despite this shakeup, DataStax has established partnerships with various other GenAI players such as AWS Bedrock, Meta's Llama 2, Cohere, Hugging Face, and GCP's PaLM 2 and Vertex.ai. Additionally, customers rely on a range of open-source solutions, clouds, and APIs to maintain resilience against rapid platform shifts. DataStax will be hosting webinars discussing app stack patterns for customers in production across LLMs, data, agents, and orchestrators, as well as challenges faced when bringing an LLM to production.

OpenAI recently announced new features enabling developers and enterprises to use GPT4 with their own data. DataStax Astra DB is now integrated with SuperAgent, an open-source framework that helps build AI assistants for production. This integration allows users to leverage real-time data feeds and LangChain's ecosystem. Retrieval augmented generation (RAG) is a common design pattern for generative AI applications, which can be applied to scale using Superagent and Astra DB with vector. An example use case involves automating the combination of structured and unstructured SEC data to better understand companies and industries.

DataStax has announced a strategic collaboration agreement with Amazon Web Services (AWS) aimed at accelerating generative AI innovation into production. The partnership includes seamless integrations with Amazon Bedrock, Amazon SageMaker, AWS Glue, and more, unlocking significant cost-effectiveness without compromising performance through the deployment of DataStax Astra DB on AWS Graviton. Security and compliance are ensured by Astra DB's AWS PrivateLink ServiceReady Designation and PCI, SOCII, and ISO 27001 certifications. The collaboration also focuses on supporting customers at every step of their journey with technical and business support from account teams working closely with AWS experts. Joint marketing initiatives include free trials, webinars, workshops, and events such as DataStax's Gold Sponsorship at AWS re:Invent 2023.

The RAGBot template, created using Next.js and hosted on Vercel, offers a customizable chat app implementing retrieval augmented generation (RAG) with the help of the Vercel AI SDK, new OpenAI Assistant API, and DataStax Astra DB vector search. Users can switch large language models (LLMs) and toggle RAG on and off to test their model's performance. Deployed via Vercel's Frontend Cloud, this template is part of the company's Template Marketplace, offering pre-built, customizable project starters for Next.js web applications. The RAGBot chatbot starter uses React and Next.js for its frontend and TypeScript with the Vercel AI SDK for its backend. It also demonstrates how to use the Vercel AI SDK to vectorize sample datasets and integrate them into an Astra DB vector database.

Today, we released the Astra Assistants API, a new service designed for users and organizations that want additional control over their data. The system uses DataStax Astra DB to store messages, assistants, threads, runs, and files, and its vector functionality for retrieval augmented generation (RAG). It calls out to OpenAI for large language model-related tasks like embedding generation and chat completion. About 70% of the OpenAI API's endpoints are currently supported by the Astra Assistants API, with plans to add more in future releases.

Today, DataStax released the Astra Assistants API, which is compatible with OpenAI's new Assistants API. This service allows users to easily switch their AI apps to use this new API by changing just a single line of code. The Astra DB Vector database supports retrieval-augmented generation (RAG) patterns for building powerful interactive assistants that combine large language models with custom data. The Astra Assistants API is now available as a developer preview, offering a simple and scalable datastore backed by the performance and reliability of Cassandra. To try it out, create an Astra account and follow the tutorial notebook provided.

Application developers are working on creating generative AI applications that can understand and enhance their business-specific contexts. Vector search is the primary solution for searching and retrieving this semantic context, but many vendors face challenges such as hallucinations, performance issues, lack of real-time context, and complexities in development and integration with GenAI ecosystems. To address these issues, DataStax has introduced a public preview of its new Astra DB vector experience, which simplifies the GenAI application development journey. The new UX allows users to upload both vector and non-vector datasets, perform cosine similarity analysis, and add or remove metadata filters for quick analysis and relevance tuning. Additionally, developers can enjoy the power of Apache Cassandra without learning Cassandra or CQL (Cassandra Query Language) through native clients for Python, JavaScript/TypeScript, and Java. The new Astra DB vector experience is compatible with large language model (LLM) orchestrators like LangChain and LlamaIndex, as well as hyperscaler services such as Google Cloud Vertex AI. DataStax customers across industries are using Astra DB to build generative AI applications, leveraging the power of simplicity, scale, and the GenAI ecosystem.

Skypoint, an AI solutions provider for the senior care industry, has successfully implemented generative AI applications using retrieval augmented generation (RAG) techniques. The company's AI application allows healthcare providers instant access to public and private data, improving patient care by streamlining administrative processes and providing conversational AI. Skypoint uses a combination of structured and unstructured data stored in its lakehouse and DataStax Astra DB with vector search. Advanced RAG techniques are employed to improve answer quality and reduce hallucinations. The application also integrates few-shot prompting for generating SQL statements from natural language queries, improving performance by using the vector store as an LLM cache. Skypoint's approach involves focusing on real-world use cases, progressive implementation, user education and training, and planning for future capabilities of generative AI.

DataStax Astra DB has won three categories in the TrustRadius 2023 Best of Awards for Vector Databases, reflecting its exceptional value, feature set, and customer relationships. Built on Apache Cassandra, Astra DB is a vector database designed to help developers quickly build accurate generative AI applications using real-time data. The platform's features include elegant APIs, powerful data pipelines, and complete ecosystem integrations. DataStax earned these awards based solely on customer feedback, which praised the feature set, alignment with sales and marketing promises, and exceptional value for the price.

DataStax Astra DB has won three categories in the TrustRadius 2023 Best of Awards for Vector Databases. The achievement highlights Astra DB's commitment to delivering top-notch features, affordability, and strong customer relationships. Built on Apache Cassandra, Astra DB is a vector database designed for developers working on accurate generative AI applications. It offers elegant APIs, powerful data pipelines, and complete ecosystem integrations. The awards were based entirely on customer feedback, praising the feature set, alignment with sales and marketing promises, and exceptional value for the price.

At the GitHub Universe global developer conference, a new collaboration between DataStax Astra DB and GitHub Copilot was announced. The DataStax Astra DB GitHub Copilot plugin will enable developers to build with Astra DB more efficiently by conversing with their database via GitHub Copilot Chat. This integration aims to enhance productivity, improve code quality, streamline collaboration, and facilitate rapid prototyping for developers using Astra DB. The plugin is set to provide contextual knowledge and memory from Astra DB, allowing developers to focus on more complex tasks and creative aspects of software development.

DataStax's research reveals that AI innovation is rapidly advancing, offering significant productivity gains and unique customer experiences. Many executives, managers, and developers are eager to utilize AI tools in their organizations. A survey of over 500 IT leaders and practitioners was conducted to understand how enterprises rely on AI for productivity, the challenges they face, and the trusted technologies driving innovation. The full report is available for download.

Hornet, an international social network and dating app with over 35 million members, is leveraging artificial intelligence (AI) to create a safe space for the global queer community. The company uses AI to enhance user safety by identifying hate speech, moderating images, and detecting malicious actors. In partnership with DataStax, Hornet utilizes vector databases to provide personalized recommendations and improve real-time matching capabilities. This collaboration showcases how AI and advanced databases can revolutionize user experiences in innovative ways for developers looking to enhance their platforms.

The recent integration of image understanding capabilities into large language models (LLMs) like ChatGPT has opened up new avenues for multimodal text and image models. By incorporating retrieval augmented generation (RAG), these models can be steered towards producing more accurate and relevant results by providing them with the most recent and accurate context from data, including images. This approach is particularly useful in mitigating hallucinations often generated by powerful LLMs and LMMs. The multimodal vector store created using CLIP and Astra DB can be queried to provide contextual understanding for multimodal models like MiniGPT-4, improving their accuracy and relevance. As multimodal models become more accessible, the potential applications of these technologies continue to expand, offering exciting possibilities for the future of AI.