November 2023 Summaries

The Advent of Tensors 2023 invites Vespa enthusiasts to participate in a festive 24-day challenge, exploring the capabilities of Vespa Tensors, a framework for real-time tensor computations on evolving datasets. Each day features a new coding challenge that highlights various AI-powered use cases, such as gift recommendation systems and facial recognition, using Vespa Tensors' advanced functionalities. Participants can solve these challenges in an online playground with zero dependencies, and solutions are shared daily. This interactive event encourages the community to engage with complex tensor operations and provides an opportunity to win Vespa-themed prizes, fostering a collaborative and festive atmosphere for both experienced users and beginners.

This blog post presents a hands-on tutorial on using Vespa's streaming mode for efficient retrieval of personal data, in conjunction with LLamaIndex, to create advanced generative AI pipelines. It details the configuration of Vespa with PyVespa, including the use of Vespa's native embedders and various ranking methods such as hybrid retrieval and Vespa Rank Fusion. It also covers the integration of LLamaIndex retrievers to build Retrieval Augmented Generation (RAG) applications that can federate and blend query results from multiple data sources like personal email and calendar data. The tutorial emphasizes Vespa's cost-effective approach by using disk-based storage and avoiding in-memory data retention, significantly reducing deployment costs. Additionally, it highlights the potential for expanding the application to include more data sources for comprehensive personal context tracking, showcasing Vespa's versatility in managing and querying large-scale personal datasets.

Vespa has announced a transition in its operating system support, moving from CentOS Stream 8 to AlmaLinux 8 for its open-source platform, due to the impending end-of-life of CentOS Stream 8 on May 31, 2024. This decision aligns with the internal migration of Vespa.ai to AlmaLinux 8 and Red Hat Enterprise Linux 8, as AlmaLinux is used in Vespa Cloud production systems and is available free of charge. While RPMs for Vespa will continue to be built on Copr and distributed through a new Cloudsmith-hosted repository, the change is expected to have minimal impact on users, as the new AlmaLinux 8 base for Vespa container images maintains compatibility with RHEL ABI standards. This transition allows Vespa to maintain an archive of releases and ensure forward compatibility for users still on CentOS Stream 8, with future support plans remaining adaptable to the evolving Linux distribution landscape.

Vespa Cloud has introduced anonymized endpoints and access token authentication to enhance security and streamline deployment processes. Anonymized endpoints, now the default for new applications, conceal application details and expedite certificate issuance, while still supporting the old format for existing applications with backward compatibility. Additionally, Vespa Cloud now supports token-based authentication for data plane access, particularly beneficial for environments where mutual TLS (mTLS) is not feasible, such as Vercel edge runtimes. While mTLS remains the preferred method, token authentication provides an alternative, requiring careful handling of token expiration and storage to maintain security. These updates aim to improve both the security and usability of Vespa Cloud for developers.

Yahoo Mail, a leading global email provider, is enhancing its service with AI-driven features that allow users to interact with their inbox through questions and commands. This innovation is powered by Retrieval Augmented Generation (RAG), which combines a large language model for text understanding and generation with a retrieval system using semantic search and vector embeddings to find relevant inbox information. Vespa.ai's technology plays a crucial role in this development by providing a cost-effective RAG system capable of handling personal data, supporting vector embeddings, structured data, and full text within queries and ranking functions. This collaboration aims to offer these advanced features to all users for free, significantly simplifying and accelerating email interactions. Vespa.ai's robust vector database technology is proven to efficiently store and search trillions of documents, making it an ideal choice for Yahoo Mail's ambitious project.

Vespa.ai, a recent spin-out from Yahoo, has secured a $31 million investment from Blossom Capital to enhance its platform for big data and AI solutions. This Series A funding will enable Vespa to expand its capabilities and serve both large and small customers via its cloud solution. The partnership with Blossom Capital is considered ideal due to the firm's expertise in building global tech companies and its quick decision-making process. This investment marks a significant step for Vespa in establishing itself as an independent entity focused on advancing its technology ecosystem.