August 2024 Summaries

Vercel has launched its first version of the Vercel Marketplace with support for three new database options - Supabase, Redis, and EdgeDB. The marketplace aims to simplify access to various services by integrating them into a single platform, making it easier for users to manage multiple logins, setup configurations, and billing systems. Currently, the marketplace supports storage solutions on Hobby, Pro, and Enterprise plans through a self-service flow. Vercel plans to add more integrations based on user feedback and aims to create a convenient place for developers to access the best tools in the ecosystem.

OpenID Connect (OIDC) is a standard for authenticating people online, introduced by the OpenID Foundation in 2014. It simplifies the authentication process and provides a seamless and secure way to log into various services. Vercel leverages OIDC to enhance the security of backend connectivity, enabling developers to replace long-lived credentials with more secure, temporary tokens. By using short-lived tokens issued by Vercel as an Identity Provider (IdP), developers can mitigate risks associated with storing static credentials and ensure consistent role-based access control across their entire infrastructure.

The Devolver team, publishers of leading independent games, experienced significant improvements in their game website deployment process after adopting Vercel. They were able to cut time spent on system management and configuration by more than half, allowing them to launch websites 73% faster. With Vercel's zero-configuration integrations, built-in CI/CD, and robust local testing features, the team was able to achieve worry-free deployments, faster feedback, and improved reliability. By moving away from their previous clunky infrastructure, Devolver was able to focus on what they love - playing games and working on their websites - and are now even considering expanding their team to take advantage of Vercel's support.

Our team at Vercel used AI-based logic to improve their domain registry experience for international domains by fixing edge-case errors in their code, specifically when dealing with Punycode format domains containing non-English characters. The issue was caused by an optional language code not being utilized and a library like CLD2 failing to accurately determine the language due to short domain names. They turned to the Vercel AI SDK and chose a model that could generate a 3-character language code, successfully identifying the correct codes for all domains with a 100% success rate in testing, leading to improved user experience and functionality of their platform.

The text discusses best practices for building scalable AI applications, focusing on infrastructure, data management, and development using Vercel's AI-native platform. It emphasizes choosing the right AI providers, architecting for multi-modality and generative UI, integrating any model into an app with the AI SDK, handling third-party, fine-tuning, or in-house models, data cleansing and management, retrieval-augmented generation (RAG) techniques, simplified implementation with Vercel, choosing the right infrastructure, optimizing AI application performance through streaming and caching, and how to create trustworthy AI applications using Vercel's platform.

On August 7, 2024, Vercel experienced a significant service disruption affecting its Edge Middleware and Edge Functions due to an upstream provider's global configuration error. The outage lasted for 26 minutes and impacted 61% of all invocations. Vercel has taken steps to mitigate the issue by initiating a config override, restoring the desired state, and implementing fail-safes for their traffic volume. They are also working on enhancing the resilience of their Edge Middleware and Edge Functions through configuration management, additional redundancy, and unified infrastructure. The company apologizes for the inconvenience caused to its customers and is committed to preventing similar issues in the future.

Artificial intelligence (AI) is becoming a crucial driver of innovation and competitive advantage in businesses. With the AI market projected to grow by 37.3% annually through 2030, organizations need to effectively integrate AI into their operations. This guide provides a strategic framework for identifying and developing AI features that align with business goals and technical capabilities. It emphasizes the importance of frontend-centric approaches in creating exceptional customer value and using AI to reimagine user interactions. The Vercel Impact Assessment for AI Initiatives (VIA-AI) is introduced as a tool to prioritize AI projects based on their potential impact, feasibility, strategic alignment, business logic enhancement, data quality and availability, performance and scalability, security, compliance, and user trust. Key performance indicators (KPIs) are also discussed for measuring both technical performance and business outcomes of AI initiatives. The guide highlights the role of various teams in AI project success and how Vercel's platform can democratize AI development by allowing engineers at any experience level to contribute meaningfully to AI projects.

Vercel AI SDK 3.3 introduces four major features: tracing with OpenTelemetry, multi-modal file attachments, useObject hook for streaming structured object generation, and additional LLM settings such as JSON schema support, stop sequences, and custom headers. These enhancements provide improved observability, flexibility, and control when building AI applications using JavaScript and TypeScript.

The text discusses various methods to protect an app against malicious traffic using Firewall rules, spend limits, code-level optimizations, and more. It highlights the importance of real-time usage monitoring and controlling, as well as proactive measures like setting maximum function durations, ensuring consistent image source values, and moving large media files to blob storage. The text also emphasizes deployment protection settings and automatic blocking of suspicious TCP connections by Vercel's Firewall.

Beyond Menu, a popular food delivery service in the US, adopted feature flags for gradual rollouts, instant rollbacks, A/B testing, trunk-based development and easier collaboration with beta users. They needed to evaluate feature flags and A/B tests on both server and client sides due to their use of Next.js and Vercel's App Router. The solution they chose was Hypertune in conjunction with Vercel's Edge Config, which provided seamless feature flag management without layout shifts. This setup enabled ultra-low latency reads at the edge, minimizing performance impact on the app and its conversion rate. Beyond Menu was able to roll out new features safely and reliably, with a robust developer and user experience, and with no negative effects on their app's performance or end-user experience. They plan to use Edge Config with Hypertune for more complex app configuration management and product analytics in the future.