February 2024 Summaries

MLB's technology transformation aimed to modernize their stack and build a dynamic API capable of fetching data dynamically from various sources. The resulting REST-based Stats API was successful but led to complexity and payload size issues. To address these challenges, MLB recognized the need for a more scalable and performant API architecture, which led them to adopt GraphQL. By exposing all features of their StatsAPI via a new GraphQL subgraph, they were able to streamline data orchestration and provide uniform representation of game-related queries across multiple consumer-facing products. The adoption of GraphQL enabled MLB to deliver personalized experiences without overburdening client-side logic, and its centralized approach prevented inconsistent data contracts and reduced complexities.

The GraphOS Schema Proposals feature is a centralized change management workflow that streamlines the process of proposing, reviewing, and approving schema changes to a supergraph. This allows developers and platform engineers across multiple teams to collaborate more effectively on designing and implementing federated GraphQL APIs. The feature includes integrated checks to validate proposed schema changes, automated publishing of approved changes, and safeguards to prevent unapproved changes from being published. With GraphOS Schema Proposals, organizations can improve developer velocity, reduce the risk of breaking client apps, and ensure that changes do not conflict with other backend teams' subgraph schemas.

The emergence of AI-driven tools has significantly transformed the tech landscape, with code generation tools like Copilot, CodeLlama, and others gaining attention from developers. Despite their advanced capabilities, a notable gap exists in their ability to generate high-quality GraphQL schemas. The author shares their experience fine-tuning GPT-3.5 for optimized GraphQL schema generation, which showed improvements but not flawless results. To address this, they created a Discord server where users can input schema descriptions and receive generated schemas based on best practices. The more specific the input, the better the resulting schema will be, allowing users to iterate and improve their schema further.

GraphQL provides several advantages for backend developers, including improved communication between API producers and consumers through a shared schema definition language (SDL), reduced bikeshedding over unnecessary discussions about minor details, and less tech debt by allowing for easier deprecation of outdated code without breaking changes. This results in clearer documentation, faster debugging, and a smoother upgrade path for clients, ultimately making building and maintaining APIs more delightful.

Federated subscriptions with HTTP callback support are now available in Apollo Router, reducing real-time cloud spend and improving efficiency at scale. This feature solves the backend WebSocket problem, allowing for more flexibility and interoperability between the router and underlying services. With HTTP callbacks, the router no longer manages a WebSocket connection with subgraphs, making it suitable for serverless infrastructures. To use federated subscriptions with self-hosted instances of Apollo Router, users need an active Apollo GraphOS Enterprise plan, a federated GraphQL API with HTTP callback-compatible servers, and the latest version of Apollo Client with HTTP-based subscriptions support. The feature is now available in general availability (GA) release of Apollo Router 1.37.