July 2016 Summaries

Stream's APIs offer developers the ability to build, scale, and personalize feeds and activity streams, servicing over 50 million end users across 300 production applications. The newly released Based UI Kit for Feeds, available for free download and featured on Product Hunt, provides a versatile and neutral design framework compatible with Material Design and iOS best practices, making it suitable for web, iOS, and Android applications. This UI Kit includes 89 module types, 51 style assets, and 20 reusable components, and is designed to be easily modified for specific use cases. Stream's team often collaborates with developers to ensure their API can support various design prototypes, and they encourage feedback for future improvements. The UI Kit, designed with Sketch, aims to help product teams efficiently build and launch compelling feed experiences.

The blog post provides a comprehensive guide on setting up a LEMP stack with PHP 7 and Laravel 5.2 on a DigitalOcean droplet, integrating Stream's scalable newsfeed technology into a Laravel application. It walks through creating a server environment, installing necessary software like Composer and Git, and setting up Laravel with MySQL. The post then delves into integrating Stream-Laravel for building social functionalities such as creating tasks as activities, implementing follow/unfollow features, and constructing a news feed using FeedManager. It provides detailed instructions on configuring database credentials, setting up user models and repositories, developing controllers for managing user interactions, and designing Blade HTML templates for frontend display. The tutorial aims to equip developers with the knowledge to build a social network-like application while promising further exploration of Stream-Laravel's capabilities in future posts.

Sportlobster, founded in 2013, is a prominent online sports community platform that provides personalized news, live scores, and prediction challenges for sports fans worldwide. Initially relying on an in-house solution, the platform faced scalability issues, leading to unpredictable query times and limitations on advanced features. To address these challenges, Sportlobster transitioned to using Stream, a third-party solution that allowed for safe scaling to billions of feed updates. This move resulted in faster feed loads, enhanced user engagement with features like ranked feeds, and reduced maintenance costs. The integration was smooth, supported by Stream's solid documentation and responsive engineering team, allowing Sportlobster to focus on further enriching the user experience with personalized and relevant content. Andy Raines, a Backend Engineer at Sportlobster, highlights the importance of early planning for data migration and leveraging Stream's capabilities to optimize the platform's feeds and user engagement.

Bandsintown, a popular concert discovery app founded in 2004, serves over 25 million users and connects more than 230,000 artists with a global audience of 120 million music fans. In 2015, the company sought to enhance its app by adding social features and a notification feed to improve artist-fan interaction and integrate social aspects among users. Rather than developing an in-house solution, Bandsintown chose Stream to implement these features due to its ease of integration, scalability, and support for various frameworks. Stream's API enabled Bandsintown to handle large fluctuations in traffic efficiently, with feeds loading in an average of 20 milliseconds, while also offering benefits like solid documentation, 99.99% uptime, and detailed usage stats. The decision to collaborate with Stream allowed Bandsintown to build a sophisticated solution quickly and cost-effectively, avoiding the prolonged development process of an in-house build.

Content discovery and recommendation are critical challenges in machine learning applications across various industries, such as social networks, news aggregators, and e-commerce, where platforms aim to personalize user experiences by suggesting relevant content. The post explores techniques for generating recommendations using the Vowpal Wabbit library, highlighting its matrix factorization capabilities, which offer an alternative to traditional collaborative filtering methods. Vowpal Wabbit can be employed to build scalable recommenders by processing data in its specific format and utilizing its command-line tools for tasks like item recommendations and purchase predictions. The post provides a practical walkthrough of setting up Vowpal Wabbit, preparing datasets, and executing commands to fit models, with an emphasis on using its features for item recommendations and purchase predictions. Through examples, it demonstrates the process of fitting models and adjusting parameters, such as importance weights, to handle unbalanced datasets. The author encourages further exploration of Vowpal Wabbit's capabilities and invites users to experiment with their datasets to optimize performance.

The Cabin tutorial series by getstream.io offers insights into best practices for building scalable React and Redux applications, emphasizing the advantages of using immutable data structures to prevent unintended state mutations. The series also highlights the importance of designing an efficient Redux state tree, either by normalizing or denormalizing data, and introduces tools like Immutable.js and Chrome's React Developer Tools for improved debugging. It advocates for a modular organization of code using the Ducks pattern and discusses enhancing error reporting with tools like Sentry's Raven. The post also touches on optimizing production builds with Webpack, ensuring performance and error management align with React's standards, and provides additional resources for developers seeking to deepen their understanding of React and Redux.