April 2024 Summaries

In the second part of their series on chat moderation in 100ms, the author delves into more ways to handle text chats in video apps using cloud services and LLMs. They begin by setting up 100ms Prebuilt for this tutorial as it comes with Chat configured. The author then discusses how various cloud services offer options for moderating text, such as Google Cloud's Natural Language API and Microsoft Azure's Content Moderator. By leveraging these cloud-based moderation tools, developers can implement robust chat moderation systems without having to build everything from scratch. They also explore the use of Large Language Models (LLMs) for chat moderation, which can sometimes be excessive due to associated costs but offer an intriguing use case in adjusting text to generate a message that is safe to share. The author concludes by discussing how implementing OpenAI's API for this task is relatively simple and provides insights into constructing your chat moderation system.

This article discusses different ways to automatically manage text chats in video apps using 100ms. It covers methods such as using regular expressions to omit personal information being shared, checking for phone numbers with an external package, and utilizing TensorFlow.js's 'Text toxicity detection' model to detect toxic content. The article also mentions that these are just ideas and not perfect solutions.

This guide explains the technical aspects of creating a notification-based one-on-one call application called Blitz. The app is powered by 100ms and Firebase, enabling users to make video or audio calls with features like Call List UI, Self-Preview, Call Options, and Call Notifications. Key technologies used include WebRTC for real-time communication, Firebase Cloud Messaging (FCM) for push notifications, and Apple's CallKit for native call interface on iOS devices. The guide also provides step-by-step instructions to set up the Blitz app using 100ms Prebuilt UI components and customizing it with a tailored calling interface from the hms_room_kit package.

The blog discusses the challenges developers face when optimizing video call quality using WebRTC. It highlights that testing in real-world conditions is crucial but often overlooked. To address this, 100ms offers dogfooding sessions where their customer success team tests customers' apps before rollout. This involves understanding end-users' setups and simulating stressful network conditions. The blog presents a case study of a dating app that faced issues with video resolution dropping frequently on slightly older Android/iOS phones. Through dogfooding, 100ms identified the issue as CPU limitations in these devices and adjusted the publish layer settings accordingly to improve video quality. The article emphasizes the importance of understanding end-users' demographics to tailor an optimal experience and reduce testing load for developers.