OpenAI Upgrades Agents SDK: Adds TypeScript Support and Voice Interruption Capability
OpenAI recently announced a comprehensive upgrade to its Agents SDK, introducing several significant enhancements for developers. The new version now supports TypeScript, aligning its capabilities with those of the existing Python version. This expansion allows developers to manage AI agent behavior more precisely, thanks to fine-grained control mechanisms and manual approval options. The SDK has been designed to simplify workflows involving multiple agents, offering configurable language models, the ability to switch agents, built-in safety controls, and a complete suite of analytics tools. One of the standout features of this update is the introduction of RealtimeAgent. Developers can now create voice agents that operate seamlessly on both client and server sides. These agents support the ability to interrupt and interject during conversations, allowing for more natural and interactive speech. Additionally, they can call various tools in real time, enhancing their functionality and flexibility. To complement this, the "Traces" dashboard has been significantly enhanced. It now displays real-time data from voice agents, including audio stream status, tool usage, and statistics on interruptions, providing developers with detailed performance insights and better monitoring capabilities. The underlying language models have also seen improvements. The newly updated "GPT-4T-Realtime-Preview-2025-06-03" version offers better accuracy in executing commands, improved reliability in tool calls, and a more graceful handling of interruptions. This model is now available to developers through the Realtime and Chat APIs. In terms of API strategy, OpenAI plans to phase out the current Assistants API by mid-2026. This decision comes as the company aims to achieve functional parity with the new Responses API. The Chat Completions API will continue to serve developers who do not require tool integration, but OpenAI recommends transitioning to the Responses API for access to advanced features and greater scalability. The Agents SDK remains fully compatible with OpenAI's API and other third-party models that adhere to the Chat Completions standard. This compatibility gives developers a wide range of choices and the flexibility to integrate with different platforms and models seamlessly. Overall, these upgrades position OpenAI to better meet the evolving needs of developers, offering them more robust tools to build and manage AI-driven applications, especially those involving real-time voice interactions. The changes reflect OpenAI's commitment to refining its offerings and maintaining a leading edge in AI technology.