OpenAI Agents SDK Upgrade: Adds TypeScript Support and Voice Interruption Feature

OpenAI recently announced a significant upgrade to its Agents SDK, introducing several new features that promise to enhance developer capabilities and user experiences. The most notable updates include expanded language support and advanced voice interaction technologies. The upgraded Agents SDK now supports TypeScript, providing developers with the same functionalities as the Python version. This expansion allows for fine-grained control and manual approval options, enabling more precise management of AI agent behavior. The SDK is designed to streamline multi-agent workflows by offering configurable language models, agent switching, built-in safety measures, and a comprehensive suite of analytics tools. A key feature of this update is the introduction of the RealtimeAgent functionality. Developers can now create voice agents that operate both on the client and server sides. These agents support real-time voice interruption and can dynamically call various tools during conversations. To enhance monitoring and debugging, the "Traces" dashboard has been significantly updated. It now displays real-time data for voice agents, including audio stream status, tool usage statistics, and interruption frequency, providing detailed performance insights. Underlying the new voice interaction capabilities is an enhanced version of the language model. The updated "GPT-4O-Realtime-Preview-2025-06-03" improves accuracy in executing commands, reliability in tool calls, and the gracefulness of handling interruptions. This model is available to developers through both the Realtime API and the Chat API, ensuring seamless integration with existing systems. In addition to these functional enhancements, OpenAI has outlined strategic changes to its API offerings. By mid-2026, the company plans to phase out the current Assistants API, replacing it with the Responses API. Developers will be encouraged to transition to the new API to leverage its advanced features and scalability. However, the Chat Completions API will remain available for those who do not require tool integration. The compatibility of the Agents SDK with OpenAI's APIs and other third-party models that adhere to the Chat Completions standard further enhances its flexibility. This ensures that developers can choose the best tools and models for their specific needs, whether they are working within OpenAI’s ecosystem or integrating with external providers. Overall, this major upgrade from OpenAI represents a significant step forward in the development of AI-driven voice interactions, offering developers powerful new tools and a clear path for future API strategies. The enhanced functionalities and improved language models are expected to drive innovation and improve user experiences across a wide range of applications.