Worked on the coze-dev/coze-js repository to enhance the Audio Player by implementing resource management improvements and configurable speech recognition. Leveraging TypeScript and front end development skills, the developer introduced mechanisms to release audio player resources and support custom speech recognition configurations, addressing diverse deployment requirements. The work included a stability fix that ensures timeouts are cleared and WebSocket connections are properly closed before destroying the player, reducing the risk of resource leaks and playback crashes. These changes improved the reliability and flexibility of speech-enabled features, with a focus on robust resource lifecycle handling and clean teardown in edge cases.