Technical Resources
Technical report
Qwen-Audio-Agent Technical Report (PDF) covers the foreground–background voice Agent architecture, ongoing conversation alongside asynchronous work, and implementations and evaluations for desktop, smart cockpit, and customer service.
Presentation
Voice Agent Architecture for Continuous Conversation and Long-Running Tasks (Marp, in Chinese) walks through the two-tier architecture, task lifecycle, delegation, result delivery, and the memory system.
Preview with the Marp for VS Code extension, or export with the CLI:
bash
npx @marp-team/marp-cli docs/voice-agent-architecture-presentation.zh.md -o slides.html