Hi All, i have face some issues , i will generate the room unlimited for each join but the agent will not connect with the all of them it should showing too quite error which is its default error after refresh 6 to 7 time then the agent is connect. and when the agent say something take 2 min approximately. how i fix these two errors can anyone help me??

Thanks for reaching out, Adi. I'm an AI assistant and I'm here to help. Give me a moment while I research this topic.

Dear LiveKit Support Team, We have developed a voice-based conversational agent integrated with real-time video proctoring using LiveKit. The solution is now production-ready. However, during our testing phase with multiple concurrent users, we have encountered significant issues that are impacting system reliability and user experience. Specifically: Room Connection Delays: When multiple users attempt to join sessions simultaneously, the room connection time becomes excessively delayed, affecting the onboarding and interaction flow. Agent Communication Freezes: During live conversations, if a user speaks and the agent momentarily goes silent, the agent fails to resume the dialogue—even when the user explicitly prompts with phrases like "Go ahead" or unmutes their microphone. This results in the agent becoming completely unresponsive, halting the conversation unexpectedly. These issues are severely affecting real-time interaction, which is critical to the effectiveness of our platform. We kindly request your assistance in investigating and resolving these problems at the earliest, as they are blocking wider deployment. Please let us know if additional logs or diagnostics are needed from our end.

Hey there! I totally get how frustrating connection delays and communication hiccups can be. Let me share some friendly tips to help smooth things out: • Parallelize Connection Steps: Think of this like having multiple checkout lanes open - generate access tokens in advance and connect users and agents at the same time, which can seriously speed things up! • Implement Worker Pool Parallelization: If you're running your own setup, spread out the workload across different CPU cores. It's like having a team working together instead of one person doing everything. • Enable Simulcast and Adaptive Streaming: These are your performance superheroes that help manage bandwidth as more people join the party. • Use Distributed Architecture: Imagine spreading your servers out like a smart network - this helps keep things running smoothly and prevents any single server from getting overwhelmed. Now, for those tricky agent communication freezes - we've got solutions: • Increase AudioSource Buffer Size: Bump up your

queue_size_ms

to 1000-2000ms in

rtc.AudioSource

. It's like giving your audio a little extra breathing room to prevent glitches. • Adjust VAD Sensitivity: Fine-tune those Voice Activity Detection settings. Think of it like adjusting a sensitive microphone so it doesn't miss important moments. • Implement Inactivity Handlers: Add some smart monitoring to catch and recover from those awkward silent moments. It's like having a helpful friend who nudges the conversation back on track. • Check Network Proximity: Make sure your agent is hanging out close to its key services. Proximity matters! If you can share more details about your setup, I'd be super happy to dive deeper and help you troubleshoot even more effectively! Sources: Going beyond a single-core | Web and mobile frontends | LiveKit Docs | Bringing Zoom's end-to-end optimizations to WebRTC | An introduction to WebRTC Simulcast | delay-handling/raw_data.txt