Thanks — I’m answering your question about whether the self‑hosted LiveKit deployment scales well. Short answer • Yes — self‑hosted LiveKit can scale, but operating WebRTC infrastructure at large scale is challenging and requires active infrastructure management. What that means in practice • Horizontal scaling is supported: you can run multiple LiveKit worker processes and distribute sessions based on CPU, memory, and network load. • You should configure autoscaling (for example in Kubernetes) against CPU/utilization with conservative thresholds so the cluster scales up before the system hits the load threshold and starts rejecting sessions. • A default self‑hosted deployment uses a single‑home architecture (one server or cluster region), which limits how far you can scale without building a more complex distributed multi‑region setup. • Achieving very large scale (tens of thousands of participants or a multi‑region global footprint) requires substantial operational effort: custom infrastructure design, networking between edge regions, monitoring, and careful capacity planning. If you want less operational overhead • LiveKit Cloud provides automatic scaling and session routing across multiple homes/edges so you don’t need to manage the low‑level scaling and networking yourself. Quick questions to help me advise better • What scale are you targeting (concurrent participants per room and total concurrent connections)? • Do you need multi‑region/global coverage or is a single region acceptable? • Will many rooms be small (1–10 people) or do you expect large rooms (hundreds)? • Are you planning to run on Kubernetes or on VMs/bare metal? Tell me those details and I’ll recommend an approach (self‑host configuration, autoscaling guidance, or whether LiveKit Cloud might better fit your needs). Sources: Pricing | LiveKit | Announcing LiveKit Cloud | LiveKit | Voice Agent Architecture: STT, LLM, and TTS Pipelines Explained | LiveKit, with interactive elements