Thanks — I’m responding to your question about how much RAM the LiveKit Agent server has. We do not have a single fixed RAM value for an agent server. Recommended and observed figures are: • Recommended starting point: 8 GB RAM per agent server (suitable for many voice-AI workloads). • Example usage: a load test with 30 agents on a 4‑core, 8 GB machine used about 2.8 GB RAM. • Small plugin example: the turn detector uses < 500 MB when run locally. • Ephemeral storage: ~*10 GB* is typically sufficient for the Docker image and temp files. If you tell me your expected concurrent jobs, features used (noise cancellation, heavy ML models, etc.), or whether this is self‑hosted or cloud, I can recommend a sizing and scaling approach. Sources: Self-hosted deployments | LiveKit Documentation | LiveKit turn detector plugin | LiveKit Documentation, with interactive elements