Echobot: Migrating a Teams Calling Bot from VMSS to a lower-cost, modern architecture?

Hi everyone,

Some time ago, I implemented a Microsoft Teams bot that joins meetings and captures real-time raw audio streams from the participants' microphones. I used the architecture from the official microsoft-graph-comms-samples (EchoBot approach) as my reference.

The system has been running in production for a while and works perfectly, but the infrastructure is becoming incredibly expensive and hard to maintain. As you know, this approach relies heavily on Azure Virtual Machine Scale Sets (VMSS) and full Windows Server environments running 24/7 just to handle the native media platform binaries.

I’m looking to modernize this stack and drastically cut down our Azure bill. I wanted to ask if anyone has successfully migrated away from VMSS for this specific production use case:

Windows Containers (AKS/ACI): Has anyone successfully moved an Application-Hosted Media Bot into Windows-based Docker containers using Azure Kubernetes Service (AKS) or Azure Container Instances (ACI)? Does it actually help reduce costs and scale better than VMSS?

Serverless / Service-Hosted: Is there any trick to get continuous raw audio streaming using Service-Hosted Media (serverless), or are we still strictly limited to just playing prompts and recording files?

Alternative architectures: If you have optimized a real-time Teams media streaming bot for production, what architecture or hosting strategy did you use to keep costs under control?

Could you please point me in the right direction regarding which specific tools or services I should use for this migration? Any links to official documentation, articles, or reference architectures would be highly appreciated.

Thanks!

Start by reviewing the EchoBot approach and the existing VMSS-based architecture described in the issue, then compare it with Azure AKS/ACI Windows container guidance and Service-Hosted Media capabilities. Done would require a documented, production-suitable migration recommendation that addresses continuous raw audio, scaling, maintenance, and cost.