Skip to main content
Each SLNG region has its own gateway and base URL, and requests are not routed between regions. Services differ by region: the regions map lists the speech models, Context Router, and agent regions in each one. For every region, its base URL, and its services, see Region availability.

​
Why regions matter

​
Latency

Voice is real time and turn-taking has a hard budget. A round trip to a distant region adds hundreds of milliseconds to every turn, which is the difference between a natural conversation and a bad user experience. Running in the caller’s region is the single largest latency lever available, ahead of any model or pipeline optimization.

​
Data residency

Voice is considered PII (personally identifiable information) under several compliance frameworks and regions, and a growing number of jurisdictions require it to stay inside their borders. For compliance details, see the SLNG Trust Center.

​
The whole stack, not just part of it

Most platforms run models or media transport close to your users but execute the agent somewhere else. That means the component holding the actual conversation, and the context it accumulates, sits outside the region. With an SLNG Managed agent region, the gateway, the models, the Context Router, and the agent workers run in the same region, so a region selection covers the entire request path.