SpeechKit Hybrid architecture
SpeechKit Hybrid comprises the following components:
- Envoy
: Routing server that receives incoming gRPC requests and determines the SpeechKit Hybrid components to handle them. - License server: Envoy dynamic configuration server for accounting Yubikey licenses or providing data required to pay for resources.
- STT server: Speech recognition server.
- TTS server: Speech synthesis server.
Relationship between SpeechKit Hybrid components
Component relationships depend on the licensing model selected by a partner.
Yubikey
The License server limits the maximum number of speech recognition sessions and maximum performance (requests per second) for speech synthesis. These parameters are read from a Yubikey hardware authentication device
Cloud Billing
No limit on the number of sessions and requests per second. Information about each service request is sent to Yandex Billing. Total service usage fees are specified in a payment invoice issued to a partner in their billing account. The component relationships are shown on the diagram below: