Pytia
Speech to text (ASR) integration
Client of ASR speech recognition services. It has a set of appropriate connectors for specific external services to which the audio stream should be transmitted and text recognition should be received in return.
Configuration
voice.gateway {
pytia {
technology = TECHMO
api {
base-url = "0.0.0.0:8081"
api-key: "xxx"
}
service-tags: ["techmo-pl"]
grpc-endpoint-port = 6443
service-availability {
ping-interval = 10d
failure-limit = 2
failure-time-window = 5m
enabled = true
}
}
techmo.engines = [
{
host = asr3-istio.example.com
use-plain-text: false
port = 8443
name = "techmo-1"
}
]
}
Field description
- technology - name of a specific supplier - selection of the right connector for the service. Can be one of the following: GOOGLE, TECHMO, PHONEXIA. When choosing technology make sure that the technology subconfig is also present.
- service-tags - Configures tags that are propagated in akka service discovery, may be used to choose a certain ASR service for communication for a certain dialog (e.g. for tests or to choose a specific engine). This list will be always supplemented by technology tag i.e.
pytia {
technology = "GOOGLE"
service-tags = ["some-tag"]
}
Will result in pytia introducing itself in cluster with tags: ["some-tag", "GOOGLE"]
- api - REST API parameters of the website
- grpc-enpoint-port - direct access port to the speech recognition service (used in the Automate chat tester)
- service-availability - configuration of checking the availability of the ASR service
- techmo.engines - configuration of access to the ASR service with a name for identification and communication method (use-plain-text option)
Phonexia - example config file
voice.gateway {
pytia {
technology = "PHONEXIA"
service-tags: ["Phonexia-pl"]
api.base-url = "http://0.0.0.0:8080"
service-availability {
ping-interval = 10d
failure-limit = 2
failure-time-window = 5m
}
}
Phonexia dictate subconfiguration.
phonexia.engines = [{
# Login to access Phonexia's API.
login = "my-login"
# Password to access Phonexia's API.
password = "secret-password"
name = "phonexia-pl"
# URL address of Phonexia's REST API.
address = "http://1.2.3.4:8600"
# Name of speech recognition model. Should be supplied by Phonexia.
model = "PL_PL_6"
# Interval for getting transcription and asking Phonexia API.
polling-interval = 1s
}]
}
Updated 7 months ago