Google Dialogflow (GDF) Plugin 1.15.0 to the UniMRCP Server (UMS) has been released.
The plugin is based on the following components:
- UniMRCP Server 1.7.0
- Google Dialogflow API v2
- gRPC 1.20.0
- Protobuf 3.7.0
The binaries are currently available for the following Linux distributions:
- Red Hat / CentOS 7 (unimrcp-gdf-1.7.1-1.el7.x86_64.rpm)
- Ubuntu 16.04 LTS (unimrcp-gdf_1.7.1-xenial_amd64.deb)
- Ubuntu 18.04 LTS (unimrcp-gdf_1.7.1-bionic_amd64.deb)
This release introduces support for multiple Google account credentials and numerous new features and parameters. The release is built against a newer version of Google APIs.
The detailed list of changes introduced in this release follows.
- Added support for multiple Google account credentials, which can be specified either per server profile or per MRCP session via feature tags.
- Added support for gRPC meta data 'x-goog-user-project'.
- Added support for an alternate way of triggering a Dialogflow welcome intent by means of 'builtin:grammar/event_welcome' in addition to 'builtin:event/welcome'. The new method is commonly supported by the majority of IVR platforms.
- Allow word-level time offset information be retrieved based on a new configuration parameter 'word-info'. The parameter can be set globally in umsgdf.xml and be specified per recognition request either via vendor-specific parameters or optional attributes passed to a built-in grammar or via metadata set in an SRGS XML grammar.
- Allow domain-specific model and model variant be specified. The parameters can be set globally in umsgdf.xml and be specified per recognition request either via vendor-specific parameters or optional attributes passed to a built-in grammar or via metadata set in an SRGS XML grammar.
- Added support for speech adaptation boost. The 'weight' attribute of the 'item' element in SRGS XML and/or a new 'boost' attribute of the 'phrase' element in 'speech-context' can be used for this purpose.
- Added support for Dialogflow agent versions and environments.
- Added support for inter-result timeout. If the specified timeout is elapsed, input is considered complete. The timeout defaults to 0 (unset) and can be overridden per recognition request.
- While composing an NLSML result based on a data structure received from Dialogflow, replace '-' with '_' in the XML parameter names to fix interoperability with Genesys PureConnect. This behavior is controlled by a new configuration attribute 'replace-dashes', which is enabled by default.
- Added a new attribute 'word-info' to the element 'streaming-recognition'. The attribute defaults to 'false'.
- Added new attributes 'model' and 'model-variant' to the element 'streaming-recognition'. The attributes are not set by default.
- Added a new attribute 'environment' to the element 'streaming-recognition'. The attribute is not set by default.
- Added a new attribute 'inter-result-timeout' to the element 'streaming-recognition'. The attribute defaults to 0 (unset).
- Added a new attribute 'replace-dashes' to the element 'results'. The attribute defaults to 'true'.
- Upgraded Google APIs (unigoogleapis-1.1.0).
- Using 'single_utterance' in 'InputAudioConfig' instead of the deprecated property of 'StreamingDetectIntentRequest".
- Using 'phrase_hints' in a speech context passed to 'InputAudioConfig' instead of the deprecated proprety of 'InputAudioConfig'
- Class tokens can now be specified as a text of 'item' elements in an SRGS XML grammar or 'phrase' elements in 'speech-context'.
- Updated the Usage Guide to reflect the changes introduced in this release.
Visit the GDF plugin page for more information.
Thank you for using UniMRCP.
Author of UniMRCP