anchorOverview

The WCCAI APIs expose contact center AI capabilities, that can be integrated with third party applications and data streams, to enable real time AI driven actions. The APIs are built and hosted in the cloud and can be integrated with other cloud deployed or premise deployed applications. They are built on a scalable real time platform, that integrates with multiple Cisco-native and 3rd party AI services and common interface to harness these services.

anchorArchitecture & Integration

The WCCAI Platform provides two APIs:

1. Speech Insights Orchestrator API: for consuming voice streams with insight options
2. and Serving API: for retrieving speech insights

The Speech Insights Orchestrator API takes in real time, voice streams via a gRPC connection along with various types of insight requests . Insight requests specify the type of insights required (Transcription, Agent Assist etc.) and the Service Provider to use (Cisco, Google etc.). Additional parameters can also be passed as appropriate. This API is built on gRPC protocol. The Speech Insights Orchestrator API will route the voice stream to the chosen STT provider. It fetches real time streaming transcripts and pushes them to the AI Data Platform. In addition, if more insights like Agent Assist and Sentiment are required, it will also send the transcripts to those respective services, fetch insights and push them again to the AI Data Platform. Each leg in the call (IVR, Caller, Agent) would be provided as separate streams, linked by a conversation ID (Call ID/ Call GUID etc.).

The Serving API provides a simple interface to subscribe to insights in real time. It is also built on gRPC protocol and supports gRPC-Web. Each request will subscribe to a given conversation ID and a list of insights needed. Conversation ID is typically the CallGUID that is used to identify the call in the ACD and all incoming streams to the Orchestrator API carry this CallGUID. Once subscribed, the insights will be published to the subscriber as they are derived in real time. A given conversation ID can have multiple subscribers, so multiple clients like Agent Desktop, Supervisor Desktop and CRM connector can subscribe at the same time. Every time a new connection is created for a conversation ID, the history of the conversation ID can also be obtained. This helps to deal with broken connections.

anchorOnboarding

The Speech Insights Orchestrator APIs support multiple Insight Service Types. For each Service Type, multiple providers are also supported. New Service Types and Providers are being added or considered for backlog. The following is the current list.

Insight Types & Providers supported

The Speech Insights Orchestrator APIs support multiple Insight Service Types. For each Service Type, multiple providers are also supported. New Service Types and Providers are being added or considered for backlog. The following is the current list.

Entitlements & Scopes

Both Speech Insights Orchestrator API and Serving API are integrated with Cisco Identity to provide authentication and authorization. Clients would need to create an Org in Control Hub, then create users under this Org. Using the User credentials, Access tokens need to be generated and used as part of the request. Please refer to Control Hub/Cisco Identity documentation on the process/APIs available to generate tokens. The following table shows the entitlements and scopes that need to be configured for users, to use the various insight types.

Voice formats supported

The following voice formats are supported for Cisco (Voicea) Encoding: LINEAR16 and MULAW Sample Rate: 8000 Hz and 16000 Hz

Languages Supported

Cisco (Voicea) Transcripts today support English. French, Spanish and German support is being added.