Showing results for 
Search instead for 
Did you mean: 

How to implement Cisco Call Studio/CVP to AWS Lex Bot communication

Hi All,


I am planning to implement AWS Lex bot integration with Cisco Call Studio (CVP).

     * Need your suggestion on how to do this?

     * Is there any connector needed like gRCP(Google Dialog flow) or Through RestAPI can this be achievable ?

     * If some one already did this suggest me steps.

Current UCCE version 11.5



Cisco Employee

There's two different approaches you could take.  One is to do transcription, bot intent processing and response TTS output as three separate operations using whatever mix of speech services you prefer.  The other is to use a single speech service for voice input and voice bot processing with media or text output.

If you opt for an MRCP based approach then I believe you can get out-the-box Amazon integration using UniMRCP.   

If you want to try something out in the lab and roll your own, then you could use the gateway media forking approach I published.   I haven't pushed out an Amazon version yet as it's not quite finished but you could use it for transcription with Google or Microsoft and then use a backend HTTP API for the bot part (text in / text out).

Hi Paul,


Thank you, I have few queries.

  • In UCCE environment UniMRCP can be used for both Agent Assist functionality(like google Agent Assist on Agent Desktop) and conversational IVR(Bot self service integration with Call Studio), Or its only for Call studio with Bot integration like GDF(Google dialog flow) and AWS Lex.
  • The Media forking connector which build by you can be used on production setup or its meant for Lab practice only.
  • And to use the Media forking connector application CUBE is mandate or it will work with Old 2200 ingress gateway?
Cisco Employee

One of the key reasons I didn't use the MRCP server approach was that it's only applicable to the IVR part of the call; I wanted a solution that would also allow various media stream processing scenarios between caller and agent.   What I provided is custom code primarily to allow anyone interested to get started very quickly, explore some innovative speech scenarios and investigate their viability without having to commit significant investment to their own science project.   

As it's custom, all the normal custom solution caveats apply with regard to lack of official support, level of testing, not being an official part of the product, etc.   It's clearly not intended for production and any use for live call traffic is at own risk.   

Regarding IOS device, I typically use vCUBE but have used it on 29xx/39xx hardware with SIP->SIP calls.   It's just making use of the gateway media forking XMF interface, exactly the same thing that CUCM uses for network based recording.  I think you might hit problems if you try to use ISDN->SIP. 

VIP Advocate

The UniMRCP approach seems to be the most widely used and will give you the most resources when you run into an issue.





I am also having the same issue with AWS Cloud platform, Thanks for the solution that i got from here.

Recognize Your Peers
Content for Community-Ad