Hey everyone, got a quick question here about implementing Rasa stack in our solution. We currently have an API layer built using flask socketIO and Deepspeech. We are trying to find the best way to programmatically take the text response from DeepSpeech and pass it into Rasa. However, we are not finding any true implementation of this. Any help, guidance would work. For example, how would we create a custom channel to pass messages when we are not using HTTP as the transport method. We are basically trying to import our rasa server module into deepspeech and then call a function that passes Deepspeech’s output into Rasa. Any thoughts?
If my understanding is right you may want to use handle_message() method. Like Agent.load().handle_message(your_message, sender_id = id) which returns the response.
Yeah, we have tried that. But from my understanding that is a single reaction. We are looking into opening a channel for continuous communication. I am looking at the Console Input Channel right now and how we can modify it to meet our needs. We are also having some issues right now running Deepspeech and Rasa Stack from the same script. Both use tensorflow, but Deepspeech somehow blocks access to Tensorflow. I think I found the issue within Tensorflow itself. It appears that it doesn’t allow for object sharing. Gotta do more digging. Thanks for the help though!
1 Like