davidnich/ondewo-bpi

Ondewo Bpi

If you have a grpc config with at least the following:

{
  "grpc_cert": "<GRPC-CERT-HERE>"
}

locally with the name 'config.json', you should be able to run the docker-compose files after building the image. Look in the makefile.

The BPI (Business Process Integration) Module

This module sits between the user and CAI:

                _______           _______
  ___          |       |         |       |
  NLU ==grpc==>|  BPI  |==grpc==>|  CAI  |
 |___|<==grpc==|       |<==grpc==|       |
               |_______|         |_______|

and has full access to the outgoing requests and the returning response. It also knows the location of CAI and has authorization, meaning it can edit the state of CAI directly with additional information.

A simple example of a server using the BPI looks like this:

import logging

from ondewo.nlu import session_pb2
from ondewo_bpi.bpi_server import BpiServer
import ondewo_bpi.helpers as helpers


class MyServer(BpiServer):
    def __init__(self) -> None:
        super().__init__()                        # initialize the server (parent class)
        self.register_handlers()                  # add handlerss for intents

    def register_handlers(self) -> None:          # the mapping from handled intents to functions
        self.register_handler(
            intent_name='i.my_handled_intent',    # a user-created intent
            handler=self.reformat_text_in_intent, # the function to handle it
        )
        self.register_handler(
            intent_name='Default Fallback Intent',  # a default (system created) intent
            handler=self.handle_default_fallback,   # the function to handle it
        )

    def reformat_text_in_intent(      # the handler functions take responses in an out, and can process parts of the response
        self, response: session_pb2.DetectIntentResponse
    ) -> session_pb2.DetectIntentResponse:
        return helpers.replace_text_in_response(
            search="<REPLACE:REPLACE_THIS_TEXT>",   # this entire pseudo-command would go in a text response for example
            replace="new text",
            response=response
        )

    def handle_default_fallback(      # the handler functions can also just trigger events and leave the response unchanged
        self, response: session_pb2.DetectIntentResponse
    ) -> session_pb2.DetectIntentResponse:
        logging.warning("Default fallback was triggered!")
        return response

There is a more complete example in ondewo_bpi/example/example.py which is hooked up to run via the Dockerfile (just do docker-compose up -d && docker-compose logs -f).

Have a look at the docker-compose file, and the sample.env. The port that the bpi listens on functionally replaces the cai port, so at the frontend you just have to point to a different port to use the bpi, there is no other difference. And the cai port will still be available as well, so switching between using CAI directly and using the BPI is just a matter of changing a single variable, the port.

CAI	BPI
HOST=localhost:50055	HOST=localhost:50051

BPI QA

There is also an example server for integrating both CAI and the QA. It sends requests to both servers and returns the response that is most accurate.

BPI SIP

There is an example for integration with ONDEWO-VTSI (Voice Telephone System Integration), however the sip client submodule is not publicly available. Please request access to it to build telephone-system compantible BPI servers.