Hello,
we're building an integration with your Bot Connector API and were able to set up an automated response mechanism through our platform.
The setup consists of a "web messenger" connected to an "inbound message flow" that immediately triggers a "call bot connector" node which forwards inputs to us.
In our scenarios, we want to forward the session to our bot for a couple of inputs until we're able to gather some required data from the user, then "complete" the bot session with a "Success" intent and additional parameters that are then consumed in the architect flow.
We're always sending a "Success" intent with a confidence of 1 with every response and rely on the "parameters" option in the "botState: COMPLETE" response to pass conversation results to Genesys.
We're able to receive messages in our platform and our responses get rendered in the web messenger, however, at some occasions, the bot connector session will "break" after a response. The "failure" branch of the "call bot connector" node gets executed and we're getting a "NoMatchError" saying "The bot session has failed", with no further information.
Our responses are shaped like this:
{
"replymessages": [
{
"type": "Text",
"text": "Hello there!"
}
],
"parameters": {},
"intent": "Success",
"confidence": 1,
"botState": "MOREDATA"
}
From our experience, even if we hardcode our platform to return that response to a request and send a couple of user messages through the web messenger, we have a ~30% chance that the bot session will break after the response. The side-effect of this is that the bot session is interrupted prematurely, breaking the end-to-end user experience.
Since the same response to the same request has no predictable outcome, we assume that this may be a bug in the bot connector API.