I tried the " Core request to execute a custom action" - Rasa & Rasa Pro Documentation to call the actions server directly from my front-end for computer generated intents to bypass the Rasa server. I could get it to run the action code, but I could not get it to return the uttered message the normal way. I got a large dict as the return message instead. I did set the input_channel to “rest” like it would if I ran any normal actions invokation.
Any idea on what is wrong? Thanks Mats
Hi Mats, what you’ve got is actually what was expected as defined by the docs you’ve linked. However, from my understanding, you’d like to bypass the NLU portion but as if the message was normally sent.
For this use case, you can use:
/mood_great type of utterance where you send a message as if the user’s intent was mood_great and then the usual processing will be done and appropriate action will get triggered (as detected by the policies).
Let me know if this answer’s your question.
Regards, Nikola
Thanks Nikola, I have tried that vesion, but Rasa server (NLU etc) still does quite a bit of work. Almost no differernce, from the standard way. When I used the method I described, I shaved off almost 50% of the processing time. But I could not get it to send the message back to the original user. I tried setting the input channel to “rest” which is the same as for the original message, but it did not help.