Rasa NLG Server

aquila0104 · March 2, 2022, 10:26pm

I am making a project in which I want to implement custom NLG for the chatbot. I have found the code on the rasa github for nlg_server.py and have got it to run in command line, receiving post requests from a python script I worked up for testing. How do I get rasa core to communicate with this NLG server to receive user inputs and respond back with a custom generated response. I have found nothing in the documentation indicating the implementation of this besides adding it to the endpoints.yml file of the project.

stephens · March 2, 2022, 11:08pm

You would create one or more custom actions (fallback action could be a catchall depending on your use case). The custom actions call the NLG and then utter_message.

aquila0104 · March 3, 2022, 7:23am

custom action calling the NLG would just be using the requests library to send a post request to the NLG server? or is in done through specifying the NLG used as a parameter within the run() method of the Action like so

 | async run(output_channel: "OutputChannel", nlg: "NaturalLanguageGenerator", tracker: "DialogueStateTracker", domain: "Domain") -> List[Event]

stephens · March 3, 2022, 4:56pm

Sorry, I gave you incorrect info on NLG. To point Rasa at your NLG, you configure this in the endpoints.yml as shown here.

nlg:
  url: http://localhost:5055/nlg

aquila0104 · March 3, 2022, 8:26pm

So beyond that, there should be no further set up? apart from of course, implementing your own response generation, as I have implemented that into my endpoints.yml file, but from what I can tell it yielded no results.

stephens · March 3, 2022, 9:07pm

Have you tested your endpoint with curl or postman? Does your endpoint calls from Rasa? What does the debug output from Rasa show both at startup and at points in the conversation when the NLG should be called?

aquila0104 · March 4, 2022, 10:00am

From my testing, when I run the nlg_server.py file, the server is running and sending requests from postman appears to work, currently getting an attribute error with the requests I am sending however.

in running rasa shell --debug the only reference I see to rasa.core.nlg is this…

2022-03-04 09:55:45 DEBUG rasa.core.nlg.generator - Instantiated NLG to ‘TemplatedNaturalLanguageGenerator’.

and there is nothing within debug referring to nlg after inputting a message to the bot

aquila0104 · March 4, 2022, 10:01am

I am confused at the stage of calling NLG, is this something done by the framework from it being tied to an endpoint or something I should be implementing into the custom action calls which I am writing?

stephens · March 4, 2022, 6:38pm

Rasa should call your nlg when the dialogue predicts an utterance.

- rule:
  steps:
  - intent: greet
  - action: utter_greet

When the call to utter_greet in this story is predicted, we should see the debug output in rasa shell -debug calling your NLG and you should add log messages or run in a debugger to confirm it’s being called.

Show the debug output when an utterance is run.

aquila0104 · March 8, 2022, 10:20am

Hi,

Apologies for a late reply, I have produced the debug output as you explained and got this below

2022-03-08 10:18:18 DEBUG 2022-03-08 10:18:18 DEBUG 2022-03-08 10:18:18 DEBUG 2022-03-08 10:18:18 DEBUG 2022-03-08 10:18:18 DEBUG 2022-03-08 10:18:18 DEBUG 2022-03-08 10:18:18 DEBUG ect at 0x00000214E3EA1130>, 2022-03-08 10:18:18 DEBUG module: None session_started_metadata: None 2022-03-08 10:18:18 DEBUG 5}’ and entities 2022-03-08 10:18:18 DEBUG module: None session_started_metadata: None 2022-03-08 10:18:18 DEBUG 2022-03-08 10:18:18 DEBUG [state 1] user intent: 2022-03-08 10:18:18 DEBUG 2022-03-08 10:18:18 DEBUG [state 1] user text: 2022-03-08 10:18:18 DEBUG 2022-03-08 10:18:18 DEBUG [state 1] user intent: 2022-03-08 10:18:18 DEBUG 2022-03-08 10:18:19 DEBUG 2022-03-08 10:18:19 DEBUG -1.3990422487258911. 2022-03-08 10:18:19 DEBUG 2022-03-08 10:18:19 DEBUG 2022-03-08 10:18:19 DEBUG 2022-03-08 10:18:19 DEBUG 2022-03-08 10:18:19 DEBUG 2022-03-08 10:18:19 DEBUG 2022-03-08 10:18:19 DEBUG tion object at 0x00000214 2022-03-08 10:18:19 DEBUG ": null, “buttons”: 2022-03-08 10:18:19 DEBUG [state 1] user intent: [state 2] user intent: 2022-03-08 10:18:19 DEBUG 2022-03-08 10:18:19 DEBUG [state 1] user intent: [state 2] user intent: 2022-03-08 10:18:19 DEBUG 2022-03-08 10:18:19 DEBUG event of type UserUttered 2022-03-08 10:18:19 DEBUG 2022-03-08 10:18:19 DEBUG 2022-03-08 10:18:19 DEBUG 2022-03-08 10:18:19 DEBUG 2022-03-08 10:18:19 DEBUG 2022-03-08 10:18:19 DEBUG rasa.core.lock_store - Issuing ticket for conversation ‘148e3342808442118b6eca0aea6234e0’. rasa.core.lock_store - Acquiring lock for conversation ‘148e3342808442118b6eca0aea6234e0’. rasa.core.lock_store - Acquired lock for conversation ‘148e3342808442118b6eca0aea6234e0’. rasa.core.tracker_store - Could not find tracker for conversation ID ‘148e3342808442118b6eca0aea6234e0’. rasa.core.processor - Starting a new session for conversation ID ‘148e3342808442118b6eca0aea6234e0’. rasa.core.processor - Policy prediction ended with events ‘[]’. rasa.core.processor - Action ‘action_session_start’ ended with events ‘[<rasa.shared.core.events.SessionStarted obj ActionExecuted(action: action_listen, policy: None, confidence: None)]’. rasa.core.processor - Current slot values: rasa.core.processor - Received user message ‘hello’ with intent ‘{‘name’: ‘greet’, ‘confidence’: 0.4546232573913670 ‘[{‘value’: ‘neu’, ‘confidence’: 1.0, ‘entity’: ‘sentiment’, ‘extractor’: ‘sentiment_extractor’}]’ rasa.core.processor - Current slot values: rasa.core.processor - Logged UserUtterance - tracker now has 4 events. rasa.core.policies.memoization - Current tracker state: greet | user entities: (‘sentiment’,) | previous action name: action_listen rasa.core.policies.memoization - There is no memorised next action rasa.core.policies.rule_policy - Current tracker state: hello | previous action name: action_listen rasa.core.policies.rule_policy - There is no applicable rule. rasa.core.policies.rule_policy - Current tracker state: greet | user entities: (‘sentiment’,) | previous action name: action_listen rasa.core.policies.rule_policy - There is no applicable rule. rasa.core.policies.unexpected_intent_policy - Querying for intent greet. rasa.core.policies.unexpected_intent_policy - Score for intent greet is 1.8118122816085815, while threshold is rasa.core.policies.unexpected_intent_policy - Top 5 intents (in ascending order) that are likely here are: [('expl ain_feeling', -1.48593), ('give_time', 1.3062135), ('ask_help', 1.3728172), ('greet', 1.8118123), ('mood_unhappy', 2.0791276)]. rasa.core.policies.ted_policy - TED predicted ‘utter_greet’ based on user intent. rasa.core.policies.ensemble - Made prediction using user intent. rasa.core.policies.ensemble - Added DefinePrevUserUtteredFeaturization(False) event. rasa.core.policies.ensemble - Predicted next action using policy_3_TEDPolicy. rasa.core.processor - Predicted next action ‘utter_greet’ with confidence 0.99. rasa.core.processor - Policy prediction ended with events ‘[<rasa.shared.core.events.DefinePrevUserUtteredFeaturiza E3EA4C40>]’. rasa.core.processor - Action ‘utter_greet’ ended with events ‘[BotUttered(‘Hey!’, {“elements”: null, "quick_replies null, “attachment”: null, “image”: null, “custom”: null}, {“utter_action”: “utter_greet”}, 1646734699.2455828)]’. rasa.core.policies.memoization - Current tracker state: greet | user entities: (‘sentiment’,) | previous action name: action_listen greet | user entities: (‘sentiment’,) | previous action name: utter_greet rasa.core.policies.memoization - There is no memorised next action rasa.core.policies.rule_policy - Current tracker state: greet | user entities: (‘sentiment’,) | previous action name: action_listen greet | user entities: (‘sentiment’,) | previous action name: utter_greet rasa.core.policies.rule_policy - There is no applicable rule. rasa.core.policies.unexpected_intent_policy - Skipping predictions for UnexpecTEDIntentPolicy as either there is no or there is an event of type ActionExecuted after the last UserUttered. rasa.core.policies.ted_policy - TED predicted ‘action_listen’ based on user intent. rasa.core.policies.ensemble - Predicted next action using policy_3_TEDPolicy. rasa.core.processor - Predicted next action ‘action_listen’ with confidence 1.00. rasa.core.processor - Policy prediction ended with events ‘[]’. rasa.core.processor - Action ‘action_listen’ ended with events ‘[]’. rasa.core.lock_store - Deleted lock for conversation ‘148e3342808442118b6eca0aea6234e0’.

keshabmanni · July 18, 2022, 7:28am

For nlg server on which route should it be listening for that post request?

localhost:5055/ or localhost:5055/nlg ?

Topic		Replies	Views
How to debug rasa-core? Rasa Open Source	14	10348	December 13, 2022
Using utter message with http server Rasa Open Source	0	703	March 31, 2019
Rasa nlu server is not called by rasa core Rasa Open Source	10	1172	October 23, 2019
Implementing Custom Actions with Python Rasa_core 0.11.1 Rasa Open Source	1	1874	September 10, 2018
When to use NLG Server vs Custom action Rasa Open Source	1	325	June 13, 2024

Rasa NLG Server

Related topics