Change mod_dialogflow to receive optional parameter for enabling sentiment analysis

[ruipfmendes]

Description

Why

• The project I'm working on is using mod_dialogflow in the infrastructure of FreeSWITCH + Drachtio Server + a Drachtio App.
• Issue https://github.com/drachtio/drachtio-freeswitch-modules/issues/85 :
  • The goal in my project is to abstract ourselves from using Dialogflow's platform as much as possible and right now Drachtio allows us to automate almost all steps. However, one issue we are having is that we would always have to go to each Agent/Flow 's page to enable Sentiment Analysis (in the Advanced tab of https://dialogflow.cloud.google.com/#/editAgent/<projectId/ ).

What

StreamingDetectIntent request ( https://cloud.google.com/dialogflow/es/docs/reference/rpc/google.cloud.dialogflow.v2beta1#google.cloud.dialogflow.v2beta1.StreamingDetectIntentRequest ) allows sending query params:

The first message must contain session, query_input plus optionally query_params.

The query parameters (see here: https://cloud.google.com/dialogflow/es/docs/reference/rpc/google.cloud.dialogflow.v2beta1#google.cloud.dialogflow.v2beta1.QueryParameters ) can allow to set fields such as sentiment_analysis_request_config (enables sentiment analysis)

Tests

• Tested with usages of "dialogflow_start" where the new fields were ignored to guarantee retro compatibility
  • project-id of dialogflow-project-id:environment:region for example still works
  • No breaking changes for existing implementations ✅
• Tested with a flow/agent that had sentiment analysis disabled in Dialogflow's platform page
  • project-id of dialogflow-project-id:environment:region:speakingRate:pitch:volume:voice-name:voice-gender:effect:sentiment-analysis
  • Works for the new use case ✅ (of course users must keep in mind their Dialogflow agent must allow the use of sentiment analysis -- i.e., not be a free-tier agent)

[davehorton]

@ruipfmendes is it not necessary to add logic to the parser to return sentiment results?

[ruipfmendes]

@ruipfmendes is it not necessary to add logic to the parser to return sentiment results?

it seems that logic is already implemented in the parser here

I tested and got this result json in my logs which contains the sentiment_analysis_result:

mod_dialogflow.c:26 json payload for type dialogflow::intent: 
{
    ...
   "query_result":{
      ...
      "sentiment_analysis_result":{
         "query_text_sentiment":{
            "score":0.89999997615814209,
            "magnitude":0.89999997615814209
         }
      },
      ...
}