cogint / dialogflow-delay-test

Test Dialogflow and webhook delays for specific sequences from the command line

Dialogflow Delay Tester

Use this script to gather response and webhook delay information from Dialogflow. This can be used to diagnose webhook delay issues and adjust your conversational design for delays. Results are saved to a CSV file for external analysis.

Input

node dfdelay.js --project my-agent-projectid --queries custom_queries.json --runs 10 --commands WelcomeEvent 15 Buy 600 . WelcomeEvent 10 Buy 14 Sell 15

Output: results.csv

SessionId, timestamp, query, responseTimeMs, webhooktimeMs
2f538d5e-b5a8-4d29-af93-07d7729ec31e, 1600815123806, WelcomeEvent, 861, 194
2f538d5e-b5a8-4d29-af93-07d7729ec31e, 1600815139669, Buy, 793, 88
e6057528-2b18-42c2-8bb8-22c78ccd889a, 1600815518319, WelcomeEvent, 848, 143
e6057528-2b18-42c2-8bb8-22c78ccd889a, 1600815534172, Buy, 570, 50
b125db2b-5484-4b9c-bf32-91688ca0dd7f, 1600816134759, Sell, 580, 63

This projects uses the Dialogflow NodeJS client to send queries to a Dialogflow agent after a specified delay.

Installation

npm clone [this repo name]
npm install

See below for setting required environment variables.

Dialogflow dependencies

Make sure the GOOGLE_APPLICATION_CREDENTIAL environment variable is set - i.e. export GOOGLE_APPLICATION_CREDENTIALS=your-df-agent-uuid-8dc0cd8a9c7c.json

Make sure to obtain the project ID for your agent - i.e. unit-converter-nyofsy. This can be found in Google Cloud Console or under the General tab of Agent Settings in the Dialogflow console.

Usage

Query file setup

Queries are stored in queries.json or specified on the command line. The query format matches the request.input value passed to sessionClient.detectIntent in the Dialogflow NodeJS client. The query should look something like:

[
  {
    "name":"WelcomeEvent",
    "query":{
      "event":{
        "name":"Welcome",
        "languageCode":"en-US"
      }
    }
  },
  {
    "name":"Buy",
    "query":{
      "text":{
        "text":"I would like to buy a widget",
        "languageCode":"en-US"
      }
    }
  
    
]

Events and text inputs are allowed. You can add contexts and any other valid sessionClient.detectIntent input. You must specify every intent you plan to use.

Commandline options

Use the --commands or -c to pass an series of query names followed by a number representing the amount of time to wait after the query has been sent. Use a period ( . ) after a wait time to start a new session instead of continuing the previously created one.

The script reads from this queries file and runs comments. The --runs option indicates the number of times the script should repeat.

Example

Run the following 10 times:

1. Fire the WelcomeEvent
2. wait 15 seconds
3. Send the Buy query
4. Wait 600 seconds
5. Start a new session: .
6. Fire the WelcomeEvent
7. Wait 10 seconds
8. Send the Buy query
9. Wait 14 seconds
10. Send the Sell query
11. Send Wait 15 seconds

node dfdelay.js --project my-agent-projectid --queries custom_queries.json --runs 10 --commands WelcomeEvent 15 Buy 600 . WelcomeEvent 10 Buy 14 Sell 15

ToDo

• Error checking - there's none of it
• Verify the queries named in the command string match the queries in the json file
• Form queries from the command line so the query.json isn't needed