qna | A simple library for interactive questions and answers | Chat library

You're trying to send the search value in the AJAX as JSON, but your PHP is not set up to read JSON. (And it makes no sense to use JSON anyway, just to send a single value).

Just send the form data in form-url-encoded format, as it would be if you submitted the form without AJAX. jQuery can help you with this - e.g.

What you want is called interruptions: https://docs.microsoft.com/azure/bot-service/bot-builder-howto-handle-user-interrupt

The core bot sample includes a somewhat complicated way of handling interruptions by creating a base CancelAndHelpDialog class that makes any dialog "interruptible," but there is a simpler way of doing it. The key is to call ContinueDialogAsync on every turn by default, and only don't call it if an interruption takes place on that turn. You can think of each turn as having three cases:

1. An interruption takes place
2. There's no interruption but there's an active dialog
3. There's no interruption and no active dialog

Here's how the logic might play out:

you cant add body in get request. but you can pass the data in the query param. like this

I'm glad that you are using "Always On". The QnA Maker team sometimes recommends adding quick availability/web test using Azure app insights monitor (see https://docs.microsoft.com/azure/azure-monitor/app/monitor-web-app-availability). This is essentially a probing service every few seconds that can resolve the cold-start problem.

It looks like the LUIS application isn't public.

Check for this in the LUIS portal:

UPDATE: Added the LUIS configuration page in Azure portal

To get the api keys and hostname, you have to check the Keys and Endpoint section on your LUIS PUBLISHING resource in Azure (you shouldn't use the AUTHORING resource, b/c it has a limited request count and you might deplete it).

Add a UniqueConstraint to Answer so no combination of Question and Profile can be repeated: https://docs.djangoproject.com/en/3.1/ref/models/constraints/#django.db.models.UniqueConstraint

You want to create a Base class that extends the MultiDex if you don't have one.

Please look over the 80.skills-simple-bot-to-bot sample for guidance on how to connect two bots. The sample demonstrates a "root bot" (your Bot A) calling the "skill bot" (your Bot B) and executing a dialog before returning control back to the root bot. The conversation Id is maintained across the root and skill bot by use of the SkillConversationIdFactory.

Reference the Implement a skill docs for additional information on setting up and consuming a skill.

I can't help you with C# specifically, but I am doing this type of testing with my nodejs bot so this may be of some help. I replied to another of you questions here where I've given my specific nodejs code for testing. This answer will focus more on the approach to automate testing and use Azure Pipelines to execute it.

I'm using mocha as my test framework, but if there is something similar in C#, this method should work. This is where my knowledge is incomplete; I don't know what the C# equivalent of npm test is.

Anyway, I'm hopeful that my approach in nodejs is transferrable for you in C#. I want to mention up from that there are really two different approaches you can take to implementing your test.

1. Create a test bot adapter and simulate your entire bot, testing specifically recognizing the utterance, sending it to QnA Maker, processing the response as necessary, and posting the response to user (and validating that response content).
2. Make a REST call directly to QnA maker and validate the result

If you are concerned about the bot processing the activities correctly, invoking the QnA Maker connecter at the right time, parsing the response correctly, AND having the correct answer, go with option 1. If you simply want to make sure your QnA Service is functioning and returning the expected results, you can go with method 2. Personally I use method 1, but I mock the webservice response (using nock from the npm library) so I'm not testing QnA Maker at all, only the bot. Having live tests in your pipeline (if your tests determine if a release is created or not) can be problematic.

If you check the answer on your other question, that should give you the approach needed for option 1. For option 2, I would suggest using an Azure Function and just write your own test to make the REST call to QnA Maker and validate the result.

Either of these tests can be done in Azure Pipelines. For nodejs/option 1, I just run the npm test command which is configured to execute all of my test files. If any tests fail, the pipeline fails and (per my configuration) no release is created. I am unsure what the equivalent is for C#. For option 2, honestly I wouldn't even worry about putting it in the pipeline because it's independent from your bot code. Just set up your testing schedule in Azure Functions (if that's what you use) or whichever execution trigger you want, and have some actionable output. E.g. I have a live integration tester in Azure Functions which is a similar concept, and it sends out service down and service restored messages based on results and changes in status.