When designing a chatbot test strategy from scratch, there are often requirements like:
We are running a chatbot for all of our five brands. The conversations are basically the same for all of our brands, but there are slight differences.
We have a development, a test and a production environment and we have to test on all of them
If you can see yourself in one of the above points, then read on. This article will present a best practice technique to prepare Botium for those requirements.
Features and Techniques Overview
For setting up Botium we will use several techniques that are actually independent of each other, but in combination they are incredibly powerful.
When asserting chatbot answers, wildcards ("jokers") can be used to accept any text. This is nothing special to Botium, but it comes in handy when asserting content for different brands.
Scripting Memory / Test Parameter Store
With the Botium Scripting Memory it is possible to inject dynamically generated or static values into your test cases. We will use this concept to set different conversation parameters for each environment your tests should run against.
For informatin about the scripting memory, see Botium Scripting Memory.
Test Set Dependencies
In Botium it is possible to define dependencies between test sets and combine them into a single test set. We will use this technique to separate the different requirements into individual test sets and combine them as needed.
Environment-specific Test Project Capabilities
In Botium it is possible to define environment-specific capabilities which will be merged with the chatbot capabilities. So it is sufficient to define the basic chatbot capabilities only once, and then add environment-specific adaptions on Test Project level (f.e. selecting a different IBM Watson Assistant workspace or a different HTTP endpoint).
Resulting Test Suite
In the end there will be several new objects in Botium:
There will be only one chatbot defined
There will be one shared test set holding the test cases valid for all brands (with placeholders)
For each brand, there will be a brand-specific test set with brand-specific test cases and brand-specific scripting memory
For each combination of brand + environment you have to run your tests, there will be one test project combining:
the chatbot, enhanced with environment-specific capabilities
the brand-specific test set with the scripting memory files
the shared test set
Step By Step
Now comes the interesting part - follow those steps to setup the basic structure in Botium.
1. Connect to IBM Watson Assistant
In this example, we will use IBM Watson Assistant, but the same principle works for all supported technologies. We are connecting the chatbot to the Assistant's development workspace, so we can use it for developing the test cases. When running test cases later we will connect to the environment-specific Assistant workspaces by overwriting this from the Test Project.
2.Create a Shared Convos Test Set with Wildcards
Create a test set named Shared Convos in Botium. Add some first Convos in the Visual Convo Designer. The convos should map the conversation structure, and they should be free from any brand-specific content by using wildcards.
We have here a convo named TC_HELLO, which sends a default greeting to the chatbot, and expects a default greeting back:
Here is the corresponding BotiumScript (for Copy&Paste):
Hello, this is Heinz, the chatbot of *. How can I help you ?
Note the use of the * as a wildcard - this is the spot where the brand name would be shown.
3. Create Brand-Specific Test Sets with Scripting Memory (optional)
The above test case would assert that
the chatbot introduces itself as Heinz
and that there is any brand name included
But for your brands, you want to make sure that
Each brand chooses a different name for the chatbot
We want to additionally assert on the brand name (not accept just anything with the wildcard)
For each of the brands, create a new test set. The brand-specific parameters will be saved in Scripting Memory files. Create a test set named Params BRAND-1 and add a YAML-file named Scripting Memory:
In this file, we define the variables that we will use in our test cases, like chatbot name and brand name:
$brand_name: My first brand
For another brand, the test set Params BRAND-2 can look roughly the same, but with different variable values:
$brand_name: Another brand
Enable the Scripting Memory for these test sets in the Settings / Scripting Settings section
Enable the switch Enable Scripting Memory
Enable the switch Enable Test Parameter Store
4. Adapt Shared Convos with brand-specific content (optional)
The shared test cases from above now have to be changed to use the placeholders for the chatbot name and brand name, instead of just using a wildcard. Replace the corresponding spots in the test case with the variable name:
Hello, this is $chatbot_name, the chatbot of $brand_name. How can I help you ?
This means that before running a test case, the variables are filled from the scripting memory files you defined upfront, therefore replacing those variables with concrete chatbot names and brand names for doing the assertions.
5. Create Test Projects
Now go the Test Projects / Register Test Project to combine everything from above and apply environment-specific settings.
As Test Project Name choose something like BRAND-1 DEV
Select the chatbot
Select the Shared Convos Test Set and the brand-specific PARAMS BRAND-1 Test Set
Save the Test Project and immediately head over to the Settings tab
In the Advanced section, it is now possible to overwrite the capabilities from the chatbot with the environment-specific settings. You can find the name of the capability (the basic configuration items for the Botium Connectors) either in the connector documentation, or in the Advanced Mode of the Chatbot connector settings.
In this case, we have to overwrite the IBM Watson Assistant workspace ID to connect it to a different workspace. Repeat the above steps for the other brands and environments, and name the Test Projects accordingly (BRAND-2 TEST, BRAND-1 PROD, ...).
Everything is ready now for running your brand- and environment-specific test cases.
In this article you learned how to use Botium to prepare a test suite testing multiple chatbots for multiple brands on multiple environments without duplicating test cases, keeping the effort for future work as low as possible.