First you need to create a new test with a test_name, which is a set of n hyper-realistic AI personas. You can run the same test to compare different versions of your models.

Create a test within your python code using these parameters and function:

ParameterExample valueTypeDescription
test_name“TestA”stringThe name you choose for your test.
chatbot_role“education tutor”stringSee supported chatbot role use cases here.
industry“Education”stringThe industry of your chatbot
n20integerNumber of AI personas in a test
topic (Optional)“Math”stringTopic of the conversations the simulated personas will chat about
intents (Optional)[“help with homework”, “study for test”]list[string]Intents (goals) the personas want to achieve (sample from list)
behavior_instructions (Optional)[“ask many follow up questions”, “ask for clarification and a summary with bullet points”]list[string]If passed, detailed instructions for the behavior logic of the AI personas (sample from list)
language (Optional)“Spanish”stringDefault is English
demographics_dict (Optional)see example belowdictDictionary with detailed values for demographics
writing_styles (Optional)[“concise, all lowercase, with abbreviations”, “lengthy and with many typos”]list[string]Writing styles of the AI personas (sample form list)

Example of demographics_dict (optional)

demographics_dict = {
  "regions": ["US", "UK", "CA"],
  "age_range": [18, 65],
  "female_ratio": 0.5,
  "non-binary_ratio": 0.1,
  "openness_range": [0.0, 1.0],
  "conscientiousness_range": [0.0, 1.0],
  "extraversion_range": [0.0, 1.0],
  "agreeableness_range": [0.0, 1.0],
  "neuroticism_range": [0.0, 1.0]
}

The function does not return any value. Only this message should be printed:

Test successfully initialized. Test name: (test_name). Personas created: (n).