Skip to content

Modular INtelligent Evaluation for a Reliable Voice Assistant

License

Notifications You must be signed in to change notification settings

NeonGeckoCom/neon-minerva

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

45 Commits
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

Neon Minerva

Neon Minerva (Modular INtelligent Evaluation for a Reliable Voice Assistant) provides tools for testing skills.

Install the Minerva Python package with: pip install neon-minerva The minerva entrypoint is available to interact with a bus via CLI. Help is available via minerva --help.

Installation

If testing Padatious intents, the following system packages must be installed before installing this package:

sudo apt install swig libfann-dev

To install this package from PyPI, simply run:

pip install neon-minerva

If testing with Padatious, install with the padatious extras:

pip install neon-minerva[padatious]

Usage

This package provides a CLI for local testing of skills. Skills installed with pip can be specified by entrypoint, or skills cloned locally can be specified by root directory.

Resource Tests

To test that skill resources are defined for all supported languages, minerva test-resources <skill-entrypoint> <test-file>

  • <skill-entrypoint> is the string entrypoint for the skill to test as specified in setup.py OR the path to the skill's root directory
  • <test-file> is a relative or absolute path to the resource test file, usually test_resources.yaml

example test_resources.yaml:

# Specify resources to test here.

# Specify languages to be tested
languages:
  - "en-us"
  - "uk-ua"

# vocab is lowercase .voc file basenames
vocab:
  - ip
  - public
  - query

# dialog is .dialog file basenames (case-sensitive)
dialog:
  - dot
  - my address is
  - my address on X is Y
  - no network connection
  - word_public
  - word_local
# regex entities, not necessarily filenames
regex: []
intents:
  # Padatious intents are the `.intent` file names
  padatious: []
  # Adapt intents are the name passed to the constructor
  adapt:
    - IPIntent

Intent Tests

To test that skill intents match as expected for all supported languages, minerva test-intents <skill-entrypoint> <test-file>

  • <skill-entrypoint> is the string entrypoint for the skill to test as specified in setup.py OR the path to the skill's root directory
  • <test-file> is a relative or absolute path to the resource test file, usually test_intents.yaml
  • The --padacioso flag can be added to test with Padacioso instead of Padatious for relevant intents

example test_intents.yaml:

en-us:
  IPIntent:
  - what is your ip address
  - what is my ip address:
    - IP
  - what is my i.p. address
  - What is your I.P. address?
  - what is my public IP address?:
    - public: public

uk-ua:
  IPIntent:
  - шо в мене за ай пі:
    - IP  
  - покажи яка в мене за мережа:
    - IP
  - покажи яка в мене публічний ай пі адреса:
    - public: публічний

Test Configuration

The following top-level sections can be added to intent test configuration:

  • unmatched intents: dict of lang to list of utterances that should match no intents. Note that this does not test for CommonQuery or CommonPlay matches.
  • common query: dict of lang to list of utterances OR dict of utterances to expected: callback_data (list keys or dict data), min_confidence, and max_confidence
  • common play: TBD

Advanced Usage

In addition to convenient CLI methods, this package also provides test cases that may be extended.

Skill Unit Tests

neon_minerva.tests.skill_unit_test_base provides SkillTestCase, a class that supplies boilerplate setup/teardown/mocking for testing a skill. An example skill test implementation could look like:

from os import environ
from neon_minerva.tests.skill_unit_test_base import SkillTestCase

environ['TEST_SKILL_ENTRYPOINT'] = "my_skill.test"

class MySkillTest(SkillTestCase):
    def test_skill_init(self):
        self.assertEqual(self.skill.skill_id, "my_skill.test")
    ...

Be sure to review the base class for mocked methods and test paths as these may change in the future.

Chatbot Unit Tests

neon_minerva.chatbots contains mocked data for testing as well as some utility methods. neon_minerva.tests.chatbot_v1_test_base provides TestSubmind which may be extended to test a submind bot in a mocked v1 environment. For example:

from os import environ
from datetime import datetime
from chatbot_core.utils.enum import ConversationState

from neon_minerva.tests.chatbot_v1_test_base import TestSubmind
from neon_minerva.chatbots.test_constants import PROMPT, RESPONSES

environ["TEST_BOT_ENTRYPOINT"] = "tester"


class TestTester(TestSubmind):
    def test_submind_chatbot(self):
        self.submind.state = ConversationState.RESP
        response = self.submind.ask_chatbot("testrunner", PROMPT,
                                            datetime.now().strftime(
                                                "%I:%M:%S %p"))
        self.assertIsInstance(response, str)
        self.assertIsNotNone(response)

Make sure to install the chatbots extra to use this test case

About

Modular INtelligent Evaluation for a Reliable Voice Assistant

Resources

License

Stars

Watchers

Forks

Packages

No packages published

Languages