sij/khoj
1
0
Fork 0
mirror of https://github.com/khoj-ai/khoj.git synced 2024-11-27 17:35:07 +01:00
Code Issues Projects Releases Packages Wiki Activity Actions

Update chat actor unit tests for greater accuracy and benchmarking

Browse source
This commit is contained in:
sabaimran 2023-08-01 12:24:43 -07:00
parent f409e16137
commit d8fa967b43
1 changed files with 28 additions and 8 deletions
Show stats Download patch file Download diff file Expand all files Collapse all files

34
tests/test_gpt4all_chat_actors.py
View file

@ -76,9 +76,29 @@ def test_extract_question_with_date_filter_from_relative_month(loaded_model):
# ----------------------------------------------------------------------------------------------------
@pytest.mark.xfail(reason="Chat actor still isn't very date aware nor capable of formatting")
@pytest.mark.chatquality
@freeze_time("1984-04-02")
def test_extract_question_with_date_filter_from_relative_year(loaded_model):
def test_extract_question_with_date_filter_from_relative_year():
# Act
response = extract_questions_offline("Which countries have I visited this year?")
# Assert
expected_responses = [
("dt>='1984-01-01'", ""),
("dt>='1984-01-01'", "dt<'1985-01-01'"),
("dt>='1984-01-01'", "dt<='1984-12-31'"),
]
assert len(response) == 1
assert any([start in response[0] and end in response[0] for start, end in expected_responses]), (
"Expected date filter to limit to 1984 in response but got: " + response[0]
)
# ----------------------------------------------------------------------------------------------------
@pytest.mark.chatquality
@freeze_time("1984-04-02")
def test_extract_question_includes_root_question(loaded_model):
# Act
response = extract_questions_offline("Which countries have I visited this year?", loaded_model=loaded_model)
@ -107,12 +127,12 @@ def test_extract_multiple_implicit_questions_from_message(loaded_model):
response = extract_questions_offline("Is Morpheus taller than Neo?", loaded_model=loaded_model)
# Assert
expected_responses = [
("morpheus", "neo", "height", "taller", "shorter"),
]
assert len(response) == 3
assert any([start in response[0].lower() and end in response[1].lower() for start, end in expected_responses]), (
"Expected two search queries in response but got: " + response[0]
expected_responses = ["height", "taller", "shorter", "heights"]
assert len(response) <= 3
for question in response:
assert any([expected_response in question.lower() for expected_response in expected_responses]), (
"Expected chat actor to ask follow-up questions about Morpheus and Neo, but got: " + question
)