# tests/test_surprise_calc.py
"""Tests for surprise score calculation."""
from daem0nmcp.surprise import calculate_surprise, SurpriseCalculator
class TestCalculateSurprise:
"""Test surprise calculation."""
def test_novel_content_high_surprise(self):
# Use vectors that point in different directions (not just different magnitudes)
# Existing embeddings point in one direction
existing_embeddings = [
[1.0] * 128 + [-1.0] * 128, # About topic A
[1.0] * 128 + [-1.0] * 128, # About topic A
]
# New embedding points in a very different direction
new_embedding = [-1.0] * 128 + [1.0] * 128 # Opposite direction
surprise = calculate_surprise(new_embedding, existing_embeddings)
assert surprise > 0.5 # Should be high surprise
def test_similar_content_low_surprise(self):
existing_embeddings = [
[0.5] * 256,
[0.5] * 256,
]
new_embedding = [0.5] * 256 # Same
surprise = calculate_surprise(new_embedding, existing_embeddings)
assert surprise < 0.5 # Should be low surprise
def test_empty_existing_returns_max_surprise(self):
new_embedding = [0.5] * 256
surprise = calculate_surprise(new_embedding, [])
assert surprise == 1.0 # First memory is always surprising
def test_surprise_bounded_0_to_1(self):
existing = [[0.1] * 256, [0.9] * 256]
new = [0.5] * 256
surprise = calculate_surprise(new, existing)
assert 0.0 <= surprise <= 1.0
class TestSurpriseCalculator:
"""Test SurpriseCalculator class."""
def test_calculator_with_k_nearest(self):
calc = SurpriseCalculator(k_nearest=2)
existing = [
[0.1] * 256,
[0.2] * 256,
[0.9] * 256, # Outlier
]
new = [0.15] * 256 # Similar to first two
# Should only compare to 2 nearest, not the outlier
surprise = calc.calculate(new, existing)
assert surprise < 0.5
