cupel 1.2.0

Context window management pipeline for LLM applications
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45

[test]
name = "ScaledScorer applies min-max normalization to inner scorer"
stage = "scoring"
scorer = "scaled"

# Inner scorer: reflexive (passes through futureRelevanceHint).
# Items have hints: 0.2, 0.5, 0.8
# min = 0.2, max = 0.8, range = 0.6
# Scaled scores:
#   0.2 → (0.2 - 0.2) / 0.6 = 0.0
#   0.5 → (0.5 - 0.2) / 0.6 = 0.5
#   0.8 → (0.8 - 0.2) / 0.6 = 1.0

[config]
inner_scorer = "reflexive"

[[items]]
content = "low"
tokens = 100
futureRelevanceHint = 0.2

[[items]]
content = "mid"
tokens = 100
futureRelevanceHint = 0.5

[[items]]
content = "high"
tokens = 100
futureRelevanceHint = 0.8

[[expected]]
content = "low"
score_approx = 0.0

[[expected]]
content = "mid"
score_approx = 0.5

[[expected]]
content = "high"
score_approx = 1.0

[tolerance]
score_epsilon = 1e-9