UNPKG

axiom

Version:

Axiom AI SDK provides - an API to wrap your AI calls with observability instrumentation. - offline evals - online evals

github.com/axiomhq/ai

axiomhq/ai

78 lines (41 loc) • 1.46 kB
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
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
[**axiom v0.51.1**](../../../README.md) *** [axiom](../../../README.md) / [evals/online](../README.md) / OnlineEvalParams # Type Alias: OnlineEvalParams\<TInput, TOutput, TScorers\> > **OnlineEvalParams**\<`TInput`, `TOutput`, `TScorers`\> = `object` Options for online evaluation. ## Type Parameters ### TInput `TInput` ### TOutput `TOutput` ### TScorers `TScorers` *extends* readonly `ScorerEntry`\<`TInput`, `TOutput`\>[] = readonly `ScorerEntry`\<`TInput`, `TOutput`\>[] ## Properties ### capability > **capability**: `string` High-level capability being evaluated (e.g., 'qa', 'summarization') *** ### input? > `optional` **input**: `TInput` Input to pass to scorers (optional - only needed for input+output scorers) *** ### links? > `optional` **links**: `SpanContext` \| `SpanContext`[] Explicit SpanContext(s) to link the eval span to originating generation span(s). When omitted, the active span's context is used automatically. Use this for deferred evaluation when onlineEval is called after the originating span has completed. Supports both single context and multiple contexts for multi-span linking. *** ### output > **output**: `TOutput` Output to evaluate *** ### scorers > **scorers**: `EnsureUniqueScorerNames`\<`TScorers`\> Scorers or precomputed scores. Supports optional per-scorer sampling. *** ### step? > `optional` **step**: `string` Specific step within the capability (e.g., 'answer', 'extract')