localarena 0.1.0 → 0.2.0
This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.
- package/CHANGELOG.md +22 -0
- package/README.md +252 -140
- package/docs/provider-and-task-support.md +165 -0
- package/examples/evaluation-config.json +157 -0
- package/javascript/src/evaluation.js +1712 -0
- package/javascript/src/index.d.ts +545 -0
- package/javascript/src/index.js +49 -0
- package/javascript/src/providers.js +1303 -0
- package/javascript/src/report.js +355 -0
- package/javascript/src/tasks.js +828 -0
- package/package.json +7 -3
- package/schema/localarena-evaluation-config-v1.schema.json +612 -0
- package/schema/localarena-evaluation-run-v1.schema.json +1422 -0
|
@@ -0,0 +1,157 @@
|
|
|
1
|
+
{
|
|
2
|
+
"$schema": "../schema/localarena-evaluation-config-v1.schema.json",
|
|
3
|
+
"name": "Live provider comparison",
|
|
4
|
+
"concurrency": 3,
|
|
5
|
+
"repetitions": 1,
|
|
6
|
+
"include_content": false,
|
|
7
|
+
"models": [
|
|
8
|
+
{
|
|
9
|
+
"name": "llama-cpp-local",
|
|
10
|
+
"provider": "llamacpp",
|
|
11
|
+
"model": "local-model",
|
|
12
|
+
"parameters": {
|
|
13
|
+
"max_tokens": 128,
|
|
14
|
+
"temperature": 0
|
|
15
|
+
}
|
|
16
|
+
},
|
|
17
|
+
{
|
|
18
|
+
"name": "ollama-local",
|
|
19
|
+
"provider": "ollama",
|
|
20
|
+
"model": "qwen3:0.6b",
|
|
21
|
+
"parameters": {
|
|
22
|
+
"max_tokens": 128,
|
|
23
|
+
"temperature": 0
|
|
24
|
+
}
|
|
25
|
+
},
|
|
26
|
+
{
|
|
27
|
+
"name": "lm-studio-local",
|
|
28
|
+
"provider": "lmstudio",
|
|
29
|
+
"model": "local-model",
|
|
30
|
+
"parameters": {
|
|
31
|
+
"max_tokens": 128,
|
|
32
|
+
"temperature": 0
|
|
33
|
+
}
|
|
34
|
+
},
|
|
35
|
+
{
|
|
36
|
+
"name": "openrouter-hosted",
|
|
37
|
+
"provider": "openrouter",
|
|
38
|
+
"model": "openai/gpt-4.1-mini",
|
|
39
|
+
"api_key_env": "OPENROUTER_API_KEY",
|
|
40
|
+
"headers": {
|
|
41
|
+
"HTTP-Referer": "https://example.com/localarena",
|
|
42
|
+
"X-OpenRouter-Title": "LocalArena example"
|
|
43
|
+
},
|
|
44
|
+
"parameters": {
|
|
45
|
+
"max_tokens": 128,
|
|
46
|
+
"temperature": 0
|
|
47
|
+
}
|
|
48
|
+
},
|
|
49
|
+
{
|
|
50
|
+
"name": "openai-hosted",
|
|
51
|
+
"provider": "openai",
|
|
52
|
+
"model": "gpt-4.1-mini",
|
|
53
|
+
"api_key_env": "OPENAI_API_KEY",
|
|
54
|
+
"parameters": {
|
|
55
|
+
"max_tokens": 128,
|
|
56
|
+
"temperature": 0
|
|
57
|
+
}
|
|
58
|
+
},
|
|
59
|
+
{
|
|
60
|
+
"name": "custom-compatible",
|
|
61
|
+
"provider": "custom",
|
|
62
|
+
"base_url": "https://inference.example.com/v1",
|
|
63
|
+
"model": "custom-model",
|
|
64
|
+
"api_key_env": "LOCALARENA_CUSTOM_API_KEY",
|
|
65
|
+
"headers_env": {
|
|
66
|
+
"X-Service-Token": "LOCALARENA_CUSTOM_HEADER"
|
|
67
|
+
},
|
|
68
|
+
"policy": {
|
|
69
|
+
"timeout": 120,
|
|
70
|
+
"max_attempts": 1
|
|
71
|
+
},
|
|
72
|
+
"parameters": {
|
|
73
|
+
"max_tokens": 128,
|
|
74
|
+
"temperature": 0
|
|
75
|
+
}
|
|
76
|
+
}
|
|
77
|
+
],
|
|
78
|
+
"tasks": [
|
|
79
|
+
{
|
|
80
|
+
"id": "capital-exact",
|
|
81
|
+
"prompt": "Reply with exactly the single word Paris.",
|
|
82
|
+
"evaluator": {
|
|
83
|
+
"type": "exact",
|
|
84
|
+
"expected": "Paris",
|
|
85
|
+
"strip": true,
|
|
86
|
+
"ignore_case": false
|
|
87
|
+
},
|
|
88
|
+
"metadata": {
|
|
89
|
+
"category": "exact-match"
|
|
90
|
+
}
|
|
91
|
+
},
|
|
92
|
+
{
|
|
93
|
+
"id": "keywords-contained",
|
|
94
|
+
"prompt": "Write one short sentence containing both alpha and beta.",
|
|
95
|
+
"evaluator": {
|
|
96
|
+
"type": "contains",
|
|
97
|
+
"expected": [
|
|
98
|
+
"alpha",
|
|
99
|
+
"beta"
|
|
100
|
+
],
|
|
101
|
+
"mode": "all",
|
|
102
|
+
"ignore_case": true
|
|
103
|
+
},
|
|
104
|
+
"metadata": {
|
|
105
|
+
"category": "contains"
|
|
106
|
+
}
|
|
107
|
+
},
|
|
108
|
+
{
|
|
109
|
+
"id": "four-digit-regex",
|
|
110
|
+
"prompt": "Return only a four-digit positive integer.",
|
|
111
|
+
"evaluator": {
|
|
112
|
+
"type": "regex",
|
|
113
|
+
"pattern": "^[0-9]{4}$",
|
|
114
|
+
"ignore_case": false,
|
|
115
|
+
"full_match": true
|
|
116
|
+
},
|
|
117
|
+
"metadata": {
|
|
118
|
+
"category": "format"
|
|
119
|
+
}
|
|
120
|
+
},
|
|
121
|
+
{
|
|
122
|
+
"id": "json-object",
|
|
123
|
+
"messages": [
|
|
124
|
+
{
|
|
125
|
+
"role": "system",
|
|
126
|
+
"content": "Return only valid JSON with no surrounding prose."
|
|
127
|
+
},
|
|
128
|
+
{
|
|
129
|
+
"role": "user",
|
|
130
|
+
"content": "Return an object whose answer field is the number 42."
|
|
131
|
+
}
|
|
132
|
+
],
|
|
133
|
+
"evaluator": {
|
|
134
|
+
"type": "json",
|
|
135
|
+
"compare": true,
|
|
136
|
+
"expected": {
|
|
137
|
+
"answer": 42
|
|
138
|
+
}
|
|
139
|
+
},
|
|
140
|
+
"metadata": {
|
|
141
|
+
"category": "structured-output"
|
|
142
|
+
}
|
|
143
|
+
},
|
|
144
|
+
{
|
|
145
|
+
"id": "numeric-answer",
|
|
146
|
+
"prompt": "Return only the result of 6 multiplied by 7.",
|
|
147
|
+
"evaluator": {
|
|
148
|
+
"type": "numeric",
|
|
149
|
+
"expected": 42,
|
|
150
|
+
"tolerance": 0
|
|
151
|
+
},
|
|
152
|
+
"metadata": {
|
|
153
|
+
"category": "numeric"
|
|
154
|
+
}
|
|
155
|
+
}
|
|
156
|
+
]
|
|
157
|
+
}
|