@ggui-ai/shared 0.12.0 → 0.14.0
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.
|
@@ -88,13 +88,70 @@ export interface BenchmarkMeta {
|
|
|
88
88
|
* degraded run rather than presenting them as full-coverage means.
|
|
89
89
|
*/
|
|
90
90
|
judgeCoverageDegraded?: true;
|
|
91
|
+
/**
|
|
92
|
+
* Per-criterion coverage of the in-loop evaluator over the run's
|
|
93
|
+
* tier-evaluated cells. A criterion that never ran fail-opens into
|
|
94
|
+
* `pass` inside the evaluator, so a pass with skipped coverage is zero
|
|
95
|
+
* evidence — this is the disclosure. `unknown` = cells whose eval
|
|
96
|
+
* predates the instrument (or lacks that criterion's row). Cells whose
|
|
97
|
+
* evaluator was bypassed by design (all rows `not-applicable`) sit
|
|
98
|
+
* outside the coverage denominator but are counted as `notApplicable`.
|
|
99
|
+
* Absent when no cell carries the instrument — an all-bypass run is
|
|
100
|
+
* PRESENT (nothing applicable), which is not the same as pre-instrument.
|
|
101
|
+
*/
|
|
102
|
+
criteriaCoverage?: CriterionCoverageSummaryDisplay[];
|
|
103
|
+
/**
|
|
104
|
+
* Present (true) when some criterion ran on fewer than the runner's
|
|
105
|
+
* floor of tier-evaluated cells — its pass/warn/fail counts are not
|
|
106
|
+
* representative. Viewers must annotate.
|
|
107
|
+
*/
|
|
108
|
+
criteriaCoverageDegraded?: true;
|
|
91
109
|
/** SPDX id of the published dataset license. */
|
|
92
110
|
dataLicense?: string;
|
|
93
111
|
}
|
|
112
|
+
/**
|
|
113
|
+
* Run-level roll-up of one eval criterion's coverage. `ran + skipped +
|
|
114
|
+
* unknown` = the cells the criterion applied to (the denominator).
|
|
115
|
+
*/
|
|
116
|
+
export interface CriterionCoverageSummaryDisplay {
|
|
117
|
+
criterion: string;
|
|
118
|
+
tier: 1 | 2;
|
|
119
|
+
ran: number;
|
|
120
|
+
skipped: number;
|
|
121
|
+
unknown: number;
|
|
122
|
+
/**
|
|
123
|
+
* Cells the evaluator bypassed by design for this criterion — outside
|
|
124
|
+
* the denominator, counted so "nothing applicable" is visible. Absent
|
|
125
|
+
* on reports published before 2026-09-03 (read as 0).
|
|
126
|
+
*/
|
|
127
|
+
notApplicable?: number;
|
|
128
|
+
}
|
|
129
|
+
/**
|
|
130
|
+
* One cell's coverage row for one eval criterion (mirrors ui-gen's
|
|
131
|
+
* `CriterionCoverage`). `not-applicable` = the evaluator was bypassed by
|
|
132
|
+
* design for this cell (same-image low-risk); `reason` says why.
|
|
133
|
+
*/
|
|
134
|
+
export interface CriterionCoverageDisplay {
|
|
135
|
+
criterion: string;
|
|
136
|
+
tier: 1 | 2;
|
|
137
|
+
status: 'ran' | 'skipped' | 'not-applicable';
|
|
138
|
+
reason?: string;
|
|
139
|
+
}
|
|
94
140
|
/** Which model + prompt version produced the quality scores. */
|
|
95
141
|
export interface JudgeDisclosureDisplay {
|
|
96
142
|
model: string;
|
|
97
143
|
promptVersion: string;
|
|
144
|
+
/**
|
|
145
|
+
* Sampling the judge ACTUALLY ran with, as reported by the harness router
|
|
146
|
+
* (the panel requests temperature 0; families that reject sampling
|
|
147
|
+
* params get `'provider-default'` plus the reason). Absent on reports
|
|
148
|
+
* published before 2026-09-02 or when the router reported nothing —
|
|
149
|
+
* unknown, never assumed.
|
|
150
|
+
*/
|
|
151
|
+
sampling?: {
|
|
152
|
+
temperature: number | 'provider-default';
|
|
153
|
+
strippedReason?: string;
|
|
154
|
+
};
|
|
98
155
|
}
|
|
99
156
|
export interface VariantInfo {
|
|
100
157
|
id: string;
|
|
@@ -165,6 +222,12 @@ export interface TierEvaluationDisplay {
|
|
|
165
222
|
description: string;
|
|
166
223
|
}>;
|
|
167
224
|
pass: string[];
|
|
225
|
+
/**
|
|
226
|
+
* Per-criterion ran/skipped rows for this cell's eval. Absent when the
|
|
227
|
+
* eval predates the instrument — coverage unknown, never assumed ran.
|
|
228
|
+
* A criterion in `pass` with status `skipped` carried no verdict.
|
|
229
|
+
*/
|
|
230
|
+
criteriaCoverage?: CriterionCoverageDisplay[];
|
|
168
231
|
}
|
|
169
232
|
export interface PostGenerationDisplay {
|
|
170
233
|
hasStreamSpec: boolean;
|
|
@@ -1 +1 @@
|
|
|
1
|
-
{"version":3,"file":"benchmark.d.ts","sourceRoot":"","sources":["../../src/types/benchmark.ts"],"names":[],"mappings":"AAAA;;;;GAIG;AAEH,MAAM,WAAW,sBAAsB;IACrC,IAAI,EAAE,aAAa,CAAC;IACpB,OAAO,EAAE,yBAAyB,EAAE,CAAC;IACrC,gBAAgB,EAAE,qBAAqB,EAAE,CAAC;IAC1C,eAAe,EAAE,oBAAoB,EAAE,CAAC;IACxC,aAAa,EAAE,MAAM,CAAC,MAAM,EAAE,MAAM,CAAC,MAAM,EAAE,kBAAkB,CAAC,CAAC,CAAC;IAClE;;;;;;;;OAQG;IACH,WAAW,CAAC,EAAE,MAAM,CAClB,MAAM,EACN,MAAM,CAAC,MAAM,EAAE,MAAM,CAAC,MAAM,EAAE,8BAA8B,CAAC,CAAC,CAC/D,CAAC;IACF;;;;OAIG;IACH,kBAAkB,CAAC,EAAE,uBAAuB,EAAE,CAAC;CAChD;AAED;;;;GAIG;AACH,MAAM,WAAW,8BAA8B;IAC7C,SAAS,EAAE,MAAM,CAAC;IAClB,QAAQ,EAAE,MAAM,CAAC;IACjB,OAAO,EAAE,MAAM,CAAC;IAChB,IAAI,EAAE,MAAM,CAAC;IACb,QAAQ,EAAE,MAAM,CAAC;IACjB,SAAS,EAAE,MAAM,CAAC;IAClB,UAAU,EAAE,MAAM,CAAC;IACnB,WAAW,EAAE,MAAM,CAAC;CACrB;AAED;;;GAGG;AACH,MAAM,WAAW,uBAAuB;IACtC,SAAS,EAAE,MAAM,CAAC;IAClB,IAAI,EAAE,MAAM,CAAC;IACb,QAAQ,EAAE,MAAM,CAAC;IACjB,SAAS,EAAE,MAAM,CAAC;IAClB,UAAU,EAAE,MAAM,CAAC;IACnB,WAAW,EAAE,MAAM,CAAC;CACrB;AAED,MAAM,WAAW,aAAa;IAC5B,QAAQ,EAAE,MAAM,CAAC;IACjB,SAAS,EAAE,MAAM,CAAC;IAClB,OAAO,EAAE,MAAM,CAAC;IAChB,aAAa,EAAE,MAAM,CAAC;IACtB,YAAY,EAAE,MAAM,CAAC;IACrB,SAAS,EAAE,MAAM,CAAC;IAClB,YAAY,EAAE,MAAM,CAAC;IACrB,YAAY,EAAE,MAAM,CAAC;IACrB,WAAW,EAAE,MAAM,CAAC;IACpB,UAAU,EAAE,MAAM,CAAC;IACnB;;;;OAIG;IACH,MAAM,CAAC,EAAE,sBAAsB,EAAE,CAAC;IAClC;;;OAGG;IACH,cAAc,CAAC,EAAE,MAAM,CAAC;IACxB;;;;OAIG;IACH,aAAa,CAAC,EAAE,MAAM,CAAC;IACvB;;;;;OAKG;IACH,qBAAqB,CAAC,EAAE,IAAI,CAAC;IAC7B,gDAAgD;IAChD,WAAW,CAAC,EAAE,MAAM,CAAC;CACtB;AAED,gEAAgE;AAChE,MAAM,WAAW,sBAAsB;IACrC,KAAK,EAAE,MAAM,CAAC;IACd,aAAa,EAAE,MAAM,CAAC;
|
|
1
|
+
{"version":3,"file":"benchmark.d.ts","sourceRoot":"","sources":["../../src/types/benchmark.ts"],"names":[],"mappings":"AAAA;;;;GAIG;AAEH,MAAM,WAAW,sBAAsB;IACrC,IAAI,EAAE,aAAa,CAAC;IACpB,OAAO,EAAE,yBAAyB,EAAE,CAAC;IACrC,gBAAgB,EAAE,qBAAqB,EAAE,CAAC;IAC1C,eAAe,EAAE,oBAAoB,EAAE,CAAC;IACxC,aAAa,EAAE,MAAM,CAAC,MAAM,EAAE,MAAM,CAAC,MAAM,EAAE,kBAAkB,CAAC,CAAC,CAAC;IAClE;;;;;;;;OAQG;IACH,WAAW,CAAC,EAAE,MAAM,CAClB,MAAM,EACN,MAAM,CAAC,MAAM,EAAE,MAAM,CAAC,MAAM,EAAE,8BAA8B,CAAC,CAAC,CAC/D,CAAC;IACF;;;;OAIG;IACH,kBAAkB,CAAC,EAAE,uBAAuB,EAAE,CAAC;CAChD;AAED;;;;GAIG;AACH,MAAM,WAAW,8BAA8B;IAC7C,SAAS,EAAE,MAAM,CAAC;IAClB,QAAQ,EAAE,MAAM,CAAC;IACjB,OAAO,EAAE,MAAM,CAAC;IAChB,IAAI,EAAE,MAAM,CAAC;IACb,QAAQ,EAAE,MAAM,CAAC;IACjB,SAAS,EAAE,MAAM,CAAC;IAClB,UAAU,EAAE,MAAM,CAAC;IACnB,WAAW,EAAE,MAAM,CAAC;CACrB;AAED;;;GAGG;AACH,MAAM,WAAW,uBAAuB;IACtC,SAAS,EAAE,MAAM,CAAC;IAClB,IAAI,EAAE,MAAM,CAAC;IACb,QAAQ,EAAE,MAAM,CAAC;IACjB,SAAS,EAAE,MAAM,CAAC;IAClB,UAAU,EAAE,MAAM,CAAC;IACnB,WAAW,EAAE,MAAM,CAAC;CACrB;AAED,MAAM,WAAW,aAAa;IAC5B,QAAQ,EAAE,MAAM,CAAC;IACjB,SAAS,EAAE,MAAM,CAAC;IAClB,OAAO,EAAE,MAAM,CAAC;IAChB,aAAa,EAAE,MAAM,CAAC;IACtB,YAAY,EAAE,MAAM,CAAC;IACrB,SAAS,EAAE,MAAM,CAAC;IAClB,YAAY,EAAE,MAAM,CAAC;IACrB,YAAY,EAAE,MAAM,CAAC;IACrB,WAAW,EAAE,MAAM,CAAC;IACpB,UAAU,EAAE,MAAM,CAAC;IACnB;;;;OAIG;IACH,MAAM,CAAC,EAAE,sBAAsB,EAAE,CAAC;IAClC;;;OAGG;IACH,cAAc,CAAC,EAAE,MAAM,CAAC;IACxB;;;;OAIG;IACH,aAAa,CAAC,EAAE,MAAM,CAAC;IACvB;;;;;OAKG;IACH,qBAAqB,CAAC,EAAE,IAAI,CAAC;IAC7B;;;;;;;;;;OAUG;IACH,gBAAgB,CAAC,EAAE,+BAA+B,EAAE,CAAC;IACrD;;;;OAIG;IACH,wBAAwB,CAAC,EAAE,IAAI,CAAC;IAChC,gDAAgD;IAChD,WAAW,CAAC,EAAE,MAAM,CAAC;CACtB;AAED;;;GAGG;AACH,MAAM,WAAW,+BAA+B;IAC9C,SAAS,EAAE,MAAM,CAAC;IAClB,IAAI,EAAE,CAAC,GAAG,CAAC,CAAC;IACZ,GAAG,EAAE,MAAM,CAAC;IACZ,OAAO,EAAE,MAAM,CAAC;IAChB,OAAO,EAAE,MAAM,CAAC;IAChB;;;;OAIG;IACH,aAAa,CAAC,EAAE,MAAM,CAAC;CACxB;AAED;;;;GAIG;AACH,MAAM,WAAW,wBAAwB;IACvC,SAAS,EAAE,MAAM,CAAC;IAClB,IAAI,EAAE,CAAC,GAAG,CAAC,CAAC;IACZ,MAAM,EAAE,KAAK,GAAG,SAAS,GAAG,gBAAgB,CAAC;IAC7C,MAAM,CAAC,EAAE,MAAM,CAAC;CACjB;AAED,gEAAgE;AAChE,MAAM,WAAW,sBAAsB;IACrC,KAAK,EAAE,MAAM,CAAC;IACd,aAAa,EAAE,MAAM,CAAC;IACtB;;;;;;OAMG;IACH,QAAQ,CAAC,EAAE;QAAE,WAAW,EAAE,MAAM,GAAG,kBAAkB,CAAC;QAAC,cAAc,CAAC,EAAE,MAAM,CAAA;KAAE,CAAC;CAClF;AAED,MAAM,WAAW,WAAW;IAC1B,EAAE,EAAE,MAAM,CAAC;IACX,OAAO,EAAE,MAAM,CAAC;IAChB,IAAI,EAAE,MAAM,CAAC;IACb,OAAO,CAAC,EAAE,MAAM,CAAC;CAClB;AAED,MAAM,WAAW,UAAU;IACzB,EAAE,EAAE,MAAM,CAAC;IACX,IAAI,EAAE,MAAM,CAAC;IACb,UAAU,EAAE,MAAM,CAAC;CACpB;AAED,MAAM,WAAW,uBAAuB;IACtC,gBAAgB,EAAE,MAAM,CAAC;IACzB,SAAS,EAAE,MAAM,CAAC;IAClB,MAAM,EAAE;QAAE,KAAK,EAAE,MAAM,CAAC;QAAC,MAAM,EAAE,MAAM,CAAC;QAAC,KAAK,EAAE,MAAM,CAAA;KAAE,CAAC;IACzD,UAAU,CAAC,EAAE,MAAM,CAAC;CACrB;AAED;;;;;GAKG;AACH,MAAM,WAAW,2BAA2B;IAC1C,MAAM,EAAE,MAAM,CAAC;IACf,YAAY,EAAE,MAAM,CAAC;IACrB,SAAS,EAAE,MAAM,CAAC;IAClB,MAAM,EAAE,MAAM,CAAC;IACf,gBAAgB,EAAE,MAAM,CAAC;CAC1B;AAED;;;;;GAKG;AACH,MAAM,WAAW,0BAA0B;IACzC,KAAK,EAAE,sBAAsB,CAAC;IAC9B,KAAK,EAAE,MAAM,CAAC;IACd,UAAU,EAAE,2BAA2B,CAAC;IACxC,QAAQ,CAAC,EAAE,MAAM,CAAC;CACnB;AAED,MAAM,WAAW,uBAAuB;IACtC,MAAM,EAAE,OAAO,CAAC;IAChB,4DAA4D;IAC5D,KAAK,EAAE,MAAM,CAAC;IACd,2CAA2C;IAC3C,UAAU,EAAE,2BAA2B,CAAC;IACxC,yEAAyE;IACzE,MAAM,EAAE,sBAAsB,EAAE,CAAC;IACjC,gEAAgE;IAChE,KAAK,EAAE,0BAA0B,EAAE,CAAC;IACpC,8EAA8E;IAC9E,MAAM,EAAE,MAAM,CAAC;IACf,QAAQ,CAAC,EAAE,MAAM,CAAC;IAClB,UAAU,CAAC,EAAE,MAAM,CAAC;CACrB;AAED,MAAM,WAAW,qBAAqB;IACpC,MAAM,EAAE,KAAK,CAAC;QACZ,IAAI,EAAE,MAAM,CAAC;QACb,MAAM,EAAE,MAAM,CAAC;QACf,QAAQ,EAAE,MAAM,CAAC;QACjB,WAAW,EAAE,MAAM,CAAC;KACrB,CAAC,CAAC;IACH,IAAI,EAAE,MAAM,EAAE,CAAC;IACf;;;;OAIG;IACH,gBAAgB,CAAC,EAAE,wBAAwB,EAAE,CAAC;CAC/C;AAED,MAAM,WAAW,qBAAqB;IACpC,aAAa,EAAE,OAAO,CAAC;IACvB,gBAAgB,EAAE,OAAO,CAAC;IAC1B,iBAAiB,EAAE,MAAM,CAAC;IAC1B,eAAe,CAAC,EAAE,MAAM,CAAC;IACzB,aAAa,CAAC,EAAE;QAAE,UAAU,EAAE,OAAO,CAAC;QAAC,UAAU,EAAE,MAAM,EAAE,CAAA;KAAE,CAAC;IAC9D;;;;;;;OAOG;IACH,WAAW,CAAC,EAAE;QACZ,QAAQ,EAAE,SAAS,MAAM,EAAE,CAAC;QAC5B,IAAI,EAAE,SAAS,MAAM,EAAE,CAAC;QACxB,OAAO,EAAE,SAAS,MAAM,EAAE,CAAC;KAC5B,CAAC;CACH;AAED,MAAM,WAAW,yBAAyB;IACxC,OAAO,EAAE,WAAW,CAAC;IACrB,MAAM,EAAE,UAAU,CAAC;IACnB,UAAU,EAAE,uBAAuB,GAAG,IAAI,CAAC;IAC3C,UAAU,EAAE,uBAAuB,GAAG,IAAI,CAAC;IAC3C,cAAc,CAAC,EAAE,qBAAqB,CAAC;IACvC,gBAAgB,EAAE,MAAM,CAAC;IACzB,KAAK,CAAC,EAAE,MAAM,CAAC;IACf,SAAS,EAAE,MAAM,CAAC;IAClB,cAAc,CAAC,EAAE,qBAAqB,CAAC;CACxC;AAED,MAAM,WAAW,qBAAqB;IACpC,SAAS,EAAE,MAAM,CAAC;IAClB,OAAO,EAAE,MAAM,CAAC;IAChB,IAAI,EAAE,MAAM,CAAC;IACb,OAAO,EAAE,MAAM,CAAC;IAChB,QAAQ,EAAE,MAAM,CAAC;IACjB,SAAS,EAAE,MAAM,CAAC;IAClB,UAAU,EAAE,MAAM,CAAC;IACnB;;;OAGG;IACH,WAAW,EAAE,MAAM,CAAC;IACpB,SAAS,EAAE,MAAM,CAAC;CACnB;AAED,MAAM,WAAW,oBAAoB;IACnC,QAAQ,EAAE,MAAM,CAAC;IACjB,IAAI,EAAE,MAAM,CAAC;IACb,UAAU,EAAE,MAAM,CAAC;IACnB,aAAa,EAAE,MAAM,CAAC;IACtB,cAAc,EAAE,MAAM,CAAC;IACvB,QAAQ,EAAE,MAAM,CAAC;CAClB;AAED,MAAM,WAAW,kBAAkB;IACjC,QAAQ,EAAE,MAAM,CAAC;IACjB,SAAS,EAAE,MAAM,CAAC;IAClB,UAAU,EAAE,MAAM,CAAC;IACnB,WAAW,EAAE,MAAM,CAAC;CACrB"}
|
package/package.json
CHANGED
|
@@ -1,6 +1,6 @@
|
|
|
1
1
|
{
|
|
2
2
|
"name": "@ggui-ai/shared",
|
|
3
|
-
"version": "0.
|
|
3
|
+
"version": "0.14.0",
|
|
4
4
|
"description": "Shared types for ggui SDK packages — self-repair, agent listings, benchmark report display",
|
|
5
5
|
"license": "Apache-2.0",
|
|
6
6
|
"keywords": [
|
|
@@ -37,7 +37,7 @@
|
|
|
37
37
|
"access": "public"
|
|
38
38
|
},
|
|
39
39
|
"engines": {
|
|
40
|
-
"node": ">=
|
|
40
|
+
"node": ">=22.0.0"
|
|
41
41
|
},
|
|
42
42
|
"author": "ggui contributors <hello@ggui.ai>",
|
|
43
43
|
"scripts": {
|