@sanity/ailf 0.1.21 → 0.1.23
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.
|
@@ -565,8 +565,30 @@ export interface RetrievalMetrics {
|
|
|
565
565
|
avgRecall: number;
|
|
566
566
|
};
|
|
567
567
|
}
|
|
568
|
+
/** Per-feature agent behavior data — how agents interacted with docs */
|
|
569
|
+
export interface FeatureAgentBehavior {
|
|
570
|
+
avgDocPagesVisited: number;
|
|
571
|
+
avgNetworkTimeMs: number;
|
|
572
|
+
avgSearchesPerformed: number;
|
|
573
|
+
docSlugsVisited: string[];
|
|
574
|
+
externalDomains: string[];
|
|
575
|
+
feature: string;
|
|
576
|
+
searchQueries: string[];
|
|
577
|
+
tasksWithBehaviorData: number;
|
|
578
|
+
}
|
|
579
|
+
/** Aggregate agent behavior stats across all features */
|
|
580
|
+
export interface OverallAgentBehavior {
|
|
581
|
+
avgDocPagesVisited: number;
|
|
582
|
+
avgNetworkTimeMs: number;
|
|
583
|
+
avgSearchesPerformed: number;
|
|
584
|
+
testsWithBehaviorData: number;
|
|
585
|
+
totalUniqueDocSlugs: number;
|
|
586
|
+
totalUniqueSearchQueries: number;
|
|
587
|
+
}
|
|
568
588
|
/** Top-level score summary (the shape of score-summary.json) */
|
|
569
589
|
export interface ScoreSummary {
|
|
590
|
+
/** Per-feature agent behavior data (only present when agentic mode ran) */
|
|
591
|
+
agentBehavior?: FeatureAgentBehavior[];
|
|
570
592
|
belowCritical: string[];
|
|
571
593
|
/**
|
|
572
594
|
* All Sanity documents used across the entire evaluation.
|
|
@@ -604,6 +626,8 @@ export interface ScoreSummary {
|
|
|
604
626
|
docLift: number;
|
|
605
627
|
}[];
|
|
606
628
|
overall: {
|
|
629
|
+
/** Aggregate agent behavior stats (only present when agentic mode ran) */
|
|
630
|
+
agentBehavior?: OverallAgentBehavior;
|
|
607
631
|
/** Average actual (agentic) score across areas. Absent if no agentic data. */
|
|
608
632
|
avgActualScore?: number;
|
|
609
633
|
/** Average ceiling score across all areas */
|
|
@@ -99,6 +99,8 @@ export async function buildRemoteRequest(options) {
|
|
|
99
99
|
raw.readiness = true;
|
|
100
100
|
if (config.discoveryReportEnabled)
|
|
101
101
|
raw.discoveryReport = true;
|
|
102
|
+
if (config.noRemoteCache)
|
|
103
|
+
raw.noRemoteCache = true;
|
|
102
104
|
// 4. Validate the assembled request
|
|
103
105
|
const parsed = PipelineRequestSchema.parse(raw);
|
|
104
106
|
return { request: parsed, taskCount: tasks.length };
|