@sanity/ailf-studio 0.2.0 → 0.3.0

package/dist/index.d.ts

@@ -278,6 +278,23 @@ declare const GLOSSARY: {
     readonly dimTaskCompletion: "Change in task completion between runs. Positive means implementations are more complete.";
     readonly dimCodeCorrectness: "Change in code correctness between runs. Positive means better code quality.";
     readonly dimDocCoverage: "Change in doc coverage between runs. Positive means the docs are providing more useful information.";
+    readonly areaDelta: "Score change for this area compared to the previous evaluation run.";
+    readonly sourceProduction: "Production source — docs fetched from the live production dataset. Scores reflect what real users and AI agents experience today.";
+    readonly sourceBranch: "Branch source — docs fetched from a branch or draft dataset. Use this to preview how content changes affect scores before publishing.";
+    readonly sourceLocal: "Local source — docs fetched from local files or a local dev server. Useful for testing doc changes before pushing.";
+    readonly reportScore: "The overall weighted score for this evaluation run: Task Completion (50%), Code Correctness (25%), and Doc Coverage (25%), averaged across all feature areas.";
+    readonly reportMode: "The evaluation mode determines which reference points are measured. Different modes test different aspects of how AI agents interact with documentation.";
+    readonly reportTrigger: "What initiated this evaluation run. Knowing the trigger helps you understand whether a score change was from a content edit, a code deploy, or a scheduled check.";
+    readonly modeBaseline: "Baseline mode — tests the model with gold-standard docs injected directly. Measures ceiling performance (best the docs can do).";
+    readonly modeFull: "Full mode — runs baseline + agentic. Compares ceiling (injected docs) against actual (agent-retrieved docs) to measure retrieval gap and infrastructure efficiency.";
+    readonly modeAgentic: "Agentic mode — the AI agent finds docs on its own via web search. Measures real-world performance: can agents actually discover and use your documentation?";
+    readonly modeObserved: "Observed mode — records how agents interact with docs without scoring. Captures search queries, pages visited, and browsing patterns for analysis.";
+    readonly modeDebug: "Debug mode — a diagnostic run for pipeline development. May use non-standard configurations or limited task sets.";
+    readonly triggerManual: "Manually triggered — someone ran the evaluation pipeline by hand, either locally or via the Studio UI.";
+    readonly triggerCi: "CI-triggered — the evaluation ran automatically as part of a pull request or merge pipeline.";
+    readonly triggerSchedule: "Scheduled — the evaluation ran on a recurring schedule (e.g. nightly or weekly) to track score trends over time.";
+    readonly triggerWebhook: "Webhook-triggered — a content change in Sanity triggered the evaluation automatically. Helps catch doc regressions early.";
+    readonly triggerCrossRepo: "Cross-repo — triggered from another repository via the dispatch API. Used when external repos want to validate their docs against AILF tasks.";
 };
 
 /**
@@ -581,6 +598,8 @@ interface ComparisonData {
         docLift: number;
         overall: number;
         actualDelta?: number;
+        /** Per-area score deltas — e.g. `{ "GROQ": 5.2, "Functions": -2.1 }` */
+        perArea?: Record<string, number>;
         retrievalGapDelta?: number;
         infrastructureEfficiencyDelta?: number;
     };