@axlsdk/studio 0.23.3 → 0.24.0

package/dist/server/index.d.cts

@@ -10,11 +10,28 @@ import { AxlRuntime, HistoricalAxlEvent, HistoricalExecutionInfo, EvalHistoryEnt
  * Each reducer is a pure (state, source) => state function — no I/O, no mutation.
  */
 
+/**
+ * How complete a trend point's spend figure is. Mirrors `@axlsdk/axl`'s
+ * `AccountingCompleteness` — restated here rather than imported so the
+ * aggregate reducers stay free of an optional peer dependency.
+ */
+type EvalTrendCompleteness = 'complete' | 'incomplete' | 'unverified';
 type EvalTrendRun = {
     timestamp: number;
     id: string;
     scores: Record<string, number>;
+    /** Known spend for this run — `accounting.knownCost`, or the legacy
+     *  `totalCost` when the artifact predates accounting. Read it WITH
+     *  `completeness`; alone it claims a precision the run may not have. */
     cost: number;
+    /**
+     * Whether that figure is measured (`complete`), a lower bound
+     * (`incomplete`), or repeated from an artifact that carried no accounting
+     * (`unverified`). Three enum values per run, so the payload stays bounded.
+     */
+    completeness: EvalTrendCompleteness;
+    /** `true` when this run's budget closed — it is short by design, not broken. */
+    budgetStopped?: boolean;
     /** Primary model for this run (first entry of `metadata.models`). Undefined
      *  when the run has no recorded models (e.g., legacy data or test harnesses). */
     model?: string;
@@ -40,12 +57,23 @@ type EvalTrendEntry = {
     scoreMean: Record<string, number>;
     scoreStd: Record<string, number>;
     costTotal: number;
+    /**
+     * The WORST completeness of every run folded into `costTotal` — including
+     * runs already evicted by the `MAX_EVAL_RUNS` cap, since `costTotal` counts
+     * them too. A window holding one legacy run is `'unverified'` as a whole.
+     */
+    costCompleteness: EvalTrendCompleteness;
+    /** Runs in this window whose budget closed. */
+    budgetStoppedRuns: number;
     runCount: number;
 };
 type EvalTrendData = {
     byEval: Record<string, EvalTrendEntry>;
     totalRuns: number;
     totalCost: number;
+    /** Worst completeness across every eval in the window. `totalCost` is never
+     *  presented as certified without it. */
+    totalCostCompleteness: EvalTrendCompleteness;
 };
 type WorkflowStatsData = {
     byWorkflow: Record<string, {

package/dist/server/index.d.ts

@@ -10,11 +10,28 @@ import { AxlRuntime, HistoricalAxlEvent, HistoricalExecutionInfo, EvalHistoryEnt
  * Each reducer is a pure (state, source) => state function — no I/O, no mutation.
  */
 
+/**
+ * How complete a trend point's spend figure is. Mirrors `@axlsdk/axl`'s
+ * `AccountingCompleteness` — restated here rather than imported so the
+ * aggregate reducers stay free of an optional peer dependency.
+ */
+type EvalTrendCompleteness = 'complete' | 'incomplete' | 'unverified';
 type EvalTrendRun = {
     timestamp: number;
     id: string;
     scores: Record<string, number>;
+    /** Known spend for this run — `accounting.knownCost`, or the legacy
+     *  `totalCost` when the artifact predates accounting. Read it WITH
+     *  `completeness`; alone it claims a precision the run may not have. */
     cost: number;
+    /**
+     * Whether that figure is measured (`complete`), a lower bound
+     * (`incomplete`), or repeated from an artifact that carried no accounting
+     * (`unverified`). Three enum values per run, so the payload stays bounded.
+     */
+    completeness: EvalTrendCompleteness;
+    /** `true` when this run's budget closed — it is short by design, not broken. */
+    budgetStopped?: boolean;
     /** Primary model for this run (first entry of `metadata.models`). Undefined
      *  when the run has no recorded models (e.g., legacy data or test harnesses). */
     model?: string;
@@ -40,12 +57,23 @@ type EvalTrendEntry = {
     scoreMean: Record<string, number>;
     scoreStd: Record<string, number>;
     costTotal: number;
+    /**
+     * The WORST completeness of every run folded into `costTotal` — including
+     * runs already evicted by the `MAX_EVAL_RUNS` cap, since `costTotal` counts
+     * them too. A window holding one legacy run is `'unverified'` as a whole.
+     */
+    costCompleteness: EvalTrendCompleteness;
+    /** Runs in this window whose budget closed. */
+    budgetStoppedRuns: number;
     runCount: number;
 };
 type EvalTrendData = {
     byEval: Record<string, EvalTrendEntry>;
     totalRuns: number;
     totalCost: number;
+    /** Worst completeness across every eval in the window. `totalCost` is never
+     *  presented as certified without it. */
+    totalCostCompleteness: EvalTrendCompleteness;
 };
 type WorkflowStatsData = {
     byWorkflow: Record<string, {

package/dist/server/index.js

@@ -4,7 +4,7 @@ import {
   ExecutionAggregator,
   TraceAggregator,
   createServer
-} from "../chunk-XK2DRXKB.js";
+} from "../chunk-PFPLZQ5X.js";
 export {
   ConnectionManager,
   EvalAggregator,

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@axlsdk/studio",
-  "version": "0.23.3",
+  "version": "0.24.0",
   "description": "Local development UI for debugging, testing, and iterating on Axl agents and workflows",
   "type": "module",
   "main": "./dist/server/index.cjs",
@@ -40,7 +40,7 @@
     "hono": "^4.7.0",
     "tsx": "^4.19.0",
     "ws": "^8.0.0",
-    "@axlsdk/axl": "0.23.3"
+    "@axlsdk/axl": "0.24.0"
   },
   "devDependencies": {
     "@tailwindcss/vite": "^4.1.0",
@@ -65,10 +65,10 @@
     "vite": "^6.0.0",
     "vitest": "^3.0.0",
     "zod": "^4.0.0",
-    "@axlsdk/testing": "0.23.3"
+    "@axlsdk/testing": "0.24.0"
   },
   "peerDependencies": {
-    "@axlsdk/eval": "0.23.3"
+    "@axlsdk/eval": "0.24.0"
   },
   "peerDependenciesMeta": {
     "@axlsdk/eval": {