sheawinkler · sheawinkler · May 30, 2026 · May 30, 2026
@@ -27,7 +27,7 @@ DC := docker compose -f docker-compose.yml
 PYTEST_FOCUS ?= app
 PYTEST_APP_TESTS := archive/services/orchestrator_legacy_python/tests/test_orchestrator_retrieval.py archive/services/orchestrator_legacy_python/tests/test_migration_runtime.py
 
-.PHONY: help launch all up up-core down status ps logs build rebuild pull clean prune             mcp-proxy-up init qdrant-init mindsdb-seed letta-seed models-pull             proxy-status doctor mem-ping monitor-open monitor-check dmg-build msi-build linux-bundle-build            storage-audit qdrant-snapshot-prune qdrant-cutover cold-snapshot-pack cold-snapshot-tier cold-snapshot-restore telemetry-archive fanout-status fanout-deadletters fanout-rehydrate retention-install retention-uninstall retention-status retention-install-daily storage-ledger-capture storage-ledger-prune storage-ledger-install storage-ledger-uninstall storage-ledger-status memory-graph-quality memory-graph-quality-install memory-graph-quality-uninstall memory-graph-quality-status weekly-lineage-rollup weekly-lineage-install weekly-lineage-uninstall weekly-lineage-status            docker-fs-watchdog-run docker-fs-watchdog-install docker-fs-watchdog-uninstall docker-fs-watchdog-status            storage-migrate-hot-bindings disk-clean-safe            mem-mode-show mem-mode-core mem-mode-balanced mem-mode-full mem-up-core mem-up-balanced mem-up-full observability-up observability-down launch-readiness-gate launch-readiness-gate-schedule launch-readiness-gate-schedule-status launch-readiness-gate-schedule-cancel paid-launch-checklist backup-restore-drill mem-up-release mem-up-lite-release release-lock-verify qdrant-cloud-check quickstart submission-preflight launch-lock launch-lock-public test-py bench-shortlist bench-qdrant-tuning bench-backend-lanes env-lock-check env-lock-apply sentrux-check sentrux-gate sentrux-gate-save agent-context-gate
+.PHONY: help launch all up up-core down status ps logs build rebuild pull clean prune             mcp-proxy-up init qdrant-init mindsdb-seed letta-seed models-pull             proxy-status doctor mem-ping monitor-open monitor-check dmg-build msi-build linux-bundle-build            storage-audit qdrant-snapshot-prune qdrant-cutover cold-snapshot-pack cold-snapshot-tier cold-snapshot-restore telemetry-archive fanout-status fanout-deadletters fanout-rehydrate retention-install retention-uninstall retention-status retention-install-daily storage-ledger-capture storage-ledger-prune storage-ledger-install storage-ledger-uninstall storage-ledger-status memory-graph-quality memory-graph-quality-install memory-graph-quality-uninstall memory-graph-quality-status recall-quality recall-quality-refresh recall-quality-tuning open-core-boundary-audit weekly-lineage-rollup weekly-lineage-install weekly-lineage-uninstall weekly-lineage-status            docker-fs-watchdog-run docker-fs-watchdog-install docker-fs-watchdog-uninstall docker-fs-watchdog-status            storage-migrate-hot-bindings disk-clean-safe            mem-mode-show mem-mode-core mem-mode-balanced mem-mode-full mem-up-core mem-up-balanced mem-up-full observability-up observability-down launch-readiness-gate launch-readiness-gate-schedule launch-readiness-gate-schedule-status launch-readiness-gate-schedule-cancel paid-launch-checklist backup-restore-drill mem-up-release mem-up-lite-release release-lock-verify qdrant-cloud-check quickstart submission-preflight launch-lock launch-lock-public test-py bench-shortlist bench-qdrant-tuning bench-backend-lanes env-lock-check env-lock-apply sentrux-check sentrux-gate sentrux-gate-save agent-context-gate
 
 help:
 > echo "Targets:"
@@ -54,6 +54,8 @@ help:
 > echo "  storage-ledger-capture|storage-ledger-prune: append/prune metadata-only storage growth ledger"
 > echo "  storage-ledger-install|storage-ledger-status: install hourly ledger runner (launchd)"
 > echo "  memory-graph-quality*: score graph coverage and install bounded repair runner"
+> echo "  recall-quality*: run saved recall eval, terminal quality view, and tuning"
+> echo "  open-core-boundary-audit: verify lite/full/paid branch feature boundaries"
 > echo "  weekly-lineage-rollup: generate weekly per-project lineage + global synergy rollups"
 > echo "  weekly-lineage-install|weekly-lineage-status: install weekly lineage runner (launchd)"
 > echo "  qdrant-cutover: set QDRANT_COLLECTION and rehydrate vectors"
@@ -319,6 +321,21 @@ memory-graph-quality-uninstall:
 memory-graph-quality-status:
 > bash scripts/install_memory_graph_quality_runner.sh status
 
+recall-quality:
+> scripts/agent/recall-quality-eval --tuning --pretty
+
+recall-quality-refresh:
+> scripts/agent/recall-quality-eval --refresh-cases --tuning --pretty
+
+recall-quality-tuning:
+> if [ -f .env ]; then source .env >/dev/null 2>&1 || true; fi
+> base="$${CONTEXTLATTICE_ORCHESTRATOR_URL:-http://127.0.0.1:8075}"; key="$${CONTEXTLATTICE_ORCHESTRATOR_API_KEY:-}"; \
+> if [ -n "$$key" ]; then curl -fsS -H "x-api-key: $$key" "$${base%/}/telemetry/recall/tuning?min_samples=1" | jq .; \
+> else curl -fsS "$${base%/}/telemetry/recall/tuning?min_samples=1" | jq .; fi
+
+open-core-boundary-audit:
+> scripts/agent/audit-open-core-boundary --pretty
+
 weekly-lineage-rollup:
 > scripts/context_storage_ops.sh weekly-lineage \
 >   --orchestrator-url "$${CONTEXTLATTICE_ORCHESTRATOR_URL:-http://127.0.0.1:8075}" \

@@ -0,0 +1,16 @@
+import { NextResponse } from "next/server";
+import { callOrchestrator } from "@/lib/orchestrator";
+
+export async function GET(request: Request) {
+  const url = new URL(request.url);
+  const params = new URLSearchParams();
+  for (const key of ["traffic_class"]) {
+    const value = url.searchParams.get(key);
+    if (value) {
+      params.set(key, value);
+    }
+  }
+  const suffix = params.toString() ? `?${params.toString()}` : "";
+  const data = await callOrchestrator(`/telemetry/recall${suffix}`);
+  return NextResponse.json(data);
+}
@@ -0,0 +1,16 @@
+import { NextResponse } from "next/server";
+import { callOrchestrator } from "@/lib/orchestrator";
+
+export async function GET(request: Request) {
+  const url = new URL(request.url);
+  const params = new URLSearchParams();
+  for (const key of ["lookback_hours", "min_samples", "max_samples"]) {
+    const value = url.searchParams.get(key);
+    if (value) {
+      params.set(key, value);
+    }
+  }
+  const suffix = params.toString() ? `?${params.toString()}` : "";
+  const data = await callOrchestrator(`/telemetry/recall/tuning${suffix}`);
+  return NextResponse.json(data);
+}
@@ -3,6 +3,7 @@
 import { useEffect, useState } from "react";
 import { RetrievalPanel } from "@/components/RetrievalPanel";
 import { MemoryGraphPanel, type MemoryGraphPayload } from "@/components/MemoryGraphPanel";
+import { RecallQualityPanel, type RecallQualityPayload, type RecallTuningPayload } from "@/components/RecallQualityPanel";
 
 type Service = {
   name: string;
@@ -63,18 +64,22 @@ export default function StatusPage() {
   const [topics, setTopics] = useState<TopicsPayload | null>(null);
   const [memoryTelemetry, setMemoryTelemetry] = useState<MemoryTelemetry | null>(null);
   const [memoryGraph, setMemoryGraph] = useState<MemoryGraphPayload | null>(null);
+  const [recallQuality, setRecallQuality] = useState<RecallQualityPayload | null>(null);
+  const [recallTuning, setRecallTuning] = useState<RecallTuningPayload | null>(null);
   const [error, setError] = useState<string | null>(null);
   const [updatedAt, setUpdatedAt] = useState<string | null>(null);
 
   async function loadStatus() {
     try {
       setError(null);
-      const [statusRes, prefRes, topicRes, memRes, graphRes] = await Promise.all([
+      const [statusRes, prefRes, topicRes, memRes, graphRes, recallRes, tuningRes] = await Promise.all([
         fetch("/api/memory/status", { cache: "no-store" }),
         fetch("/api/memory/preferences", { cache: "no-store" }),
         fetch("/api/memory/topics", { cache: "no-store" }),
         fetch("/api/telemetry/memory", { cache: "no-store" }),
         fetch("/api/telemetry/memory/graph", { cache: "no-store" }),
+        fetch("/api/telemetry/recall", { cache: "no-store" }),
+        fetch("/api/telemetry/recall/tuning", { cache: "no-store" }),
       ]);
       const statusData = await statusRes.json();
       if (!statusRes.ok) {
@@ -93,6 +98,12 @@ export default function StatusPage() {
       if (graphRes.ok) {
         setMemoryGraph(await graphRes.json());
       }
+      if (recallRes.ok) {
+        setRecallQuality(await recallRes.json());
+      }
+      if (tuningRes.ok) {
+        setRecallTuning(await tuningRes.json());
+      }
       setUpdatedAt(new Date().toLocaleTimeString());
     } catch (err: any) {
       setError(err?.message || "Status unavailable");
@@ -214,6 +225,8 @@ export default function StatusPage() {
 
       <MemoryGraphPanel graph={memoryGraph} />
 
+      <RecallQualityPanel recall={recallQuality} tuning={recallTuning} />
+
       <RetrievalPanel />
     </div>
   );

@@ -0,0 +1,195 @@
+"use client";
+
+type QualityTotals = {
+  requests?: number;
+  timeouts?: number;
+  errors?: number;
+  sourceErrorRate?: number;
+  noHitRate?: number;
+  lowConfidenceRate?: number;
+  staleHitRate?: number;
+  recallAtK?: number | null;
+  mrr?: number | null;
+  numericExactness?: number | null;
+  citationCoverage?: number | null;
+  sourceDiversity?: number | null;
+  graphLift?: number | null;
+  evalP95Ms?: number | null;
+  lastEvalAt?: string | null;
+};
+
+export type RecallQualityPayload = {
+  updatedAt?: string;
+  trafficClass?: string;
+  quality?: {
+    status?: string;
+    totals?: QualityTotals;
+    sampleCount?: number;
+    recommendations?: string[];
+  };
+  alerts?: {
+    count?: number;
+  };
+};
+
+export type RecallTuningPayload = {
+  window?: {
+    samples?: number;
+    minSamples?: number;
+    sufficient?: boolean;
+  };
+  recommended?: {
+    quality?: {
+      graphExpansion?: {
+        enabled?: boolean;
+        depth?: number;
+        neighborLimit?: number;
+      };
+      sourceOrder?: string[];
+      recommendations?: string[];
+    };
+  };
+};
+
+function numberValue(value: unknown): number {
+  return typeof value === "number" && Number.isFinite(value) ? value : 0;
+}
+
+function percentText(value: unknown): string {
+  if (typeof value !== "number" || !Number.isFinite(value)) {
+    return "-";
+  }
+  return `${Math.round(value * 100)}%`;
+}
+
+function statusTone(status: string) {
+  if (status === "repair_recommended" || status === "insufficient_cases") {
+    return "bg-amber-500 text-amber-950";
+  }
+  if (status === "watch" || status === "unknown") {
+    return "bg-cyan-500 text-cyan-950";
+  }
+  return "bg-emerald-500 text-emerald-950";
+}
+
+function QualityBar({ value, tone = "good" }: { value: number; tone?: "good" | "warn" | "neutral" }) {
+  const pct = Math.max(0, Math.min(100, value * 100));
+  const color = tone === "warn" ? "bg-amber-300" : tone === "neutral" ? "bg-cyan-300" : "bg-emerald-300";
+  return (
+    <div className="h-2 w-full rounded bg-slate-800 overflow-hidden" aria-hidden="true">
+      <div className={`h-full ${color}`} style={{ width: `${pct}%` }} />
+    </div>
+  );
+}
+
+function Metric({ label, value, tone }: { label: string; value: string; tone?: "warn" | "good" }) {
+  const toneClass =
+    tone === "warn" ? "text-amber-200 border-amber-500/70" : tone === "good" ? "text-emerald-200 border-emerald-500/70" : "text-slate-200 border-slate-600";
+  return (
+    <div className={`rounded border px-3 py-2 ${toneClass}`}>
+      <div className="text-xs uppercase tracking-wide text-slate-400">{label}</div>
+      <div className="text-lg font-semibold">{value}</div>
+    </div>
+  );
+}
+
+export function RecallQualityPanel({
+  recall,
+  tuning,
+}: {
+  recall: RecallQualityPayload | null;
+  tuning?: RecallTuningPayload | null;
+}) {
+  if (!recall) {
+    return (
+      <section className="card">
+        <h3 className="text-lg font-semibold">Recall quality</h3>
+        <p className="text-sm text-slate-400 mt-2">Recall telemetry unavailable.</p>
+      </section>
+    );
+  }
+
+  const totals = recall.quality?.totals ?? {};
+  const status = String(recall.quality?.status || "unknown");
+  const graphExpansion = tuning?.recommended?.quality?.graphExpansion;
+  const sourceOrder = tuning?.recommended?.quality?.sourceOrder ?? [];
+  const recallAtK = typeof totals.recallAtK === "number" ? totals.recallAtK : 0;
+  const mrr = typeof totals.mrr === "number" ? totals.mrr : 0;
+  const citationCoverage = typeof totals.citationCoverage === "number" ? totals.citationCoverage : 0;
+  const graphLift = typeof totals.graphLift === "number" ? totals.graphLift : 0;
+  const recommendations = [
+    ...(recall.quality?.recommendations ?? []),
+    ...(tuning?.recommended?.quality?.recommendations ?? []),
+  ].slice(0, 4);
+
+  return (
+    <section className="card space-y-5">
+      <div className="flex flex-wrap items-start justify-between gap-3">
+        <div>
+          <h3 className="text-lg font-semibold">Recall quality</h3>
+          <p className="text-xs text-slate-500 mt-1">
+            {totals.lastEvalAt ? `Last eval ${new Date(totals.lastEvalAt).toLocaleTimeString()}` : "No saved eval sample yet"}
+          </p>
+        </div>
+        <span className={`text-xs px-2 py-1 rounded ${statusTone(status)}`}>{status}</span>
+      </div>
+
+      <div className="grid md:grid-cols-6 gap-3 text-sm">
+        <Metric label="Recall@K" value={percentText(totals.recallAtK)} tone={recallAtK >= 0.75 ? "good" : "warn"} />
+        <Metric label="MRR" value={mrr ? mrr.toFixed(2) : "-"} tone={mrr >= 0.55 ? "good" : "warn"} />
+        <Metric label="Citations" value={percentText(totals.citationCoverage)} tone={citationCoverage >= 0.9 ? "good" : "warn"} />
+        <Metric label="Graph lift" value={percentText(totals.graphLift)} tone={graphLift > 0 ? "good" : undefined} />
+        <Metric label="Diversity" value={numberValue(totals.sourceDiversity).toFixed(1)} />
+        <Metric label="Eval p95" value={totals.evalP95Ms ? `${Math.round(totals.evalP95Ms)} ms` : "-"} />
+      </div>
+
+      <div className="grid lg:grid-cols-[minmax(0,1fr)_18rem] gap-5">
+        <div className="space-y-3">
+          <div className="grid grid-cols-[5rem_minmax(0,1fr)_3.5rem] items-center gap-3 text-xs">
+            <span className="text-slate-400">recall</span>
+            <QualityBar value={recallAtK} tone={recallAtK >= 0.75 ? "good" : "warn"} />
+            <span className="text-right text-slate-300">{percentText(totals.recallAtK)}</span>
+          </div>
+          <div className="grid grid-cols-[5rem_minmax(0,1fr)_3.5rem] items-center gap-3 text-xs">
+            <span className="text-slate-400">mrr</span>
+            <QualityBar value={mrr} tone={mrr >= 0.55 ? "good" : "warn"} />
+            <span className="text-right text-slate-300">{mrr ? mrr.toFixed(2) : "-"}</span>
+          </div>
+          <div className="grid grid-cols-[5rem_minmax(0,1fr)_3.5rem] items-center gap-3 text-xs">
+            <span className="text-slate-400">graph</span>
+            <QualityBar value={Math.min(1, graphLift * 4)} tone={graphLift > 0 ? "neutral" : "good"} />
+            <span className="text-right text-slate-300">{percentText(totals.graphLift)}</span>
+          </div>
+        </div>
+
+        <div className="rounded border border-slate-700/70 p-3 text-xs text-slate-300">
+          <div className="font-semibold text-slate-100 mb-2">Tuning</div>
+          <div className="flex justify-between gap-3">
+            <span className="text-slate-500">Graph depth</span>
+            <span>{graphExpansion?.enabled ? `${graphExpansion.depth ?? 0} / ${graphExpansion.neighborLimit ?? 0}` : "off"}</span>
+          </div>
+          <div className="mt-2 text-slate-500">Sources</div>
+          <div className="mt-1 flex flex-wrap gap-1">
+            {sourceOrder.slice(0, 5).map((source) => (
+              <span key={source} className="rounded border border-slate-700 px-1.5 py-0.5 text-slate-300">
+                {source}
+              </span>
+            ))}
+            {!sourceOrder.length ? <span>-</span> : null}
+          </div>
+        </div>
+      </div>
+
+      {recommendations.length ? (
+        <div className="rounded border border-slate-700/70 p-3 text-xs text-slate-300">
+          <div className="font-semibold text-slate-100 mb-1">Recommended next action</div>
+          <ul className="space-y-1">
+            {recommendations.map((item) => (
+              <li key={item}>{item}</li>
+            ))}
+          </ul>
+        </div>
+      ) : null}
+    </section>
+  );
+}
@@ -14,7 +14,7 @@
     "billing:reconcile:coinbase": "TS_NODE_COMPILER_OPTIONS='{\"module\":\"commonjs\",\"jsx\":\"react-jsx\"}' node --require tsconfig-paths/register --require ts-node/register scripts/reconcile_coinbase.ts",
     "audit:export": "TS_NODE_COMPILER_OPTIONS='{\"module\":\"commonjs\",\"jsx\":\"react-jsx\"}' node --require tsconfig-paths/register --require ts-node/register scripts/export_audit_logs.ts",
     "audit:prune": "TS_NODE_COMPILER_OPTIONS='{\"module\":\"commonjs\",\"jsx\":\"react-jsx\"}' node --require tsconfig-paths/register --require ts-node/register scripts/prune_audit_logs.ts",
-    "test": "TS_NODE_COMPILER_OPTIONS='{\"module\":\"commonjs\",\"jsx\":\"react-jsx\"}' node --require tsconfig-paths/register --test --require ts-node/register tests/overrides-api.test.ts tests/overrides-panel.test.tsx tests/sidecar-health-api.test.ts tests/sidecar-health-panel.test.tsx tests/charts-panel.test.tsx tests/alerts-panel.test.tsx tests/strategy-panel.test.tsx"
+    "test": "TS_NODE_COMPILER_OPTIONS='{\"module\":\"commonjs\",\"jsx\":\"react-jsx\"}' node --require tsconfig-paths/register --test --require ts-node/register tests/overrides-api.test.ts tests/overrides-panel.test.tsx tests/sidecar-health-api.test.ts tests/sidecar-health-panel.test.tsx tests/charts-panel.test.tsx tests/alerts-panel.test.tsx tests/strategy-panel.test.tsx tests/recall-quality-panel.test.tsx"
   },
   "dependencies": {
     "@next-auth/prisma-adapter": "^1.0.7",

@@ -0,0 +1,41 @@
+import assert from "node:assert/strict";
+import test from "node:test";
+import React from "react";
+import { renderToStaticMarkup } from "react-dom/server";
+import { RecallQualityPanel } from "@/components/RecallQualityPanel";
+
+test("RecallQualityPanel renders quality metrics and tuning", () => {
+  const html = renderToStaticMarkup(
+    <RecallQualityPanel
+      recall={{
+        quality: {
+          status: "healthy",
+          totals: {
+            recallAtK: 0.91,
+            mrr: 0.78,
+            citationCoverage: 1,
+            sourceDiversity: 2.4,
+            graphLift: 0.12,
+            evalP95Ms: 184,
+            lastEvalAt: "2026-05-29T12:00:00Z",
+          },
+          recommendations: ["Recall quality telemetry is inside current production thresholds."],
+        },
+      }}
+      tuning={{
+        recommended: {
+          quality: {
+            graphExpansion: { enabled: true, depth: 1, neighborLimit: 12 },
+            sourceOrder: ["qdrant", "postgres_pgvector", "topic_rollups"],
+          },
+        },
+      }}
+    />,
+  );
+
+  assert.match(html, /Recall quality/);
+  assert.match(html, /91%/);
+  assert.match(html, /Graph lift/);
+  assert.match(html, /qdrant/);
+  assert.match(html, /1 \/ 12/);
+});