add deterministic pilot benchmarks

docs/benchmarks/README.md

@@ -0,0 +1,19 @@
+# Phase 16 pilot benchmark
+
+Run from the repository root:
+
+```bash
+python3 scripts/benchmark_phase16.py
+```
+
+The command is offline-only and writes `docs/benchmarks/phase16.latest.json`. Use
+`--no-latency` when comparing deterministic result sections in CI. The committed
+fixture is `apps/api/fixtures/phase16.json`; it is intentionally synthetic and
+small. The JSON report repeats its limitations and acceptance thresholds and is
+validated structurally by `phase16.schema.json`.
+
+Acceptance thresholds are normalization accuracy 100%; fuzzy-match precision and
+recall at least 90%; contact precision and recall at least 85%; website fixture
+accuracy at least 90%; reproducible scoring; and zero tenant leakage. Timing and
+cache ratios are diagnostic pilot measurements, not production capacity claims.
+No benchmark performs network outreach or website fetching.

docs/benchmarks/phase16.latest.json

@@ -0,0 +1,136 @@
+{
+  "acceptance_thresholds": {
+    "contact_precision": 0.85,
+    "contact_recall": 0.85,
+    "matching_precision": 0.9,
+    "matching_recall": 0.9,
+    "normalization_accuracy": 1.0,
+    "score_reproducible": true,
+    "tenant_leakage": 0,
+    "website_accuracy": 0.9
+  },
+  "benchmark": "phase16",
+  "checks": {
+    "contact_precision": true,
+    "contact_recall": true,
+    "matching_precision": true,
+    "matching_recall": true,
+    "normalization_accuracy": true,
+    "score_reproducible": true,
+    "tenant_leakage": true,
+    "website_accuracy": true
+  },
+  "fixture": "apps/api/fixtures/phase16.json",
+  "limitations": [
+    "Synthetic, small, English-heavy cases are not representative of production traffic.",
+    "Contact labels cover parser false positives but do not establish consent, identity, or deliverability.",
+    "Website fixtures classify supplied HTML only; they do not measure network, DNS, redirects, or adversarial pages.",
+    "Latency and cache measurements are local pilot signals and vary by host load."
+  ],
+  "offline": true,
+  "passed": true,
+  "results": {
+    "cache": {
+      "enabled": true,
+      "timing_omitted": true
+    },
+    "contacts": {
+      "cases": [
+        {
+          "expected_count": 2,
+          "false_positives": [],
+          "name": "public contacts and ignored markup",
+          "predicted_count": 2
+        },
+        {
+          "expected_count": 1,
+          "false_positives": [],
+          "name": "forms are provenance not email destinations",
+          "predicted_count": 1
+        }
+      ],
+      "false_negative": 0,
+      "false_positive": 0,
+      "precision": 1.0,
+      "recall": 1.0,
+      "true_positive": 3
+    },
+    "latency": {},
+    "matching": {
+      "cases": [
+        {
+          "expected_ids": [
+            1
+          ],
+          "name": "shared exact domain is a positive",
+          "predicted_ids": [
+            1
+          ]
+        },
+        {
+          "expected_ids": [
+            4
+          ],
+          "name": "similar name without corroboration is not a positive",
+          "predicted_ids": [
+            4
+          ]
+        },
+        {
+          "expected_ids": [],
+          "name": "unrelated businesses are negatives",
+          "predicted_ids": []
+        }
+      ],
+      "false_negative": 0,
+      "false_positive": 0,
+      "false_positive_cases": 0,
+      "precision": 1.0,
+      "recall": 1.0,
+      "true_positive": 2
+    },
+    "normalization": {
+      "accuracy": 1.0,
+      "correct": 2,
+      "total": 2
+    },
+    "scoring": {
+      "fixture_scores": [
+        {
+          "eligible": true,
+          "name": "complete active business",
+          "score": 80
+        },
+        {
+          "eligible": false,
+          "name": "suppressed remains ineligible",
+          "score": 15
+        },
+        {
+          "eligible": true,
+          "name": "stale evidence does not create positive",
+          "score": 15
+        }
+      ],
+      "reproducible": true
+    },
+    "tenant_isolation": {
+      "isolated": true,
+      "leakage": 0,
+      "tenant_a_visible_ids": [
+        1,
+        2
+      ],
+      "tenant_b_ids": [
+        3,
+        4
+      ]
+    },
+    "website_classification": {
+      "accuracy": 1.0,
+      "correct": 6,
+      "total": 6
+    }
+  },
+  "version": 1
+}

docs/benchmarks/phase16.schema.json

@@ -0,0 +1,16 @@
+{
+  "$schema": "https://json-schema.org/draft/2020-12/schema",
+  "title": "Prospect Platform Phase 16 benchmark report",
+  "type": "object",
+  "required": ["benchmark", "version", "offline", "limitations", "acceptance_thresholds", "results", "checks", "passed"],
+  "properties": {
+    "benchmark": {"const": "phase16"},
+    "version": {"type": "integer", "const": 1},
+    "offline": {"const": true},
+    "limitations": {"type": "array", "items": {"type": "string"}},
+    "acceptance_thresholds": {"type": "object"},
+    "results": {"type": "object"},
+    "checks": {"type": "object"},
+    "passed": {"type": "boolean"}
+  }
+}