feat(verify): report the verified rate per product, not just per SKU

78% of the smartphone set is regional and RAM variants of the same
phone: 57,216 records stand for 20,594 actual models, with slugs like
"…k61-2020-latam-q630ha-…-costa-rica-4gb-128gb". No source documents an
individual SKU, so counting verification per record measures the shape
of the import rather than how well the products are sourced.

`status` now also aggregates by model — variants collapse onto their
base model, a standalone record is its own model, and a model counts as
verified once any of its SKUs is, since the variants of one phone share
a provenance. The per-record figures are unchanged and still reported;
this is a second view of the same verifications, not a replacement.

Refs #1

Author: Seungpyo1007

app/verify/cli.py

@@ -254,14 +254,24 @@ def cmd_status(args: argparse.Namespace) -> int:
 
     by_category: dict[str, dict[str, Any]] = {}
     tot = ver = g = y = r = 0
+    all_models: set[tuple[str, ...]] = set()
+    all_verified_models: set[tuple[str, ...]] = set()
     for cat in CATEGORIES:
         ct = cv = cg = cy = cr = 0
+        # A model counts as verified once any of its SKUs is: the variants of one
+        # phone share a provenance, so the source that verifies one describes the
+        # product the others are configurations of.
+        models: set[tuple[str, ...]] = set()
+        verified_models: set[tuple[str, ...]] = set()
         for rec in records[cat]:
             if not rec.slug:
                 continue
             ct += 1
+            key = rec.model_key
+            models.add(key)
             if rec.verified:
                 cv += 1
+                verified_models.add(key)
             band = offline.score_record(rec, now_year, soc_release).band
             cg += band == "green"
             cy += band == "yellow"
@@ -270,6 +280,11 @@ def cmd_status(args: argparse.Namespace) -> int:
             "total": ct,
             "verified": cv,
             "verified_pct": round(100 * cv / ct, 2) if ct else 0.0,
+            "models": len(models),
+            "verified_models": len(verified_models),
+            "verified_models_pct": (
+                round(100 * len(verified_models) / len(models), 2) if models else 0.0
+            ),
             "green": cg,
             "yellow": cy,
             "red": cr,
@@ -281,6 +296,8 @@ def cmd_status(args: argparse.Namespace) -> int:
         g += cg
         y += cy
         r += cr
+        all_models |= models
+        all_verified_models |= verified_models
 
     status = {
         "generated_at": _now_iso(),
@@ -289,6 +306,13 @@ def cmd_status(args: argparse.Namespace) -> int:
             "records": tot,
             "verified": ver,
             "verified_pct": round(100 * ver / tot, 2) if tot else 0.0,
+            # Same verifications counted per product rather than per SKU.
+            "models": len(all_models),
+            "verified_models": len(all_verified_models),
+            "verified_models_pct": (
+                round(100 * len(all_verified_models) / len(all_models), 2)
+                if all_models else 0.0
+            ),
             "green": g,
             "yellow": y,
             "red": r,

app/verify/common.py

@@ -54,6 +54,22 @@ def slug(self) -> str | None:
     def verified(self) -> bool:
         return self.data.get("verified") is True
 
+    @property
+    def model_key(self) -> tuple[str, ...]:
+        """Identifies the product, not the SKU.
+
+        78% of the smartphone set is regional/RAM variants of the same phone —
+        one model can carry dozens of records ("…k61…latam-q630ha…costa-rica-
+        4gb-128gb"). Counting per record makes the dataset look far less
+        verified than the products in it are, because no source documents an
+        individual SKU. Variants collapse onto their base model; a standalone
+        record is its own model.
+        """
+        base = self.data.get("base_model_slug")
+        if isinstance(base, str) and base:
+            return (self.category, str(self.data.get("brand") or ""), base)
+        return (self.category, self.slug or self.path)
+
     def content_hash(self) -> str:
         """Stable hash of the record body — invalidates stale ledger decisions on edit."""
         blob = json.dumps(self.data, sort_keys=True, ensure_ascii=False)

tests/verify/test_offline.py

@@ -63,3 +63,26 @@ def test_future_release_red():
         "source_urls": ["https://en.wikipedia.org/wiki/x"],
     }
     assert _score("cpu", rec).band == "red"
+
+
+def test_model_key_collapses_variants_of_one_phone():
+    """Regional/RAM SKUs of one phone are one product, not many."""
+    from app.verify.common import Record
+
+    def variant(slug, base):
+        return Record("smartphone", f"smartphone/lg/2020/{base}/{slug}.json",
+                      {"slug": slug, "brand": "lg", "base_model_slug": base})
+
+    a = variant("lg-k61-costa-rica-4gb-128gb", "k61-2020")
+    b = variant("lg-k61-colombia-3gb-64gb", "k61-2020")
+    other = variant("lg-k51-usa-3gb-32gb", "k51-2020")
+    assert a.model_key == b.model_key
+    assert a.model_key != other.model_key
+
+
+def test_model_key_of_a_standalone_record_is_itself():
+    from app.verify.common import Record
+
+    rec = Record("cpu", "cpu/intel/2023/desktop/core-i9-14900k.json",
+                 {"slug": "core-i9-14900k"})
+    assert rec.model_key == ("cpu", "core-i9-14900k")