fix

apps/api/internal/company/synth_heuristic_test.go

@@ -0,0 +1,58 @@
+package company
+
+import "testing"
+
+// Regression: the generic invent shapes (" is a … product from …" and
+// " je … znamk|kategor|produkt …") matched any length of copy. In Slovenian " je "
+// appears in almost every sentence and A1 spec lists start with "Znamka: <brand>",
+// so real formula output was classified as invent fallback, discarded, and replaced
+// with the supplier feed text — Review then showed Original ≈ Enriched.
+func TestLooksLikeHeuristicSynthesize_keepsRealFormulaCopy(t *testing.T) {
+	t.Parallel()
+	slovenianFormulaCopy := "<h2>BRUNNER zložljiv stol ONE SHOT za udobje na poti</h2>" +
+		"<p>Zložljiv stol iz aluminija je zasnovan za dolge ure udobnega sedenja na terenu, na snemanju ali ob " +
+		"kampiranju. Okvir se razpre v enem gibu in ostane stabilen tudi na neravnih tleh, medtem ko tkanina " +
+		"sedišča prijetno diha in se hitro suši. Naslonjala za roke razbremenijo ramena, hrbtni del pa podpira " +
+		"naravno držo. Zaradi majhne teže ga brez napora prenesete od avtomobila do prizorišča.</p>" +
+		"<h2>Stabilnost in dolga življenjska doba</h2>" +
+		"<p>Konstrukcija združuje trpežne materiale in premišljene detajle. Aluminijasti profili so odporni proti " +
+		"koroziji, spoji pa so ojačani na mestih največjih obremenitev. Sivo črna kombinacija ostaja videti " +
+		"urejena tudi po sezoni uporabe na prostem, saj se madeži manj poznajo.</p>" +
+		"<ul><li>Znamka: BRUNNER</li><li>Model: ONE SHOT</li><li>Barva: sivo črna</li></ul>"
+	if LooksLikeHeuristicSynthesize(slovenianFormulaCopy) {
+		t.Fatal("real Slovenian formula copy must not read as invent fallback")
+	}
+
+	englishFormulaCopy := "<h2>Acme Monitor Pro</h2><p>This monitor is a strong choice for colour-critical work, " +
+		"with a factory-calibrated panel and a stand that adjusts through a wide range. Cable routing keeps the " +
+		"desk tidy, and the built-in hub means one cable to the laptop. Panel uniformity holds up across the " +
+		"whole surface, so gradients stay smooth from edge to edge in long editing sessions.</p>" +
+		"<ul><li>Brand: Acme</li><li>Key features: USB-C hub, height adjustment</li></ul>"
+	// "Key features" is ordinary section wording: suggestive enough to force a
+	// re-enhance, never enough to discard the copy.
+	if IsInventFallbackDescription(englishFormulaCopy) {
+		t.Fatal("real English formula copy must not read as invent fallback")
+	}
+	if IsInventFallbackDescription(slovenianFormulaCopy) {
+		t.Fatal("real Slovenian formula copy must not read as invent fallback")
+	}
+}
+
+func TestLooksLikeHeuristicSynthesize_stillCatchesInvent(t *testing.T) {
+	t.Parallel()
+	cases := []string{
+		"Vogel's WALL 3245 is a TV Mounts product from Vogel's. Key specs: width 45 cm, max_load 40 kg.",
+		"Gorenje EHT6020 je izdelek znamke Gorenje.",
+		"Samsung QLED je izdelek v kategoriji Televizorji.",
+		"Acme Widget is a catalog product with the known attributes.",
+		"Acme Widget — catalog product.",
+		"BRUNNER ONE SHOT — Stolčki, znamka BRUNNER. Širina: 44 cm.",
+		"<h2>Gorenje EHT6020</h2><p>Gorenje EHT6020 — Štedilniki, znamka Gorenje.</p>" +
+			"<h3>Ključne lastnosti</h3><ul><li>Širina: 60 cm</li></ul>",
+	}
+	for _, c := range cases {
+		if !LooksLikeHeuristicSynthesize(c) {
+			t.Fatalf("invent fallback must still be detected: %q", c)
+		}
+	}
+}

apps/api/internal/company/weak_desc.go

@@ -31,10 +31,12 @@ func IsWeakPriorEnhanceDescription(priorDesc string, titles ...string) bool {
 	return reason == "weak" || reason == "title-echo"
 }
 
-// heuristicSynthesizePhrases are distinctive invent / formula-skeleton snippets from
+// heuristicSynthesizePhrases are distinctive invent snippets from
 // synthesizeDescriptionFromTitle / synthesizeDescriptionFromFormula / factualDescriptionIntro.
 // These may look "strong" enough to pass IsWeakPriorEnhanceDescription but must never
 // hash-skip enhance (would leave fallback copy forever on reprocess).
+// Only whole sentences that no copywriter would produce belong here — anything a
+// real category formula could legitimately emit goes in synthSkeletonPhrases.
 var heuristicSynthesizePhrases = []string{
 	"is a catalog product with the known attributes",
 	"is listed in the ",
@@ -43,19 +45,44 @@ var heuristicSynthesizePhrases = []string{
 	"je izdelek v kategoriji",
 	"je izdelek znamke",
 	". ključne specifikacije:",
+	" — katalogski izdelek",
+	" — catalog product",
+}
+
+// synthSkeletonPhrases also appear in synthesize output but are ordinary words a
+// real description may use ("Ključne lastnosti" as a heading, "znamka X" in a spec
+// line, "Key features" as a section title). They are suggestive, not proof, so only
+// the hash-skip gate consults them — see IsInventFallbackDescription.
+var synthSkeletonPhrases = []string{
 	// factualDescriptionIntro (post-phrase-avoidance invent)
 	" — znamka ",
 	", znamka ",
-	" — katalogski izdelek",
 	" — from ",
-	" — catalog product",
 	// synthesizeDescriptionFromFormula heading fallbacks
 	"ključne lastnosti",
 	"key features",
 }
 
-// LooksLikeHeuristicSynthesize reports invent / formula-skeleton fallback copy.
-func LooksLikeHeuristicSynthesize(desc string) bool {
+// maxSynthSkeletonRunes bounds the suggestive signals above. Every synthesize helper
+// emits at most a title heading, one intro sentence, and a short "Label: value"
+// list, so it stays well under this; real category-formula prose (multiple 100-word
+// paragraphs plus a spec list) runs several times longer.
+const maxSynthSkeletonRunes = 600
+
+// maxSynthInventRunes bounds the generic "<title> is a … product from …" /
+// "<title> je … znamke …" shapes. Those describe a single invent sentence, so they
+// may only match copy that is itself about one sentence long. Unbounded, " je "
+// plus "znamk" matched almost any Slovenian description.
+const maxSynthInventRunes = 2 * shortBoilerplateDescRunes
+
+// IsInventFallbackDescription reports copy that is unmistakably machine invent —
+// whole sentences emitted by the synthesize helpers, or the one-sentence
+// "<title> is a <category> product from <brand>" shape.
+//
+// This is the strict test, safe for discarding a description outright. The looser
+// LooksLikeHeuristicSynthesize adds ambiguous wording and is only used where a
+// false positive costs one extra enhance call rather than the copy itself.
+func IsInventFallbackDescription(desc string) bool {
 	plain := strings.ToLower(plainTextForWeakCheck(desc))
 	if plain == "" {
 		return false
@@ -65,19 +92,39 @@ func LooksLikeHeuristicSynthesize(desc string) bool {
 			return true
 		}
 	}
+	if len([]rune(plain)) > maxSynthInventRunes {
+		return false
+	}
 	// EN invent: "<title> is a <category> product from <brand>"
 	if strings.Contains(plain, " is a ") && strings.Contains(plain, " product from ") {
 		return true
 	}
 	// SL/CS short invent: "<title> je … znamk|kategor|produkt …"
-	if strings.Contains(plain, " je ") &&
+	return strings.Contains(plain, " je ") &&
 		(strings.Contains(plain, "znamk") ||
 			strings.Contains(plain, "kategor") ||
 			strings.Contains(plain, "produkt") ||
 			strings.Contains(plain, "televiz") ||
-			strings.Contains(plain, "monitor")) {
+			strings.Contains(plain, "monitor"))
+}
+
+// LooksLikeHeuristicSynthesize reports invent / formula-skeleton fallback copy for
+// the enhance hash-skip gate: IsInventFallbackDescription plus wording that merely
+// suggests a skeleton, inside a body short enough to be one. A false positive here
+// forces one re-enhance, which is cheap; never use it to drop a description.
+func LooksLikeHeuristicSynthesize(desc string) bool {
+	if IsInventFallbackDescription(desc) {
 		return true
 	}
+	plain := strings.ToLower(plainTextForWeakCheck(desc))
+	if plain == "" || len([]rune(plain)) > maxSynthSkeletonRunes {
+		return false
+	}
+	for _, p := range synthSkeletonPhrases {
+		if p != "" && strings.Contains(plain, p) {
+			return true
+		}
+	}
 	return false
 }