Add trust signals to enrichment output

Author: Chnurok

skill/scripts/enrich_lead.py

@@ -445,10 +445,37 @@ def summarize(text, snippets, source_url=None):
     return None
 
 
-def compute_confidence(domain, emails, phones, summary, warnings, best_contact_meta=None):
+def build_trust_signals(domain, emails, phones, summary, warnings, best_contact_meta=None, site_verification=None):
+    warning_penalty = round(min(0.3, 0.05 * len(warnings)), 2)
+    signals = {
+        "has_domain": bool(domain),
+        "has_summary": bool(summary),
+        "email_count": len(emails or []),
+        "phone_count": len(phones or []),
+        "warning_count": len(warnings or []),
+        "warning_penalty": warning_penalty,
+        "site_verified": bool(site_verification and site_verification.get("verified")),
+        "site_verification_score": round((site_verification or {}).get("score", 0.0), 2),
+        "best_contact": {
+            "present": bool(best_contact_meta),
+            "official": bool(best_contact_meta and best_contact_meta.get("official")),
+            "strong": bool(best_contact_meta and best_contact_meta.get("strong")),
+            "weak": bool(best_contact_meta and best_contact_meta.get("weak")),
+            "tier": best_contact_meta.get("tier") if best_contact_meta else None,
+        },
+    }
+    return signals
+
+
+def compute_confidence(domain, emails, phones, summary, warnings, best_contact_meta=None, site_verification=None):
+    signals = build_trust_signals(domain, emails, phones, summary, warnings, best_contact_meta, site_verification)
     score = 0.05
-    if domain:
-        score += 0.3
+    if signals["has_domain"]:
+        score += 0.25
+    if signals["site_verified"]:
+        score += 0.1
+    elif signals["site_verification_score"] >= 1.0:
+        score += 0.05
     if emails:
         score += 0.15
     if phones:
@@ -466,8 +493,8 @@ def compute_confidence(domain, emails, phones, summary, warnings, best_contact_m
             score -= 0.1
     elif emails:
         score -= 0.1
-    score -= min(0.3, 0.05 * len(warnings))
-    return round(max(0.0, min(1.0, score)), 2)
+    score -= signals["warning_penalty"]
+    return round(max(0.0, min(1.0, score)), 2), signals
 
 
 def enrich(company, region=None, domain=None, query_mode="smart"):
@@ -553,9 +580,18 @@ def enrich(company, region=None, domain=None, query_mode="smart"):
         "social_links": social_values,
         "snippets": snippets[:5],
         "confidence": 0.0,
+        "trust_signals": {},
         "warnings": dedupe(warnings),
     }
-    result["confidence"] = compute_confidence(result["primary_domain"], result["emails"], result["phones"], result["summary"], result["warnings"], best_contact_meta)
+    result["confidence"], result["trust_signals"] = compute_confidence(
+        result["primary_domain"],
+        result["emails"],
+        result["phones"],
+        result["summary"],
+        result["warnings"],
+        best_contact_meta,
+        site_verification,
+    )
     return result
 
 

tests/test_enrich_lead.py

@@ -55,9 +55,11 @@ def test_summarize_removes_obvious_junk(self):
     def test_compute_confidence_penalizes_weak_external_contact(self):
         weak_meta = mod.classify_contact_email("press@gmail.com", "acme.com")
         strong_meta = mod.classify_contact_email("hello@acme.com", "acme.com")
-        weak = mod.compute_confidence("acme.com", ["press@gmail.com"], [], "summary", ["Only weak outreach contacts found"], weak_meta)
-        strong = mod.compute_confidence("acme.com", ["hello@acme.com"], [], "summary", [], strong_meta)
+        weak, weak_signals = mod.compute_confidence("acme.com", ["press@gmail.com"], [], "summary", ["Only weak outreach contacts found"], weak_meta)
+        strong, strong_signals = mod.compute_confidence("acme.com", ["hello@acme.com"], [], "summary", [], strong_meta)
         self.assertLess(weak, strong)
+        self.assertTrue(weak_signals["best_contact"]["weak"])
+        self.assertTrue(strong_signals["best_contact"]["official"])
 
     def test_enrich_warns_when_only_weak_contacts_exist(self):
         original_build_queries = mod.build_queries
@@ -96,6 +98,7 @@ def fake_parse_page(url, base_domain):
         self.assertEqual(result["best_contact_source"]["source_url"], "https://acme.com")
         self.assertIn("Only weak outreach contacts found", result["warnings"])
         self.assertLess(result["confidence"], 0.6)
+        self.assertTrue(result["trust_signals"]["best_contact"]["weak"])
 
     def test_summarize_uses_longest_snippet_fallback(self):
         out = mod.summarize(None, ["short", "this is a much longer snippet about a company and what it does"])
@@ -194,6 +197,8 @@ def fake_parse_page(url, base_domain):
         self.assertEqual(result["phone_sources"]["+49 30 123456"]["source_url"], "https://acme.com")
         self.assertEqual(result["summary_source"]["source_url"], "https://acme.com")
         self.assertTrue(result["site_verification"]["verified"])
+        self.assertTrue(result["trust_signals"]["site_verified"])
+        self.assertEqual(result["trust_signals"]["email_count"], 2)
 
 
 if __name__ == "__main__":