improved scanner

src/repo_registry/repo_scanning/scanner.py

@@ -358,6 +358,10 @@ class DeterministicScanner:
             source_role = self._source_role(relative)
             if source_role == "agent_guidance":
                 continue
+            utility_relationship = self._provider_utility_relationship(
+                source_role,
+                relative,
+            )
             for needle, provider in LLM_PROVIDER_HINTS.items():
                 if not self._has_provider_signal(lower_text, needle):
                     continue
@@ -372,6 +376,7 @@ class DeterministicScanner:
                         metadata={
                             "source": "provider_hint",
                             "source_role": source_role,
+                            "utility_relationship": utility_relationship,
                         },
                     ),
                 )
@@ -389,6 +394,7 @@ class DeterministicScanner:
                         metadata={
                             "source": "environment_variable",
                             "source_role": source_role,
+                            "utility_relationship": "configure",
                         },
                     ),
                 )
@@ -412,6 +418,7 @@ class DeterministicScanner:
                             metadata={
                                 "source": "provider_registry_hint",
                                 "source_role": source_role,
+                                "utility_relationship": utility_relationship,
                             },
                         ),
                     )
@@ -429,11 +436,30 @@ class DeterministicScanner:
                         metadata={
                             "source": "fallback_hint",
                             "source_role": source_role,
+                            "utility_relationship": utility_relationship,
                         },
                     ),
                 )
         return facts
 
+    def _provider_utility_relationship(
+        self,
+        source_role: str,
+        relative_path: str,
+    ) -> str:
+        if source_role == "implementation_source":
+            lower = relative_path.lower()
+            if "adapter" in lower or "provider" in lower:
+                return "adapter"
+            return "owned"
+        if source_role == "configuration":
+            return "configure"
+        if source_role == "dependency_declaration":
+            return "dependency"
+        if source_role in {"ci_tooling", "agent_guidance"}:
+            return "tooling"
+        return "mention"
+
     def _source_role(self, relative_path: str) -> str:
         lower = relative_path.lower()
         parts = lower.split("/")