configuration support for optional LLM extraction

README.md

@@ -127,6 +127,9 @@ When an extractor is present and returns candidates, analysis stores those
 reviewable candidates; when it returns no candidates, the deterministic
 heuristic generator remains the fallback.
 
+The FastAPI settings object also accepts `llm_provider` and `llm_model`. By
+default `llm_provider` is unset, so analysis is fully offline and deterministic.
+
 ## Agent-Facing Endpoints
 
 The v0.1 API covers the main registration, analysis, review, search, and inspection loop:

src/repo_registry/web_api/app.py

@@ -7,6 +7,7 @@ from fastapi import Depends, FastAPI, HTTPException
 from pydantic import BaseModel, Field
 
 from repo_registry.core.service import RegistryService
+from repo_registry.llm_extraction import LLMCandidateExtractor, create_llm_connect_adapter
 from repo_registry.repo_ingestion.git import GitIngestionService
 from repo_registry.storage.sqlite import NotFoundError, RegistryStore
 
@@ -14,6 +15,8 @@ from repo_registry.storage.sqlite import NotFoundError, RegistryStore
 class Settings(BaseModel):
     database_path: str = Field(default="var/repo-registry.sqlite3")
     checkout_root: str = Field(default="var/checkouts")
+    llm_provider: str | None = Field(default=None)
+    llm_model: str | None = Field(default=None)
 
 
 def get_settings() -> Settings:
@@ -25,7 +28,18 @@ def get_service(settings: Settings = Depends(get_settings)) -> RegistryService:
     database_path.parent.mkdir(parents=True, exist_ok=True)
     store = RegistryStore(database_path)
     store.initialize()
-    return RegistryService(store, ingestion=GitIngestionService(settings.checkout_root))
+    llm_extractor = None
+    if settings.llm_provider:
+        adapter = create_llm_connect_adapter(
+            settings.llm_provider,
+            model=settings.llm_model,
+        )
+        llm_extractor = LLMCandidateExtractor(adapter)
+    return RegistryService(
+        store,
+        ingestion=GitIngestionService(settings.checkout_root),
+        llm_extractor=llm_extractor,
+    )
 
 
 class RepositoryCreate(BaseModel):

tests/test_web_api.py

@@ -1,6 +1,7 @@
 from fastapi.testclient import TestClient
 
-from repo_registry.web_api.app import Settings, app, get_settings
+from repo_registry.web_api import app as app_module
+from repo_registry.web_api.app import Settings, app, get_service, get_settings
 
 
 def test_api_manual_registry_loop(tmp_path):
@@ -164,6 +165,34 @@ def test_api_registers_repository_from_url_metadata(tmp_path):
         app.dependency_overrides.clear()
 
 
+def test_api_service_settings_can_enable_llm_extractor(monkeypatch, tmp_path):
+    class Response:
+        content = '{"abilities": [{"name": "Configured LLM Ability"}]}'
+
+    class Adapter:
+        def execute_prompt(self, prompt, config):
+            return Response()
+
+    calls = []
+
+    def fake_create_adapter(provider, model=None):
+        calls.append((provider, model))
+        return Adapter()
+
+    monkeypatch.setattr(app_module, "create_llm_connect_adapter", fake_create_adapter)
+    service = get_service(
+        Settings(
+            database_path=str(tmp_path / "llm-settings.sqlite3"),
+            checkout_root=str(tmp_path / "checkouts"),
+            llm_provider="mock",
+            llm_model="demo-model",
+        )
+    )
+
+    assert calls == [("mock", "demo-model")]
+    assert service.llm_extractor is not None
+
+
 def test_api_analysis_run_loop(tmp_path):
     source = tmp_path / "repo"
     source.mkdir()