refactor, security updates, cv extraction upgrades

tools/summarizer/tests/test_app.py

@@ -1,5 +1,4 @@
 import importlib
-import os
 import sys
 from pathlib import Path
 
@@ -11,16 +10,23 @@ if str(ROOT) not in sys.path:
     sys.path.insert(0, str(ROOT))
 
 
-def load_app_module(monkeypatch):
-    monkeypatch.setenv("AI_SERVICE_SKIP_MODEL_LOAD", "1")
-    monkeypatch.delenv("OLLAMA_MODEL", raising=False)
+def load_app_module(monkeypatch, *, skip_model_load=True, ollama_model=None):
+    if skip_model_load:
+        monkeypatch.setenv("AI_SERVICE_SKIP_MODEL_LOAD", "1")
+    else:
+        monkeypatch.delenv("AI_SERVICE_SKIP_MODEL_LOAD", raising=False)
+    monkeypatch.delenv("AI_SERVICE_EAGER_MODEL_LOAD", raising=False)
+    if ollama_model is None:
+        monkeypatch.delenv("OLLAMA_MODEL", raising=False)
+    else:
+        monkeypatch.setenv("OLLAMA_MODEL", ollama_model)
     if "app" in sys.modules:
         del sys.modules["app"]
     module = importlib.import_module("app")
     return importlib.reload(module)
 
 
-def test_health_reports_runtime_without_ollama(monkeypatch):
+def test_health_reports_runtime_without_ollama_and_without_forcing_model_load(monkeypatch):
     module = load_app_module(monkeypatch)
     client = TestClient(module.app)
 
@@ -30,12 +36,46 @@ def test_health_reports_runtime_without_ollama(monkeypatch):
     payload = response.json()
     assert payload["ok"] is True
     assert payload["device"] == "cpu"
+    assert payload["model_loaded"] is False
+    assert payload["model_disabled"] is True
+    assert payload["summarize_available"] is False
+    assert "disabled" in payload["model_load_error"].lower()
     assert payload["ollama_configured"] is False
     assert payload["ollama_model"] is None
     assert payload["ollama_installed_models"] == []
     assert payload["ollama_loaded_models"] == []
 
 
+def test_summarize_returns_503_with_explicit_reason_when_model_loading_is_disabled(monkeypatch):
+    module = load_app_module(monkeypatch)
+    client = TestClient(module.app)
+
+    response = client.post("/summarize", json={"text": "Platform engineering role with APIs and Python experience."})
+
+    assert response.status_code == 503
+    payload = response.json()
+    assert "disabled" in payload["detail"].lower()
+
+
+def test_health_reports_ollama_unreachable_when_configured_but_not_available(monkeypatch):
+    module = load_app_module(monkeypatch, ollama_model="qwen2.5:7b")
+
+    def boom(path: str):
+        raise OSError("connection refused")
+
+    monkeypatch.setattr(module, "_ollama_json", boom)
+    client = TestClient(module.app)
+
+    response = client.get("/health")
+
+    assert response.status_code == 200
+    payload = response.json()
+    assert payload["ollama_configured"] is True
+    assert payload["ollama_reachable"] is False
+    assert payload["ollama_model"] == "qwen2.5:7b"
+    assert payload["ollama_model_available"] is False
+
+
 def test_classify_block_returns_structured_json(monkeypatch):
     module = load_app_module(monkeypatch)