fix: use in-memory checkpointer for API tests to avoid database dependency

- Replace PostgreSQL AsyncPostgresSaver with MemorySaver in test fixtures
- Create test_app fixture with test_lifespan using in-memory dependencies
- Mock LLM model to avoid requiring API keys during tests
- Update test_api.py to use test_app fixture instead of production app
- Fix test_root endpoint URL to match root_path configuration
- All API tests now pass without requiring external services (0.06s vs 30s+ timeout)

This change allows tests to run in isolation without PostgreSQL or API keys,
making them faster and more reliable for CI/CD and local development.

tests/test_api.py

@@ -1,16 +1,16 @@
 import pytest
 from fastapi.testclient import TestClient
-from src.naliiabotapi.main import app
 
 
 @pytest.fixture
-def client():
-    with TestClient(app) as client:
+def client(test_app):
+    """Create a test client using the test app with in-memory dependencies."""
+    with TestClient(test_app) as client:
         yield client
 
 
 def test_root(client):
-    response = client.get("/")
+    response = client.get("/agent/api")
 
     assert response.status_code == 200
 
@@ -22,7 +22,9 @@ def test_health(client):
 
 
 def test_chat(client):
-    response = client.post("/chat", json={"messages": "Hello, how are you?"})
+    response = client.post(
+        "/chat", json={"messages": "Hello, how are you?"}
+    )
 
     assert response.status_code == 200
     assert response.json() is not None