From d74b21a3469535eb1a598246b2ad53a83fd331de Mon Sep 17 00:00:00 2001
From: phact <estevezsebastian@gmail.com>
Date: Tue, 14 Oct 2025 01:39:15 -0400
Subject: [PATCH] onboarding in conftest

---
 tests/conftest.py                       | 49 +++++++++++++++++++++++++
 tests/integration/test_api_endpoints.py | 49 +++++--------------------
 2 files changed, 59 insertions(+), 39 deletions(-)

diff --git a/tests/conftest.py b/tests/conftest.py
index 7c2ffc1d..9498d6e9 100644
--- a/tests/conftest.py
+++ b/tests/conftest.py
@@ -20,6 +20,55 @@ from src.session_manager import SessionManager
 from src.main import generate_jwt_keys
 
 
+@pytest_asyncio.fixture(scope="session")
+async def onboard_system():
+    """Perform initial onboarding once for all tests in the session.
+
+    This ensures the OpenRAG config is marked as edited and properly initialized
+    so that tests can use the /settings endpoint.
+    """
+    from pathlib import Path
+
+    # Delete any existing config to ensure clean onboarding
+    config_file = Path("config/config.yaml")
+    if config_file.exists():
+        config_file.unlink()
+
+    # Initialize clients
+    await clients.initialize()
+
+    # Create app and perform onboarding via API
+    from src.main import create_app, startup_tasks
+    import httpx
+
+    app = await create_app()
+    await startup_tasks(app.state.services)
+
+    transport = httpx.ASGITransport(app=app)
+    async with httpx.AsyncClient(transport=transport, base_url="http://testserver") as client:
+        onboarding_payload = {
+            "model_provider": "openai",
+            "embedding_model": "text-embedding-3-small",
+            "llm_model": "gpt-4o-mini",
+            "endpoint": "https://api.openai.com/v1",
+            "sample_data": False,
+        }
+        resp = await client.post("/onboarding", json=onboarding_payload)
+        if resp.status_code not in (200, 204):
+            # If it fails, it might already be onboarded, which is fine
+            print(f"[DEBUG] Onboarding returned {resp.status_code}: {resp.text}")
+        else:
+            print(f"[DEBUG] Session onboarding completed successfully")
+
+    yield
+
+    # Cleanup after all tests
+    try:
+        await clients.close()
+    except Exception:
+        pass
+
+
 @pytest.fixture(scope="session")
 def event_loop():
     """Create an instance of the default event loop for the test session."""
diff --git a/tests/integration/test_api_endpoints.py b/tests/integration/test_api_endpoints.py
index d49cf224..0e7942a9 100644
--- a/tests/integration/test_api_endpoints.py
+++ b/tests/integration/test_api_endpoints.py
@@ -38,45 +38,6 @@ def dump_docker_logs(container_name_pattern: str = "langflow", tail: int = 100):
         print(f"[DEBUG] Failed to fetch docker logs for {container_name_pattern}: {e}")
 
 
-async def perform_onboarding(client: httpx.AsyncClient, embedding_model: str = "text-embedding-3-small", llm_model: str = "gpt-4o-mini"):
-    """Perform onboarding configuration to properly initialize the system.
-
-    This should be called once at the beginning of test session, or per-test if config is reset.
-    It configures the embedding model, LLM model, and initializes the OpenSearch index.
-
-    If onboarding has already been performed (config.edited=True), this will use the /settings
-    endpoint instead to update the configuration.
-    """
-    onboarding_payload = {
-        "model_provider": "openai",
-        "embedding_model": embedding_model,
-        "llm_model": llm_model,
-        "endpoint": "https://api.openai.com/v1",
-        "sample_data": False,
-    }
-    onboarding_resp = await client.post("/onboarding", json=onboarding_payload)
-
-    # If onboarding fails because config is already edited, use /settings endpoint instead
-    if onboarding_resp.status_code == 403:
-        print(f"[DEBUG] Config already onboarded, using /settings endpoint instead")
-        settings_payload = {
-            "embedding_model": embedding_model,
-            "llm_model": llm_model,
-        }
-        settings_resp = await client.post("/settings", json=settings_payload)
-        if settings_resp.status_code not in (200, 204):
-            raise AssertionError(
-                f"Settings update failed: {settings_resp.status_code} {settings_resp.text}"
-            )
-        print(f"[DEBUG] Settings updated: embedding_model={embedding_model}, llm_model={llm_model}")
-    elif onboarding_resp.status_code not in (200, 204):
-        raise AssertionError(
-            f"Onboarding failed: {onboarding_resp.status_code} {onboarding_resp.text}"
-        )
-    else:
-        print(f"[DEBUG] Onboarding completed: embedding_model={embedding_model}, llm_model={llm_model}")
-
-
 async def wait_for_service_ready(client: httpx.AsyncClient, timeout_s: float = 30.0):
     """Poll existing endpoints until the app and OpenSearch are ready.
 
@@ -433,6 +394,16 @@ async def test_langflow_chat_and_nudges_endpoints():
         async with httpx.AsyncClient(transport=transport, base_url="http://testserver") as client:
             await wait_for_service_ready(client)
 
+            # Ensure embedding model is configured via settings
+            resp = await client.post(
+                "/settings",
+                json={
+                    "embedding_model": "text-embedding-3-small",
+                    "llm_model": "gpt-4o-mini",
+                },
+            )
+            assert resp.status_code == 200, resp.text
+
             warmup_file = Path("./nudges_seed.md")
             warmup_file.write_text(
                 "The user may care about different fruits including apples, hardy kiwi, and bananas"