fix(test): gate models.dev background refresh out of tests (main) (#13597)

fix(test): gate models.dev background refresh out of tests

Integration tests failed twice in nightly run 27260425158 with pyleak
EventLoopBlockError - first Integration Tests 3.14, then 3.12 on the
rerun, each time in a different test. The blocking stack points at
refresh_models_dev_periodically: every app boot unconditionally starts
a lifespan task that immediately fetches https://models.dev/api.json,
so the request lands mid-test in whatever test happens to be running.
Under pyleak's asyncio debug instrumentation the fetch blocked the loop
0.797s against a 0.2s threshold. Whichever test draws the short straw
flakes - which is why it looked transient and moved between versions.

Add a LANGFLOW_MODELS_DEV_REFRESH env gate (default unchanged: enabled)
and disable it session-wide in the backend test conftest. Tests fall
back to the bundled static model lists, which is also deterministic.

Verified: with the gate set, app boot makes zero models.dev requests;
the previously failing integration test passes.

Author: erichare

.secrets.baseline

@@ -2837,15 +2837,15 @@
         "filename": "src/backend/tests/conftest.py",
         "hashed_secret": "8bb6118f8fd6935ad0876a3be34a717d32708ffd",
         "is_verified": false,
-        "line_number": 490,
+        "line_number": 515,
         "is_secret": false
       },
       {
         "type": "Secret Keyword",
         "filename": "src/backend/tests/conftest.py",
         "hashed_secret": "61fbb5a12cd7b1f1fe1624120089efc0cd299e43",
         "is_verified": false,
-        "line_number": 700,
+        "line_number": 725,
         "is_secret": false
       }
     ],
@@ -9287,5 +9287,5 @@
       }
     ]
   },
-  "generated_at": "2026-06-08T16:38:06Z"
+  "generated_at": "2026-06-10T08:36:23Z"
 }

src/backend/base/langflow/main.py

@@ -479,7 +479,14 @@ async def refresh_models_dev_periodically() -> None:
 
                     await asyncio.sleep(refresh_interval_seconds)
 
-            models_dev_refresh_task = asyncio.create_task(refresh_models_dev_periodically())
+            # LANGFLOW_MODELS_DEV_REFRESH=false disables the live models.dev
+            # fetch. Tests set this: the startup fetch otherwise fires from a
+            # background task during whatever test is running, hitting the
+            # network and tripping event-loop-block detectors (pyleak).
+            if os.getenv("LANGFLOW_MODELS_DEV_REFRESH", "true").lower() not in ("false", "0", "no"):
+                models_dev_refresh_task = asyncio.create_task(refresh_models_dev_periodically())
+            else:
+                await logger.adebug("models.dev refresh disabled via LANGFLOW_MODELS_DEV_REFRESH")
 
             # v1 and project MCP server context managers
             from langflow.api.v1.mcp import start_streamable_http_manager

src/backend/tests/conftest.py

@@ -50,6 +50,21 @@ def disable_rate_limiting():
     os.environ.pop("LANGFLOW_RATE_LIMIT_ENABLED", None)
 
 
+@pytest.fixture(scope="session", autouse=True)
+def disable_models_dev_refresh():
+    """Keep the models.dev background refresh out of tests.
+
+    Every app boot otherwise launches a lifespan task that fetches
+    https://models.dev/api.json mid-test, which both hits the network and
+    trips event-loop-block detectors (pyleak) in whatever test happens to be
+    running when the request lands. The bundled static model lists are used
+    instead, which is also deterministic.
+    """
+    os.environ["LANGFLOW_MODELS_DEV_REFRESH"] = "false"
+    yield
+    os.environ.pop("LANGFLOW_MODELS_DEV_REFRESH", None)
+
+
 # TODO: Revert this to True once bb.functions[func].can_block_in("http/client.py", "_safe_read") is fixed
 @pytest.fixture(autouse=False)
 def blockbuster(request):