fix(simple_shuffle): honor weights when first deployment has none

litellm/router_strategy/simple_shuffle.py

@@ -41,9 +41,9 @@ def simple_shuffle(
 
     ############## Check if 'weight' or 'rpm' or 'tpm' param set for a weighted pick #################
     for weight_by in ["weight", "rpm", "tpm"]:
-        weight = healthy_deployments[0].get("litellm_params").get(weight_by, None)
-        if weight is not None:
-            weights = [m["litellm_params"].get(weight_by, 0) for m in healthy_deployments]
+        raw_weights = [m["litellm_params"].get(weight_by, None) for m in healthy_deployments]
+        if any(weight is not None for weight in raw_weights):
+            weights = [weight if weight is not None else 0 for weight in raw_weights]
             verbose_router_logger.debug(f"\nweight {weights}")
             total_weight = sum(weights)
             if total_weight <= 0:

tests/test_litellm/router_strategy/test_simple_shuffle.py

@@ -0,0 +1,34 @@
+from collections import Counter
+from unittest.mock import MagicMock
+
+from litellm.router_strategy.simple_shuffle import simple_shuffle
+
+
+def _deployment(deployment_id: str, **litellm_params) -> dict:
+    return {
+        "model_name": "test-model",
+        "litellm_params": {"model": "gpt-4o", **litellm_params},
+        "model_info": {"id": deployment_id},
+    }
+
+
+def test_weights_respected_when_first_deployment_has_no_weight():
+    """Regression for #33329: `simple_shuffle` decided whether weighted routing
+    was enabled by inspecting only the first healthy deployment. When the first
+    deployment had no `weight`, every deployment's weight was ignored and the
+    pick fell back to uniform random. The weighted pick must be driven by any
+    deployment declaring a weight, regardless of ordering.
+    """
+    healthy_deployments = [
+        _deployment("A"),
+        _deployment("B", weight=1),
+        _deployment("C", weight=99),
+    ]
+
+    counts: Counter = Counter()
+    for _ in range(2000):
+        deployment = simple_shuffle(MagicMock(), healthy_deployments, "test-model")
+        counts[deployment["model_info"]["id"]] += 1
+
+    assert counts["A"] == 0
+    assert counts["C"] > counts["B"] * 2