From 7817f92d1952f78cec8c17711032c0c8ef5cec50 Mon Sep 17 00:00:00 2001
From: Sameer Kankute <sameer@berri.ai>
Date: Fri, 3 Apr 2026 01:46:40 +0530
Subject: [PATCH] fix team routing

---
 litellm/proxy/health_check.py |  2 +-
 litellm/router.py             | 11 ++++++++---
 litellm/utils.py              | 35 +++++++++++++++++++++++++++++++----
 3 files changed, 40 insertions(+), 8 deletions(-)

diff --git a/litellm/proxy/health_check.py b/litellm/proxy/health_check.py
index bd2f4734fb2..5a98daec5f9 100644
--- a/litellm/proxy/health_check.py
+++ b/litellm/proxy/health_check.py
@@ -348,7 +348,7 @@ async def perform_health_check(
                     source,
                     cycle_id,
                 )
-            return [], []
+            return [], [], {}
 
     cycle_start_time = time.monotonic()
     requested_model_count = len(model_list)
diff --git a/litellm/router.py b/litellm/router.py
index 26daf2b1e95..f390d2c70fc 100644
--- a/litellm/router.py
+++ b/litellm/router.py
@@ -5315,11 +5315,16 @@ class Router:
             e,
             (litellm.ContextWindowExceededError, litellm.ContentPolicyViolationError),
         )
-        all_deployments = self._get_all_deployments(model_name=original_model_group)
+        _request_team_id: Optional[str] = (
+            kwargs.get("metadata", {}) or {}
+        ).get("user_api_key_team_id")
+        all_deployments = self._get_all_deployments(
+            model_name=original_model_group, team_id=_request_team_id
+        )
         _order_set: set = {
-            d.get("litellm_params", {}).get("order")
+            litellm.utils._get_deployment_order(d)
             for d in all_deployments
-            if d.get("litellm_params", {}).get("order") is not None
+            if litellm.utils._get_deployment_order(d) is not None
         }
         order_values: list = sorted(_order_set)
         if len(order_values) > 1 and not _skip_order_fallback:
diff --git a/litellm/utils.py b/litellm/utils.py
index 088ee07d630..c68a935c640 100644
--- a/litellm/utils.py
+++ b/litellm/utils.py
@@ -4866,12 +4866,39 @@ def calculate_max_parallel_requests(
     return None
 
 
-def _get_order_filtered_deployments(healthy_deployments: List[Dict]) -> List:
+def _get_deployment_order(deployment: Dict) -> Optional[int]:
+    """
+    Returns the routing order for a deployment.
+
+    Checks litellm_params first (static config), then model_info (dynamic/team
+    models added via API where order lives in model_info, not litellm_params).
+    """
+    order = deployment.get("litellm_params", {}).get("order")
+    if order is None:
+        order = deployment.get("model_info", {}).get("order")
+    return order
+
+
+def _get_order_filtered_deployments(
+    healthy_deployments: List[Dict], target_order: Optional[int] = None
+) -> List:
+    if target_order is not None:
+        filtered = [
+            d
+            for d in healthy_deployments
+            if _get_deployment_order(d) == target_order
+        ]
+        if filtered:
+            return filtered
+        # target_order doesn't match any deployment (e.g., external fallback model) — return all
+        return healthy_deployments
+
+    # Default: pick min order group
     min_order = min(
         (
-            deployment["litellm_params"]["order"]
+            _get_deployment_order(deployment)
             for deployment in healthy_deployments
-            if "order" in deployment["litellm_params"]
+            if _get_deployment_order(deployment) is not None
         ),
         default=None,
     )
@@ -4880,7 +4907,7 @@ def _get_order_filtered_deployments(healthy_deployments: List[Dict]) -> List:
         filtered_deployments = [
             deployment
             for deployment in healthy_deployments
-            if deployment["litellm_params"].get("order") == min_order
+            if _get_deployment_order(deployment) == min_order
         ]
 
         return filtered_deployments