Merge pull request #25551 from BerriAI/litellm_backport_patch4_to_stable_1_82_3

[Fix] Backport v1.82.3-stable.patch.4 commits to stable branch

docs/my-website/docs/proxy/config_settings.md

@@ -801,6 +801,7 @@ router_settings:
 | LITELLM_OTEL_INTEGRATION_ENABLE_EVENTS | Optionally enable semantic logs for OTEL
 | LITELLM_OTEL_INTEGRATION_ENABLE_METRICS | Optionally enable emantic metrics for OTEL
 | LITELLM_ENABLE_PYROSCOPE | If true, enables Pyroscope CPU profiling. Profiles are sent to PYROSCOPE_SERVER_ADDRESS. Off by default. See [Pyroscope profiling](/proxy/pyroscope_profiling).
+| LITELLM_ENABLE_TEAM_STALE_ALIAS_BYPASS | When `true`, if a team's legacy `model_aliases` entry maps a public model name to an internal `model_name_<team_id>_<uuid>` deployment, pre-call handling can skip that rewrite when team-scoped sibling deployments exist for the public name—so load balancing / `order` apply across siblings. Default is `false` for backwards compatibility. See [Team-scoped models and legacy aliases](./load_balancing#team-scoped-models-and-legacy-model_aliases). When stale aliases are detected and this flag is off, the proxy may log a one-time warning.
 | PYROSCOPE_APP_NAME | Application name reported to Pyroscope. Required when LITELLM_ENABLE_PYROSCOPE is true. No default.
 | PYROSCOPE_SERVER_ADDRESS | Pyroscope server URL to send profiles to. Required when LITELLM_ENABLE_PYROSCOPE is true. No default.
 | PYROSCOPE_SAMPLE_RATE | Optional. Sample rate for Pyroscope profiling (integer). No default; when unset, the pyroscope-io library default is used.

docs/my-website/docs/proxy/load_balancing.md

@@ -336,6 +336,21 @@ The `order` parameter requires `enable_pre_call_checks: true` in `router_setting
 
 If `order=1` deployment is unavailable (e.g., rate-limited), the router falls back to `order=2` deployments.
 
+### Team-scoped models and legacy `model_aliases` {#team-scoped-models-and-legacy-model_aliases}
+
+Team-scoped deployments are identified by `model_info.team_id` and `model_info.team_public_model_name`. Requests should use the **public** model name; the router resolves all sibling deployments (same public name, different `api_base` / `order`, etc.) for routing, failover, and deployment `order`.
+
+For router internals: when a `team_id` is in scope, optimized lookups key off `(team_id, team_public_model_name)`. If code passes an internal deployment id (e.g. `model_name_<team_id>_<uuid>`) instead of the public name, routing still works via the usual deployment-name paths, but the team-specific fast path applies only to the public name.
+
+**Legacy teams:** Older proxy versions could persist `model_aliases` on the team row mapping a public name to a single internal deployment id (`model_name_<team_id>_<uuid>`). On each request, pre-call logic may still rewrite `model` to that internal name **before** routing, which collapses to one deployment and can make newer sibling deployments unreachable.
+
+**Migration options:**
+
+1. **Recommended for upgrades:** Set environment variable `LITELLM_ENABLE_TEAM_STALE_ALIAS_BYPASS=true` so that when sibling team deployments exist for the public name, the stale alias rewrite is skipped and team-scoped routing (including `order` and failover) applies. See the [Environment variables](./config_settings) table in the proxy settings doc.
+2. **Data cleanup:** Remove obsolete `model_aliases` entries for team public names from the team record in the database so only `team_public_model_name` + team model list drive access.
+
+If a stale alias is detected and the bypass is **not** enabled, the proxy may emit a **one-time** warning in logs explaining that sibling deployments may be unreachable until the flag is set or aliases are cleaned up.
+
 ### When You'll See Load Balancing in Action
 
 **Immediate Effects:**

litellm/proxy/litellm_pre_call_utils.py

@@ -1,6 +1,7 @@
 import asyncio
 import copy
 import time
+from collections import OrderedDict
 from typing import TYPE_CHECKING, Any, Dict, List, Optional, Union
 
 from fastapi import Request
@@ -26,6 +27,7 @@ _SPECIAL_HEADERS_CACHE = frozenset(
     v.value.lower() for v in SpecialHeaders._member_map_.values()
 )
 from litellm.router import Router
+from litellm.secret_managers.main import get_secret_bool
 from litellm.types.llms.anthropic import ANTHROPIC_API_HEADERS
 from litellm.types.services import ServiceTypes
 from litellm.types.utils import (
@@ -36,6 +38,9 @@ from litellm.types.utils import (
 )
 
 service_logger_obj = ServiceLogging()  # used for tracking latency on OTEL
+# Bounded dedup for stale-alias warnings (FIFO eviction when over cap).
+_MAX_STALE_ALIAS_WARNING_KEYS = 10_000
+_STALE_TEAM_ALIAS_WARNING_KEYS: OrderedDict[str, None] = OrderedDict()
 
 
 if TYPE_CHECKING:
@@ -1295,6 +1300,10 @@ def _update_model_if_team_alias_exists(
             "gpt-4o": "gpt-4o-team-1"
         }
         - requested_model = "gpt-4o-team-1"
+
+    Note: model_aliases for team models are deprecated. This function only applies
+    to legacy non-team-scoped aliases. Team-scoped deployments use team_public_model_name
+    and are resolved via map_team_model in route_llm_request.
     """
     _model = data.get("model")
     if (
@@ -1302,7 +1311,48 @@ def _update_model_if_team_alias_exists(
         and user_api_key_dict.team_model_aliases
         and _model in user_api_key_dict.team_model_aliases
     ):
-        data["model"] = user_api_key_dict.team_model_aliases[_model]
+        from litellm.proxy.proxy_server import llm_router
+
+        # Skip alias rewrite if this model resolves to team-specific deployments
+        # (team models use team_public_model_name, not model_aliases)
+        aliased_target = user_api_key_dict.team_model_aliases[_model]
+
+        # Optional bypass for stale aliases from pre-PR deployments:
+        # only enabled via feature flag to preserve backwards compatibility.
+        enable_stale_alias_bypass = get_secret_bool(
+            "LITELLM_ENABLE_TEAM_STALE_ALIAS_BYPASS", False
+        )
+        # Check if the alias points to a team-scoped UUID name
+        # (format: "model_name_{team_id}_{uuid}")
+        is_stale_team_alias = aliased_target.startswith(
+            f"model_name_{user_api_key_dict.team_id}_"
+        )
+        if is_stale_team_alias and llm_router:
+            # This is a stale alias from pre-PR deployments.
+            # Check if current team deployments exist for the public name.
+            key = (user_api_key_dict.team_id, _model)
+            if key in llm_router.team_model_to_deployment_indices:
+                if enable_stale_alias_bypass:
+                    # Team deployments exist; skip stale alias
+                    return
+                warning_key = f"{user_api_key_dict.team_id}:{_model}:{aliased_target}"
+                if warning_key not in _STALE_TEAM_ALIAS_WARNING_KEYS:
+                    _STALE_TEAM_ALIAS_WARNING_KEYS[warning_key] = None
+                    while (
+                        len(_STALE_TEAM_ALIAS_WARNING_KEYS)
+                        > _MAX_STALE_ALIAS_WARNING_KEYS
+                    ):
+                        _STALE_TEAM_ALIAS_WARNING_KEYS.popitem(last=False)
+                    verbose_proxy_logger.warning(
+                        "Stale team model alias detected for model='%s', team_id='%s'. "
+                        "New sibling deployments may be unreachable. "
+                        "Set LITELLM_ENABLE_TEAM_STALE_ALIAS_BYPASS=true to enable "
+                        "team-scoped sibling routing.",
+                        _model,
+                        user_api_key_dict.team_id,
+                    )
+
+        data["model"] = aliased_target
     return
 
 

litellm/proxy/management_endpoints/model_management_endpoints.py

@@ -13,13 +13,13 @@ model/{model_id}/update - PATCH endpoint for model update.
 import asyncio
 import datetime
 import json
-from litellm._uuid import uuid
 from typing import Dict, List, Literal, Optional, Tuple, Union, cast
 
 from fastapi import APIRouter, Depends, HTTPException, Request, status
 from pydantic import BaseModel, ConfigDict, Field
 
 from litellm._logging import verbose_proxy_logger
+from litellm._uuid import uuid
 from litellm.constants import LITELLM_PROXY_ADMIN_NAME
 from litellm.proxy._types import (
     CommonProxyErrors,
@@ -32,7 +32,7 @@ from litellm.proxy._types import (
     ProxyErrorTypes,
     ProxyException,
     TeamModelAddRequest,
-    UpdateTeamRequest,
+    TeamModelDeleteRequest,
     UserAPIKeyAuth,
 )
 from litellm.proxy.auth.user_api_key_auth import user_api_key_auth
@@ -40,7 +40,7 @@ from litellm.proxy.common_utils.encrypt_decrypt_utils import encrypt_value_helpe
 from litellm.proxy.management_endpoints.common_utils import _is_user_team_admin
 from litellm.proxy.management_endpoints.team_endpoints import (
     team_model_add,
-    update_team,
+    team_model_delete,
 )
 from litellm.proxy.management_helpers.audit_logs import create_object_audit_log
 from litellm.proxy.utils import PrismaClient
@@ -324,7 +324,7 @@ async def _add_team_model_to_db(
 
     - generate a unique 'model_name' for the model (e.g. 'model_name_{team_id}_{uuid})
     - store the model in the db with the unique 'model_name'
-    - store a team model alias mapping {"model_name": "model_name_{team_id}_{uuid}"}
+    - add the public model name to the team's allowed models list
     """
     _team_id = model_params.model_info.team_id
     if _team_id is None:
@@ -344,25 +344,15 @@ async def _add_team_model_to_db(
         prisma_client=prisma_client,
     )
 
-    ## CREATE MODEL ALIAS IN DB ##
-    await update_team(
-        data=UpdateTeamRequest(
-            team_id=_team_id,
-            model_aliases={original_model_name: unique_model_name},
-        ),
-        user_api_key_dict=user_api_key_dict,
-        http_request=Request(scope={"type": "http"}),
-    )
-
-    # add model to team object
-    await team_model_add(
-        data=TeamModelAddRequest(
-            team_id=_team_id,
-            models=[original_model_name],
-        ),
-        http_request=Request(scope={"type": "http"}),
-        user_api_key_dict=user_api_key_dict,
-    )
+    if original_model_name:
+        await team_model_add(
+            data=TeamModelAddRequest(
+                team_id=_team_id,
+                models=[original_model_name],
+            ),
+            http_request=Request(scope={"type": "http"}),
+            user_api_key_dict=user_api_key_dict,
+        )
 
     return model_response
 
@@ -428,6 +418,7 @@ async def _update_team_model_in_db(
             db_model=db_model,
             patch_data=patch_data,
             user_api_key_dict=user_api_key_dict,
+            prisma_client=prisma_client,
         )
 
     return update_db_model(db_model=db_model, updated_patch=patch_data)
@@ -453,19 +444,10 @@ async def _setup_new_team_model_assignment(
     patch_data: updateDeployment,
     user_api_key_dict: UserAPIKeyAuth,
 ) -> None:
-    """Set up a new team model with unique name, alias, and team membership."""
+    """Set up a new team model with unique name and team membership."""
     unique_model_name = f"model_name_{team_id}_{uuid.uuid4()}"
     patch_data.model_name = unique_model_name
 
-    await update_team(
-        data=UpdateTeamRequest(
-            team_id=team_id,
-            model_aliases={public_model_name: unique_model_name},
-        ),
-        user_api_key_dict=user_api_key_dict,
-        http_request=Request(scope={"type": "http"}),
-    )
-
     await team_model_add(
         data=TeamModelAddRequest(
             team_id=team_id,
@@ -482,24 +464,95 @@ async def _update_existing_team_model_assignment(
     db_model: Deployment,
     patch_data: updateDeployment,
     user_api_key_dict: UserAPIKeyAuth,
+    prisma_client: Optional[PrismaClient],
 ) -> None:
-    """Update an existing team model if the public name changed."""
+    """Update an existing team model if the public name changed.
+
+    Note on DB scan: Prisma's JSON filtering does not support compound AND conditions
+    across multiple JSON paths, so we fetch all deployments for the team and filter
+    team_public_model_name in Python. For teams with many deployments this scan grows
+    linearly; if team deployment counts become large this should be revisited.
+    """
+
+    def _get_team_public_model_name(
+        model_info: Optional[Union[dict, str]]
+    ) -> Optional[str]:
+        if isinstance(model_info, dict):
+            value = model_info.get("team_public_model_name")
+            return value if isinstance(value, str) else None
+        if isinstance(model_info, str):
+            try:
+                parsed = json.loads(model_info)
+            except (TypeError, ValueError):
+                return None
+            if isinstance(parsed, dict):
+                value = parsed.get("team_public_model_name")
+                return value if isinstance(value, str) else None
+        return None
+
     old_public_name = (
         db_model.model_info.team_public_model_name if db_model.model_info else None
     )
 
-    # Update alias only if public name changed
     if old_public_name and public_model_name != old_public_name:
-        await update_team(
-            data=UpdateTeamRequest(
+        # Clear user-supplied public name from patch before any early return so the
+        # caller does not overwrite the internal UUID-based model_name in the DB.
+        patch_data.model_name = None
+        if prisma_client is None:
+            verbose_proxy_logger.warning(
+                "prisma_client not initialized; skipping public name update entirely to avoid orphaned entries"
+            )
+            return
+
+        response = await prisma_client.db.litellm_proxymodeltable.find_many(
+            where={
+                "model_info": {
+                    "path": ["team_id"],
+                    "equals": team_id,
+                }
+            }
+        )
+        if not response:
+            other_deployments_with_old_name = []
+        else:
+            other_deployments_with_old_name = [
+                d
+                for d in response
+                if d.model_name != db_model.model_name
+                and _get_team_public_model_name(d.model_info) == old_public_name
+            ]
+
+        # Add new name first, then delete old name to prevent access loss on partial failure
+        await team_model_add(
+            data=TeamModelAddRequest(
                 team_id=team_id,
-                model_aliases={public_model_name: db_model.model_name},
+                models=[public_model_name],
             ),
-            user_api_key_dict=user_api_key_dict,
             http_request=Request(scope={"type": "http"}),
+            user_api_key_dict=user_api_key_dict,
+        )
+
+        if not other_deployments_with_old_name:
+            await team_model_delete(
+                data=TeamModelDeleteRequest(
+                    team_id=team_id,
+                    models=[old_public_name],
+                ),
+                http_request=Request(scope={"type": "http"}),
+                user_api_key_dict=user_api_key_dict,
+            )
+    elif not old_public_name and public_model_name:
+        # First-time assignment of public name on an existing team deployment:
+        # ensure the team's models list is updated so team routing can resolve it.
+        await team_model_add(
+            data=TeamModelAddRequest(
+                team_id=team_id,
+                models=[public_model_name],
+            ),
+            http_request=Request(scope={"type": "http"}),
+            user_api_key_dict=user_api_key_dict,
         )
 
-    # Keep existing unique model_name
     patch_data.model_name = None
 
 

litellm/router.py

@@ -466,6 +466,8 @@ class Router:
         # Initialize model name to deployment indices mapping for O(1) lookups
         # Maps model_name -> list of indices in model_list
         self.model_name_to_deployment_indices: Dict[str, List[int]] = {}
+        # Maps (team_id, team_public_model_name) -> list of indices in model_list
+        self.team_model_to_deployment_indices: Dict[Tuple[str, str], List[int]] = {}
 
         if model_list is not None:
             # set_model_list will build indices automatically
@@ -6757,6 +6759,7 @@ class Router:
         self.model_list = []
         self.model_id_to_deployment_index_map = {}  # Reset the index
         self.model_name_to_deployment_indices = {}  # Reset the model_name index
+        self.team_model_to_deployment_indices = {}  # Reset the team_model index
         self._invalidate_model_group_info_cache()
         self._invalidate_access_groups_cache()
         # we add api_base/api_key each model so load balancing between azure/gpt on api_base1 and api_base2 works
@@ -7054,16 +7057,17 @@ class Router:
 
         # Update model_name_to_deployment_indices
         for model_name, indices in list(self.model_name_to_deployment_indices.items()):
-            # Remove the deleted index
-            if removal_idx in indices:
-                indices.remove(removal_idx)
-
-            # Decrement all indices greater than removal_idx
+            # Build new list without mutating the original
             updated_indices = []
             for idx in indices:
-                if idx > removal_idx:
+                if idx == removal_idx:
+                    # Skip the removed index
+                    continue
+                elif idx > removal_idx:
+                    # Decrement indices after removal
                     updated_indices.append(idx - 1)
                 else:
+                    # Keep indices before removal unchanged
                     updated_indices.append(idx)
 
             # Update or remove the entry
@@ -7072,6 +7076,45 @@ class Router:
             else:
                 del self.model_name_to_deployment_indices[model_name]
 
+        # Update team_model_to_deployment_indices
+        for key, indices in list(self.team_model_to_deployment_indices.items()):
+            # Build new list without mutating the original
+            updated_indices = []
+            for idx in indices:
+                if idx == removal_idx:
+                    # Skip the removed index
+                    continue
+                elif idx > removal_idx:
+                    # Decrement indices after removal
+                    updated_indices.append(idx - 1)
+                else:
+                    # Keep indices before removal unchanged
+                    updated_indices.append(idx)
+
+            # Update or remove the entry
+            if len(updated_indices) > 0:
+                self.team_model_to_deployment_indices[key] = updated_indices
+            else:
+                del self.team_model_to_deployment_indices[key]
+
+    def _update_team_model_index(self, model: dict, idx: int) -> None:
+        """
+        Helper to update team_model_to_deployment_indices for a single deployment.
+
+        Parameters:
+        - model: dict - the deployment to index
+        - idx: int - the index in model_list
+        """
+        team_id = (model.get("model_info") or {}).get("team_id")
+        team_public_model_name = (model.get("model_info") or {}).get(
+            "team_public_model_name"
+        )
+        if team_id and team_public_model_name:
+            key = (team_id, team_public_model_name)
+            if key not in self.team_model_to_deployment_indices:
+                self.team_model_to_deployment_indices[key] = []
+            self.team_model_to_deployment_indices[key].append(idx)
+
     def _add_model_to_list_and_index_map(
         self, model: dict, model_id: Optional[str] = None
     ) -> None:
@@ -7100,6 +7143,9 @@ class Router:
                 self.model_name_to_deployment_indices[model_name] = []
             self.model_name_to_deployment_indices[model_name].append(idx)
 
+        # Update team_model index for O(1) team-scoped lookup
+        self._update_team_model_index(model, idx)
+
     def upsert_deployment(self, deployment: Deployment) -> Optional[Deployment]:
         """
         Add or update deployment
@@ -7118,7 +7164,10 @@ class Router:
             )
             if _deployment_on_router is not None:
                 # deployment with this model_id exists on the router
-                if deployment.litellm_params == _deployment_on_router.litellm_params:
+                if (
+                    deployment.litellm_params == _deployment_on_router.litellm_params
+                    and deployment.model_info == _deployment_on_router.model_info
+                ):
                     # No need to update
                     return None
 
@@ -7930,6 +7979,7 @@ class Router:
         instead of O(n) linear scan through the entire model_list.
         """
         self.model_name_to_deployment_indices.clear()
+        self.team_model_to_deployment_indices.clear()
 
         for idx, model in enumerate(model_list):
             model_name = model.get("model_name")
@@ -7938,6 +7988,8 @@ class Router:
                     self.model_name_to_deployment_indices[model_name] = []
                 self.model_name_to_deployment_indices[model_name].append(idx)
 
+            self._update_team_model_index(model, idx)
+
     def _build_model_id_to_deployment_index_map(self, model_list: list):
         """
         Build model index from model list to enable O(1) lookups immediately.
@@ -8070,20 +8122,23 @@ class Router:
 
     def map_team_model(self, team_model_name: str, team_id: str) -> Optional[str]:
         """
-        Map a team model name to a team-specific model name.
+        Check if team_model_name resolves to team-specific deployments.
+
+        Returns the public model name (unchanged) so the router can find all
+        sibling deployments via team_id filtering, instead of collapsing to a
+        single internal model_name.
 
         Returns:
-        - deployment id: str - the deployment id of the team-specific model
-        - None: if no team-specific model name is found
+        - str: the team_model_name if team deployments exist for this team
+        - None: if no team-specific model is found
         """
         models = self.get_model_list(model_name=team_model_name, team_id=team_id)
         if not models:
             return None
         for model in models:
             if model.get("model_info", {}).get("team_id") == team_id:
-                return model.get("model_name")
+                return team_model_name
 
-        ## wildcard models
         return None
 
     def should_include_deployment(
@@ -8094,12 +8149,19 @@ class Router:
         """
         if (
             team_id is not None
-            and model["model_info"].get("team_id") == team_id
-            and model_name == model["model_info"].get("team_public_model_name")
+            and (model.get("model_info") or {}).get("team_id") == team_id
+            and model_name
+            == (model.get("model_info") or {}).get("team_public_model_name")
         ):
             return True
         elif model_name is not None and model["model_name"] == model_name:
-            return True
+            model_team_id = (model.get("model_info") or {}).get("team_id")
+            if (
+                team_id is None
+                or model_team_id is None  # global deployment - accessible to all teams
+                or model_team_id == team_id
+            ):
+                return True
         return False
 
     def _get_all_deployments(
@@ -8116,9 +8178,36 @@ class Router:
         if team_id specified, only return team-specific models
 
         Optimized with O(1) index lookup instead of O(n) linear scan.
+
+        Note: when team_id is provided, O(1) lookup in
+        `team_model_to_deployment_indices` only applies when `model_name` is the
+        team public model name. If a caller passes an internal deployment model
+        name (for example, `model_name_<team_id>_<uuid>`), this method falls back
+        to the standard model-name index / scan path.
         """
         returned_models: List[DeploymentTypedDict] = []
 
+        # O(1) lookup in team_model index when team_id is provided
+        if team_id is not None:
+            key = (team_id, model_name)
+            if key in self.team_model_to_deployment_indices:
+                indices = self.team_model_to_deployment_indices[key]
+                # O(k) where k = team deployments for this model_name (typically 1-10)
+                for idx in indices:
+                    model = self.model_list[idx]
+                    if not self.should_include_deployment(
+                        model_name=model_name, model=model, team_id=team_id
+                    ):
+                        continue
+                    if model_alias is not None:
+                        alias_model = model.copy()
+                        alias_model["model_name"] = model_alias
+                        returned_models.append(alias_model)
+                    else:
+                        returned_models.append(model)
+                if returned_models:
+                    return returned_models
+
         # O(1) lookup in model_name index
         if model_name in self.model_name_to_deployment_indices:
             indices = self.model_name_to_deployment_indices[model_name]
@@ -8778,6 +8867,16 @@ class Router:
             model = _model_from_alias
 
         if model not in self.model_names:
+            # Check for team-specific deployments by team_public_model_name.
+            # This intentionally takes priority over team pattern routers below,
+            # so that named team deployments shadow wildcard/pattern routes.
+            if request_team_id is not None:
+                team_deployments = self._get_all_deployments(
+                    model_name=model, team_id=request_team_id
+                )
+                if team_deployments:
+                    return model, team_deployments
+
             # check if provider/ specific wildcard routing use pattern matching
             pattern_deployments = self.pattern_router.get_deployments_by_pattern(
                 model=model,

pyproject.toml

@@ -61,7 +61,7 @@ boto3 = { version = "^1.40.76", optional = true }
 redisvl = {version = "^0.4.1", optional = true, markers = "python_version >= '3.9' and python_version < '3.14'"}
 mcp = {version = ">=1.25.0,<2.0.0", optional = true, python = ">=3.10"}
 a2a-sdk = {version = "^0.3.22", optional = true, python = ">=3.10"}
-litellm-proxy-extras = {version = "^0.4.56", optional = true}
+litellm-proxy-extras = {version = "^0.4.63", optional = true}
 rich = {version = "^13.7.1", optional = true}
 litellm-enterprise = {version = "^0.1.33", optional = true}
 diskcache = {version = "^5.6.1", optional = true}

requirements.txt

@@ -57,7 +57,7 @@ grpcio>=1.75.0; python_version >= "3.14"
 sentry_sdk==2.21.0 # for sentry error handling
 detect-secrets==1.5.0 # Enterprise - secret detection / masking in LLM requests
 tzdata==2025.1 # IANA time zone database
-litellm-proxy-extras==0.4.56 # for proxy extras - e.g. prisma migrations
+litellm-proxy-extras==0.4.63 # for proxy extras - e.g. prisma migrations
 llm-sandbox==0.3.31 # for skill execution in sandbox
 ### LITELLM PACKAGE DEPENDENCIES
 python-dotenv==1.0.1 # for env

tests/proxy_unit_tests/test_proxy_utils.py

@@ -2044,6 +2044,49 @@ def test_update_model_if_team_alias_exists(data, user_api_key_dict, expected_mod
     assert test_data.get("model") == expected_model
 
 
+def test_team_alias_stale_bypass_disabled_by_default():
+    from litellm.proxy.litellm_pre_call_utils import _update_model_if_team_alias_exists
+
+    class _MockRouter:
+        team_model_to_deployment_indices = {("team-1", "gpt-4o"): [0]}
+
+    test_data = {"model": "gpt-4o"}
+    user_api_key_dict = UserAPIKeyAuth(
+        api_key="test_key",
+        team_id="team-1",
+        team_model_aliases={"gpt-4o": "model_name_team-1_legacy-uuid"},
+    )
+
+    with patch("litellm.proxy.proxy_server.llm_router", _MockRouter()):
+        _update_model_if_team_alias_exists(
+            data=test_data, user_api_key_dict=user_api_key_dict
+        )
+
+    assert test_data.get("model") == "model_name_team-1_legacy-uuid"
+
+
+def test_team_alias_stale_bypass_enabled_by_flag(monkeypatch):
+    from litellm.proxy.litellm_pre_call_utils import _update_model_if_team_alias_exists
+
+    class _MockRouter:
+        team_model_to_deployment_indices = {("team-1", "gpt-4o"): [0]}
+
+    test_data = {"model": "gpt-4o"}
+    user_api_key_dict = UserAPIKeyAuth(
+        api_key="test_key",
+        team_id="team-1",
+        team_model_aliases={"gpt-4o": "model_name_team-1_legacy-uuid"},
+    )
+    monkeypatch.setenv("LITELLM_ENABLE_TEAM_STALE_ALIAS_BYPASS", "true")
+
+    with patch("litellm.proxy.proxy_server.llm_router", _MockRouter()):
+        _update_model_if_team_alias_exists(
+            data=test_data, user_api_key_dict=user_api_key_dict
+        )
+
+    assert test_data.get("model") == "gpt-4o"
+
+
 @pytest.fixture
 def mock_prisma_client():
     client = MagicMock()

tests/router_unit_tests/test_get_model_list_alias_optimization.py

@@ -46,5 +46,5 @@ def test_map_team_model_should_not_iterate_aliases_for_non_alias_team_model_name
 
     assert (
         router.map_team_model(team_model_name="team-model", team_id="team-1")
-        == "gpt-3.5-turbo"
+        == "team-model"
     )

tests/router_unit_tests/test_router_index_management.py

@@ -118,6 +118,28 @@ class TestRouterIndexManagement:
         assert router.model_id_to_deployment_index_map["id-2"] == 1
         assert router.model_id_to_deployment_index_map["id-3"] == 2
 
+    def test_update_team_model_index(self, router):
+        """Test _update_team_model_index updates team_model_to_deployment_indices."""
+        model = {
+            "model_name": "team-alias",
+            "model_info": {
+                "id": "dep-1",
+                "team_id": "team-abc",
+                "team_public_model_name": "gpt-4o",
+            },
+        }
+        router._update_team_model_index(model, 0)
+        assert router.team_model_to_deployment_indices[("team-abc", "gpt-4o")] == [0]
+        router._update_team_model_index(model, 2)
+        assert router.team_model_to_deployment_indices[("team-abc", "gpt-4o")] == [0, 2]
+
+        router._update_team_model_index(
+            {"model_name": "x", "model_info": {"id": "dep-2"}}, 5
+        )
+        assert router.team_model_to_deployment_indices == {
+            ("team-abc", "gpt-4o"): [0, 2],
+        }
+
     def test_has_model_id(self, router):
         """Test has_model_id function for O(1) membership check"""
         # Setup: Add models to router

tests/test_litellm/proxy/management_endpoints/test_model_management_endpoints.py

@@ -1,13 +1,14 @@
 import json
 import os
 import sys
-from litellm._uuid import uuid
 from typing import Dict, Optional
 from unittest.mock import AsyncMock, MagicMock, patch
 
 import pytest
 from fastapi.testclient import TestClient
 
+from litellm._uuid import uuid
+
 sys.path.insert(
     0, os.path.abspath("../../../..")
 )  # Adds the parent directory to the system path
@@ -27,9 +28,15 @@ from litellm.types.router import Deployment, LiteLLM_Params, updateDeployment
 
 
 class MockPrismaClient:
-    def __init__(self, team_exists: bool = True, user_admin: bool = True):
+    def __init__(
+        self,
+        team_exists: bool = True,
+        user_admin: bool = True,
+        sibling_deployments: list = None,
+    ):
         self.team_exists = team_exists
         self.user_admin = user_admin
+        self.sibling_deployments = sibling_deployments or []
         self.db = self
 
     async def find_unique(self, where):
@@ -45,10 +52,53 @@ class MockPrismaClient:
             )
         return None
 
+    async def find_many(self, where):
+        # Filter sibling deployments by team_id if where clause specifies it
+        if not self.sibling_deployments:
+            return []
+
+        # Extract team_id from where clause if present
+        team_id_filter = None
+        if where and "model_info" in where:
+            model_info_filter = where["model_info"]
+            if isinstance(model_info_filter, dict) and "path" in model_info_filter:
+                if (
+                    model_info_filter["path"] == ["team_id"]
+                    and "equals" in model_info_filter
+                ):
+                    team_id_filter = model_info_filter["equals"]
+
+        # Filter deployments by team_id if specified
+        if team_id_filter:
+
+            def _get_team_id(model_info):
+                if isinstance(model_info, dict):
+                    return model_info.get("team_id")
+                if isinstance(model_info, str):
+                    try:
+                        parsed = json.loads(model_info)
+                    except (TypeError, ValueError):
+                        return None
+                    if isinstance(parsed, dict):
+                        return parsed.get("team_id")
+                return None
+
+            return [
+                d
+                for d in self.sibling_deployments
+                if _get_team_id(d.model_info) == team_id_filter
+            ]
+
+        return self.sibling_deployments
+
     @property
     def litellm_teamtable(self):
         return self
 
+    @property
+    def litellm_proxymodeltable(self):
+        return self
+
 
 class MockLLMRouter:
     def __init__(self):
@@ -399,7 +449,9 @@ class TestClearCache:
         """
         Test that clear_cache clears DB models and preserves config models.
         """
-        from litellm.proxy.management_endpoints.model_management_endpoints import clear_cache
+        from litellm.proxy.management_endpoints.model_management_endpoints import (
+            clear_cache,
+        )
 
         # Create mock router with mixed DB and config models
         mock_router = MagicMock()
@@ -407,18 +459,18 @@ class TestClearCache:
             {
                 "model_name": "gpt-4",
                 "model_info": {"id": "db-model-1", "db_model": True},
-                "litellm_params": {"model": "gpt-4"}
+                "litellm_params": {"model": "gpt-4"},
             },
             {
-                "model_name": "gpt-3.5-turbo", 
+                "model_name": "gpt-3.5-turbo",
                 "model_info": {"id": "config-model-1", "db_model": False},
-                "litellm_params": {"model": "gpt-3.5-turbo"}
+                "litellm_params": {"model": "gpt-3.5-turbo"},
             },
             {
                 "model_name": "claude-3",
                 "model_info": {"id": "db-model-2", "db_model": True},
-                "litellm_params": {"model": "claude-3"}
-            }
+                "litellm_params": {"model": "claude-3"},
+            },
         ]
         mock_router.delete_deployment = MagicMock(return_value=True)
         mock_router.auto_routers = MagicMock()
@@ -466,8 +518,8 @@ class TestUpdatePublicModelGroups:
         """
         import litellm
         from litellm.proxy.management_endpoints.model_management_endpoints import (
-            update_public_model_groups,
             UpdatePublicModelGroupsRequest,
+            update_public_model_groups,
         )
 
         old_db_models = ["db-model-1", "db-model-2"]
@@ -525,7 +577,10 @@ class TestUpdatePublicModelGroups:
         )
 
         old_links = {"Old Doc": "https://old.example.com"}
-        new_links = {"New Doc": "https://new.example.com", "API Ref": "https://api.example.com"}
+        new_links = {
+            "New Doc": "https://new.example.com",
+            "API Ref": "https://api.example.com",
+        }
 
         async def mock_get_config(*args, **kwargs):
             litellm.public_model_groups_links = old_links
@@ -558,6 +613,161 @@ class TestUpdatePublicModelGroups:
             litellm.public_model_groups_links = original_value
 
 
+class TestTeamModelSiblingRouting:
+    """
+    Verify that sibling team deployments (same public model name, different
+    api_base) are all reachable through routing — no alias overwrite, no
+    collapse to a single deployment.
+    """
+
+    @pytest.mark.asyncio
+    async def test_no_model_aliases_written_for_team_models(self):
+        """
+        _add_team_model_to_db must NOT write model_aliases (which caused
+        the second sibling to overwrite the first). It should only call
+        team_model_add to register the public name on the team's models list.
+        """
+        from litellm.proxy.management_endpoints.model_management_endpoints import (
+            _add_team_model_to_db,
+        )
+        from litellm.types.router import ModelInfo
+
+        team_id = "team_no_alias"
+        public_name = "gpt-4.1-mini"
+
+        async def mock_add_model_to_db(model_params, user_api_key_dict, prisma_client):
+            return MagicMock(model_id=str(uuid.uuid4()))
+
+        mock_team_model_add = AsyncMock()
+
+        user = UserAPIKeyAuth(user_id="admin", user_role=LitellmUserRoles.PROXY_ADMIN)
+        prisma_client = MockPrismaClient(team_exists=True)
+
+        for api_base in ["https://eastus.example.com", "https://westus.example.com"]:
+            dep = Deployment(
+                model_name=public_name,
+                litellm_params=LiteLLM_Params(
+                    model="azure/gpt-4o-mini",
+                    api_key="key",
+                    api_base=api_base,
+                ),
+                model_info=ModelInfo(team_id=team_id),
+            )
+            with patch(
+                "litellm.proxy.management_endpoints.model_management_endpoints._add_model_to_db",
+                side_effect=mock_add_model_to_db,
+            ), patch(
+                "litellm.proxy.management_endpoints.model_management_endpoints.team_model_add",
+                mock_team_model_add,
+            ):
+                await _add_team_model_to_db(
+                    model_params=dep,
+                    user_api_key_dict=user,
+                    prisma_client=prisma_client,
+                )
+
+        assert mock_team_model_add.call_count == 2
+
+    @pytest.mark.asyncio
+    async def test_router_finds_all_sibling_team_deployments(self):
+        """
+        When two team deployments share team_public_model_name="gpt-4.1-mini",
+        the router's _common_checks_available_deployment must return BOTH as
+        healthy_deployments (not collapse to one).
+        """
+        import litellm
+
+        team_id = "teamA"
+        public_name = "gpt-4.1-mini"
+
+        router = litellm.Router(
+            model_list=[
+                {
+                    "model_name": f"model_name_{team_id}_uuid1",
+                    "litellm_params": {
+                        "model": "azure/gpt-4o-mini",
+                        "api_key": "key-1",
+                        "api_base": "https://eastus.openai.azure.com",
+                    },
+                    "model_info": {
+                        "team_id": team_id,
+                        "team_public_model_name": public_name,
+                    },
+                },
+                {
+                    "model_name": f"model_name_{team_id}_uuid2",
+                    "litellm_params": {
+                        "model": "azure/gpt-4o-mini",
+                        "api_key": "key-2",
+                        "api_base": "https://westus.openai.azure.com",
+                    },
+                    "model_info": {
+                        "team_id": team_id,
+                        "team_public_model_name": public_name,
+                    },
+                },
+                {
+                    "model_name": "global-gpt-4o",
+                    "litellm_params": {
+                        "model": "azure/gpt-4o",
+                        "api_key": "global-key",
+                        "api_base": "https://global.openai.azure.com",
+                    },
+                    "model_info": {},  # No team_id - global deployment
+                },
+            ],
+        )
+
+        # map_team_model should return the public name (not an internal UUID)
+        result = router.map_team_model(public_name, team_id)
+        assert result == public_name
+
+        # _common_checks_available_deployment should return both deployments
+        model, healthy = router._common_checks_available_deployment(
+            model=public_name,
+            request_kwargs={"metadata": {"user_api_key_team_id": team_id}},
+        )
+        assert isinstance(healthy, list)
+        assert len(healthy) == 2
+        api_bases = {d["litellm_params"]["api_base"] for d in healthy}
+        assert api_bases == {
+            "https://eastus.openai.azure.com",
+            "https://westus.openai.azure.com",
+        }
+
+    def test_global_deployments_accessible_to_teams(self):
+        """Test that global deployments (no team_id) are accessible to all teams"""
+        import litellm
+
+        router = litellm.Router(
+            model_list=[
+                {
+                    "model_name": "global-gpt-4o",
+                    "litellm_params": {
+                        "model": "azure/gpt-4o",
+                        "api_key": "global-key",
+                        "api_base": "https://global.openai.azure.com",
+                    },
+                    "model_info": {},  # No team_id - global deployment
+                },
+            ],
+        )
+
+        # Global deployment should be accessible when team_id is provided
+        deployments = router._get_all_deployments(
+            model_name="global-gpt-4o", team_id="teamA"
+        )
+        assert len(deployments) == 1
+        assert deployments[0]["model_name"] == "global-gpt-4o"
+
+        # should_include_deployment should return True for global deployments
+        assert router.should_include_deployment(
+            model_name="global-gpt-4o",
+            model={"model_name": "global-gpt-4o", "model_info": {}},
+            team_id="teamA",
+        )
+
+
 class TestTeamModelUpdate:
     """Test team model update handles team_id consistently with model creation"""
 
@@ -591,8 +801,6 @@ class TestTeamModelUpdate:
             "litellm.proxy.proxy_server.premium_user",
             True,
         ), patch(
-            "litellm.proxy.management_endpoints.model_management_endpoints.update_team"
-        ) as mock_update_team, patch(
             "litellm.proxy.management_endpoints.model_management_endpoints.team_model_add"
         ) as mock_team_model_add:
             result = await _update_team_model_in_db(
@@ -604,9 +812,200 @@ class TestTeamModelUpdate:
 
             assert result.get("model_name", "").startswith("model_name_test_team_123_")
             assert "team_public_model_name" in str(result.get("model_info", ""))
-            mock_update_team.assert_called_once()
+            # team_model_add must be called to add public name to team's models list
             mock_team_model_add.assert_called_once()
 
+    @pytest.mark.asyncio
+    async def test_rename_preserves_old_name_when_siblings_exist(self):
+        """Test that renaming a deployment preserves old public name when sibling deployments still use it"""
+        from unittest.mock import MagicMock
+
+        from litellm.proxy.management_endpoints.model_management_endpoints import (
+            _update_existing_team_model_assignment,
+        )
+        from litellm.types.router import ModelInfo
+
+        # Create a deployment being renamed
+        db_model = Deployment(
+            model_name="model_name_team_123_uuid1",
+            litellm_params=LiteLLM_Params(model="azure/gpt-4o-mini"),
+            model_info=ModelInfo(
+                team_id="team_123", team_public_model_name="old-public-name"
+            ),
+        )
+
+        # Create a sibling deployment that still uses the old public name
+        sibling_deployment = MagicMock()
+        sibling_deployment.model_name = "model_name_team_123_uuid2"
+        sibling_deployment.model_info = {
+            "team_id": "team_123",
+            "team_public_model_name": "old-public-name",
+        }
+
+        prisma_client = MockPrismaClient(
+            team_exists=True, sibling_deployments=[sibling_deployment]
+        )
+
+        patch_data = updateDeployment(
+            model_name="new-public-name",
+            model_info=ModelInfo(team_id="team_123"),
+        )
+
+        user_api_key_dict = UserAPIKeyAuth(
+            user_id="test_user",
+            user_role=LitellmUserRoles.PROXY_ADMIN,
+        )
+
+        with patch(
+            "litellm.proxy.management_endpoints.model_management_endpoints.team_model_delete"
+        ) as mock_delete, patch(
+            "litellm.proxy.management_endpoints.model_management_endpoints.team_model_add"
+        ) as mock_add:
+            await _update_existing_team_model_assignment(
+                team_id="team_123",
+                public_model_name="new-public-name",
+                db_model=db_model,
+                patch_data=patch_data,
+                user_api_key_dict=user_api_key_dict,
+                prisma_client=prisma_client,  # type: ignore
+            )
+
+            # team_model_delete should NOT be called because sibling exists
+            mock_delete.assert_not_called()
+            # team_model_add should be called to add new public name
+            mock_add.assert_called_once()
+
+    @pytest.mark.asyncio
+    async def test_first_time_public_name_assignment_adds_team_model(self):
+        """If existing team deployment had no public name, first assignment must call team_model_add."""
+        from litellm.proxy.management_endpoints.model_management_endpoints import (
+            _update_existing_team_model_assignment,
+        )
+        from litellm.types.router import ModelInfo
+
+        db_model = Deployment(
+            model_name="model_name_team_123_uuid1",
+            litellm_params=LiteLLM_Params(model="azure/gpt-4o-mini"),
+            model_info=ModelInfo(team_id="team_123"),
+        )
+
+        patch_data = updateDeployment(
+            model_name="new-public-name",
+            model_info=ModelInfo(team_id="team_123"),
+        )
+
+        user_api_key_dict = UserAPIKeyAuth(
+            user_id="test_user",
+            user_role=LitellmUserRoles.PROXY_ADMIN,
+        )
+
+        with patch(
+            "litellm.proxy.management_endpoints.model_management_endpoints.team_model_delete"
+        ) as mock_delete, patch(
+            "litellm.proxy.management_endpoints.model_management_endpoints.team_model_add"
+        ) as mock_add:
+            await _update_existing_team_model_assignment(
+                team_id="team_123",
+                public_model_name="new-public-name",
+                db_model=db_model,
+                patch_data=patch_data,
+                user_api_key_dict=user_api_key_dict,
+                prisma_client=None,
+            )
+
+            mock_add.assert_called_once()
+            mock_delete.assert_not_called()
+
+    @pytest.mark.asyncio
+    async def test_rename_with_prisma_none_clears_patch_model_name(self):
+        """Rename path must clear patch_data.model_name even when prisma is unavailable (P1)."""
+        from litellm.proxy.management_endpoints.model_management_endpoints import (
+            _update_existing_team_model_assignment,
+        )
+        from litellm.types.router import ModelInfo
+
+        db_model = Deployment(
+            model_name="model_name_team_123_uuid1",
+            litellm_params=LiteLLM_Params(model="azure/gpt-4o-mini"),
+            model_info=ModelInfo(
+                team_id="team_123", team_public_model_name="old-public-name"
+            ),
+        )
+        patch_data = updateDeployment(
+            model_name="new-public-name",
+            model_info=ModelInfo(team_id="team_123"),
+        )
+        user_api_key_dict = UserAPIKeyAuth(
+            user_id="test_user",
+            user_role=LitellmUserRoles.PROXY_ADMIN,
+        )
+
+        await _update_existing_team_model_assignment(
+            team_id="team_123",
+            public_model_name="new-public-name",
+            db_model=db_model,
+            patch_data=patch_data,
+            user_api_key_dict=user_api_key_dict,
+            prisma_client=None,
+        )
+
+        assert patch_data.model_name is None
+
+    @pytest.mark.asyncio
+    async def test_rename_handles_legacy_string_model_info(self):
+        """Test rename path handles legacy string-encoded model_info rows without crashing."""
+        from unittest.mock import MagicMock
+
+        from litellm.proxy.management_endpoints.model_management_endpoints import (
+            _update_existing_team_model_assignment,
+        )
+        from litellm.types.router import ModelInfo
+
+        db_model = Deployment(
+            model_name="model_name_team_123_uuid1",
+            litellm_params=LiteLLM_Params(model="azure/gpt-4o-mini"),
+            model_info=ModelInfo(
+                team_id="team_123", team_public_model_name="old-public-name"
+            ),
+        )
+
+        sibling_deployment = MagicMock()
+        sibling_deployment.model_name = "model_name_team_123_uuid2"
+        sibling_deployment.model_info = (
+            '{"team_id":"team_123","team_public_model_name":"old-public-name"}'
+        )
+
+        prisma_client = MockPrismaClient(
+            team_exists=True, sibling_deployments=[sibling_deployment]
+        )
+
+        patch_data = updateDeployment(
+            model_name="new-public-name",
+            model_info=ModelInfo(team_id="team_123"),
+        )
+
+        user_api_key_dict = UserAPIKeyAuth(
+            user_id="test_user",
+            user_role=LitellmUserRoles.PROXY_ADMIN,
+        )
+
+        with patch(
+            "litellm.proxy.management_endpoints.model_management_endpoints.team_model_delete"
+        ) as mock_delete, patch(
+            "litellm.proxy.management_endpoints.model_management_endpoints.team_model_add"
+        ) as mock_add:
+            await _update_existing_team_model_assignment(
+                team_id="team_123",
+                public_model_name="new-public-name",
+                db_model=db_model,
+                patch_data=patch_data,
+                user_api_key_dict=user_api_key_dict,
+                prisma_client=prisma_client,  # type: ignore
+            )
+
+            mock_delete.assert_not_called()
+            mock_add.assert_called_once()
+
     @pytest.mark.asyncio
     async def test_patch_model_with_team_id_validates_permissions(self):
         """Test PATCH with team_id runs same validation as POST for team permissions"""
@@ -657,27 +1056,37 @@ class TestModelInfoEndpoint:
             user_id="test_user",
             api_key="test_key",
             models=["gpt-4", "claude-3"],
-            team_models=["gpt-3.5-turbo"]
+            team_models=["gpt-3.5-turbo"],
         )
 
-        with patch("litellm.proxy.proxy_server.llm_router") as mock_router, \
-             patch("litellm.proxy.proxy_server.get_key_models") as mock_get_key_models, \
-             patch("litellm.proxy.proxy_server.get_team_models") as mock_get_team_models, \
-             patch("litellm.proxy.proxy_server.get_complete_model_list") as mock_get_complete_models, \
-             patch("litellm.get_llm_provider") as mock_get_provider:
-            
+        with patch("litellm.proxy.proxy_server.llm_router") as mock_router, patch(
+            "litellm.proxy.proxy_server.get_key_models"
+        ) as mock_get_key_models, patch(
+            "litellm.proxy.proxy_server.get_team_models"
+        ) as mock_get_team_models, patch(
+            "litellm.proxy.proxy_server.get_complete_model_list"
+        ) as mock_get_complete_models, patch(
+            "litellm.get_llm_provider"
+        ) as mock_get_provider:
             # Setup mocks
-            mock_router.get_model_names.return_value = ["gpt-4", "claude-3", "gpt-3.5-turbo"]
+            mock_router.get_model_names.return_value = [
+                "gpt-4",
+                "claude-3",
+                "gpt-3.5-turbo",
+            ]
             mock_router.get_model_access_groups.return_value = {}
             mock_get_key_models.return_value = ["gpt-4", "claude-3"]
             mock_get_team_models.return_value = ["gpt-3.5-turbo"]
-            mock_get_complete_models.return_value = ["gpt-4", "claude-3", "gpt-3.5-turbo"]
+            mock_get_complete_models.return_value = [
+                "gpt-4",
+                "claude-3",
+                "gpt-3.5-turbo",
+            ]
             mock_get_provider.return_value = (None, "openai", None, None)
 
             # Test accessible model
             result = await model_info(
-                model_id="gpt-4",
-                user_api_key_dict=user_api_key_dict
+                model_id="gpt-4", user_api_key_dict=user_api_key_dict
             )
 
             assert result["id"] == "gpt-4"
@@ -688,22 +1097,25 @@ class TestModelInfoEndpoint:
     @pytest.mark.asyncio
     async def test_model_info_inaccessible_model_returns_404(self):
         """Test model_info returns 404 for inaccessible models"""
-        from litellm.proxy.proxy_server import model_info
         from fastapi import HTTPException
 
+        from litellm.proxy.proxy_server import model_info
+
         # Mock user with limited access
         user_api_key_dict = UserAPIKeyAuth(
             user_id="test_user",
             api_key="test_key",
             models=["gpt-4"],  # Only has access to gpt-4
-            team_models=[]
+            team_models=[],
         )
 
-        with patch("litellm.proxy.proxy_server.llm_router") as mock_router, \
-             patch("litellm.proxy.proxy_server.get_key_models") as mock_get_key_models, \
-             patch("litellm.proxy.proxy_server.get_team_models") as mock_get_team_models, \
-             patch("litellm.proxy.proxy_server.get_complete_model_list") as mock_get_complete_models:
-            
+        with patch("litellm.proxy.proxy_server.llm_router") as mock_router, patch(
+            "litellm.proxy.proxy_server.get_key_models"
+        ) as mock_get_key_models, patch(
+            "litellm.proxy.proxy_server.get_team_models"
+        ) as mock_get_team_models, patch(
+            "litellm.proxy.proxy_server.get_complete_model_list"
+        ) as mock_get_complete_models:
             # Setup mocks - user only has access to gpt-4
             mock_router.get_model_names.return_value = ["gpt-4", "claude-3"]
             mock_router.get_model_access_groups.return_value = {}
@@ -715,32 +1127,35 @@ class TestModelInfoEndpoint:
             with pytest.raises(HTTPException) as exc_info:
                 await model_info(
                     model_id="claude-3",  # Not in user's accessible models
-                    user_api_key_dict=user_api_key_dict
+                    user_api_key_dict=user_api_key_dict,
                 )
-            
+
             assert exc_info.value.status_code == 404
             assert "does not exist or is not accessible" in exc_info.value.detail
 
-    @pytest.mark.asyncio 
+    @pytest.mark.asyncio
     async def test_model_info_team_model_access(self):
         """Test model_info works with team model access"""
         from litellm.proxy.proxy_server import model_info
-        
+
         # Mock user with team access
         user_api_key_dict = UserAPIKeyAuth(
             user_id="test_user",
-            api_key="test_key", 
+            api_key="test_key",
             team_id="test_team",
             models=[],  # No direct key models
-            team_models=["team-model-1"]
+            team_models=["team-model-1"],
         )
 
-        with patch("litellm.proxy.proxy_server.llm_router") as mock_router, \
-             patch("litellm.proxy.proxy_server.get_key_models") as mock_get_key_models, \
-             patch("litellm.proxy.proxy_server.get_team_models") as mock_get_team_models, \
-             patch("litellm.proxy.proxy_server.get_complete_model_list") as mock_get_complete_models, \
-             patch("litellm.get_llm_provider") as mock_get_provider:
-            
+        with patch("litellm.proxy.proxy_server.llm_router") as mock_router, patch(
+            "litellm.proxy.proxy_server.get_key_models"
+        ) as mock_get_key_models, patch(
+            "litellm.proxy.proxy_server.get_team_models"
+        ) as mock_get_team_models, patch(
+            "litellm.proxy.proxy_server.get_complete_model_list"
+        ) as mock_get_complete_models, patch(
+            "litellm.get_llm_provider"
+        ) as mock_get_provider:
             # Setup mocks
             mock_router.get_model_names.return_value = ["team-model-1"]
             mock_router.get_model_access_groups.return_value = {}
@@ -751,10 +1166,9 @@ class TestModelInfoEndpoint:
 
             # Test team model access
             result = await model_info(
-                model_id="team-model-1",
-                user_api_key_dict=user_api_key_dict
+                model_id="team-model-1", user_api_key_dict=user_api_key_dict
             )
 
             assert result["id"] == "team-model-1"
-            assert result["object"] == "model" 
+            assert result["object"] == "model"
             assert result["owned_by"] == "custom"