Merge e0aa0a093c into 635085ac14

litellm/router.py

@@ -10447,6 +10447,10 @@ class Router:
         configured or discovered token limits. Resolved via O(1) index
         lookup.
 
+        A ``model_group_alias`` name that is not itself a deployment name resolves to its
+        target group, as routing does, so a listed alias reports the limits of the
+        deployments it routes to rather than none at all.
+
         Returns None for wildcard-expanded or unknown names, where the listed name is the
         real model name and no deployment-specific information exists, and treats a
         malformed configured limit as absent rather than failing the listing.
@@ -10459,7 +10463,9 @@ class Router:
         this never triggers pattern matching or deep copies, so it is safe to call per
         listed model on the /v1/models hot path.
         """
-        indices: Final = self.model_name_to_deployment_indices.get(model_name)
+        indices: Final = self.model_name_to_deployment_indices.get(
+            model_name
+        ) or self.model_name_to_deployment_indices.get(self._get_model_from_alias(model_name) or "")
         if not indices:
             return None
 

tests/unit/test_router_model_listing_alias.py

@@ -0,0 +1,52 @@
+"""
+A model_group_alias listed by /v1/models reports the token limits of the group it
+routes to. Before this, the listing lookup read only deployment names, so an alias
+row carried no max_input_tokens while /model/info showed them, and Claude Code's
+picker view could not mark a 1M-context alias with [1m].
+"""
+
+from typing import Final
+
+from litellm import Router
+from litellm.proxy.utils import create_model_info_response
+
+
+def _router(model_group_alias: dict) -> Router:
+    return Router(
+        model_list=[
+            {
+                "model_name": "long-context",
+                "litellm_params": {"model": "openai/org/long-context-model", "api_key": "k"},
+                "model_info": {"id": "lc", "max_input_tokens": 1_000_000, "max_output_tokens": 128_000},
+            },
+            {
+                "model_name": "short",
+                "litellm_params": {"model": "openai/org/short-model", "api_key": "k"},
+                "model_info": {"id": "s", "max_input_tokens": 8192, "max_output_tokens": 4096},
+            },
+        ],
+        model_group_alias=model_group_alias,
+    )
+
+
+def test_alias_reports_its_target_groups_limits() -> None:
+    router: Final = _router({"big": "long-context", "item": {"model": "long-context", "hidden": False}})
+    for alias in ("big", "item"):
+        listing = router.get_model_listing_info(alias)
+        assert listing is not None
+        assert (listing.max_input_tokens, listing.max_output_tokens) == (1_000_000, 128_000)
+        row = create_model_info_response(model_id=alias, provider="openai", llm_router=router)
+        assert (row["id"], row.get("max_input_tokens"), row.get("max_output_tokens")) == (alias, 1_000_000, 128_000)
+
+
+def test_a_deployment_name_wins_over_an_alias_of_the_same_name() -> None:
+    router: Final = _router({"short": "long-context"})
+    listing: Final = router.get_model_listing_info("short")
+    assert listing is not None
+    assert listing.max_input_tokens == 8192
+
+
+def test_alias_to_an_unknown_group_and_unknown_names_stay_none() -> None:
+    router: Final = _router({"dangling": "no-such-group"})
+    assert router.get_model_listing_info("dangling") is None
+    assert router.get_model_listing_info("not-listed") is None