fix(proxy): keep the flat DB where shape for single-condition searches

The upstream CI test
test_apply_search_filter_honours_exact_model_name_in_db_query pins the
Prisma where shape: exact model_name stays a top-level key, and plain
substring search keeps {model_name: {contains, mode}} at the top level.
Only genuinely multi-condition queries (status filter and/or router-id
exclusion) now wrap in AND.

litellm/proxy/proxy_server.py

@@ -14467,11 +14467,19 @@ async def _fetch_db_models_for_search(
     filter for `team_public_model_name` instead and keep the DB cost
     bounded by `search`.
     """
+    model_name_condition: Final[dict[str, Any]] = (
+        {"model_name": {"contains": search_lower, "mode": "insensitive"}}
+        if model_name is None
+        else {"model_name": model_name}
+    )
     match_conditions: list[dict[str, Any]] = (
         [
             {
                 "OR": [
-                    {"model_name": {"contains": search_lower, "mode": "insensitive"}},
+                    model_name_condition,
+                    # Substring search also matches the underlying LiteLLM model
+                    # name (e.g. "openrouter/deepseek/deepseek-chat"), so users
+                    # can find deployments by provider or upstream model id.
                     # JSON string_contains is case-sensitive on Postgres (see
                     # note above); router-side matching below covers the
                     # case-insensitive path for rows already in the router.
@@ -14480,7 +14488,7 @@ async def _fetch_db_models_for_search(
             }
         ]
         if model_name is None
-        else [{"model_name": model_name}]
+        else [model_name_condition]
     )
     # Status filter runs inside the DB query too: the fetch is capped, so
     # matches of the other status must not consume the page budget.
@@ -14488,7 +14496,12 @@ async def _fetch_db_models_for_search(
         match_conditions.append({"model_info": {"path": ["blocked"], "equals": blocked}})
     if db_model_ids_in_router:
         match_conditions.append({"model_id": {"not": {"in": list(db_model_ids_in_router)}}})
-    db_where_condition: Final[dict[str, Any]] = {"AND": match_conditions}
+    # Keep the single-condition shape flat: it is what existing callers (and
+    # tests) assert, and Prisma treats both forms identically.
+    if len(match_conditions) == 1:
+        db_where_condition: Final[dict[str, Any]] = match_conditions[0]
+    else:
+        db_where_condition = {"AND": match_conditions}
 
     # Unsorted searches only need enough DB rows to fill the current
     # page after counting router-side matches. Sorted searches need

tests/test_litellm/proxy/test_proxy_server.py

@@ -2594,7 +2594,10 @@ async def test_apply_search_filter_honours_exact_model_name_in_db_query():
         proxy_config=proxy_config,
     )
     where = prisma_client.db.litellm_proxymodeltable.count.call_args.kwargs["where"]
-    assert where["model_name"] == {"contains": "sonnet", "mode": "insensitive"}
+    assert where["OR"] == [
+        {"model_name": {"contains": "sonnet", "mode": "insensitive"}},
+        {"litellm_params": {"path": ["model"], "string_contains": "sonnet"}},
+    ]
 
 
 @pytest.mark.asyncio

ui/litellm-dashboard/src/app/(dashboard)/models-and-endpoints/components/AllModelsTable.tsx

@@ -312,13 +312,13 @@ export function AllModelsTable({
                       ] ?? ROUTING_STATUS_FILTER_VALUES.all}
                     </SelectTrigger>
                     <SelectContent>
-                      {(Object.keys(ROUTING_STATUS_FILTER_VALUES) as Array<keyof typeof ROUTING_STATUS_FILTER_VALUES>).map(
-                        (value) => (
-                          <SelectItem key={value} value={value}>
-                            {ROUTING_STATUS_FILTER_VALUES[value]}
-                          </SelectItem>
-                        ),
-                      )}
+                      {(
+                        Object.keys(ROUTING_STATUS_FILTER_VALUES) as Array<keyof typeof ROUTING_STATUS_FILTER_VALUES>
+                      ).map((value) => (
+                        <SelectItem key={value} value={value}>
+                          {ROUTING_STATUS_FILTER_VALUES[value]}
+                        </SelectItem>
+                      ))}
                     </SelectContent>
                   </Select>
                 </DataTableFilterField>