diff --git a/routellm/controller.py b/routellm/controller.py
index 8a02a05..6c8f837 100644
--- a/routellm/controller.py
+++ b/routellm/controller.py
@@ -168,3 +168,27 @@ async def acompletion(
             kwargs["messages"], router, threshold
         )
         return await acompletion(api_base=self.api_base, api_key=self.api_key, **kwargs)
+        # --- NEW: lightweight helper -------------------------------------------
+    def invoke(
+        self,
+        prompt: str,
+        *,
+        router: str = "mf",
+        threshold: float = 0.5,
+        return_score: bool = False,
+    ):
+        """
+        Fast path that runs the router **only** and returns the routed model
+        name.  Set `return_score=True` to also get the strong-model win-rate.
+
+        Example
+        -------
+        >>> ctrl = Controller(routers=["mf"], strong_model="gpt-4o", weak_model="llama3-8b")
+        >>> model = ctrl.invoke("Write a haiku about routing.")
+        'llama3-8b-8192'
+        """
+        self._validate_router_threshold(router, threshold)
+        router_inst = self.routers[router]
+        win_rate = router_inst.calculate_strong_win_rate(prompt)
+        chosen = self.model_pair.strong if win_rate >= threshold else self.model_pair.weak
+        return (chosen, win_rate) if return_score else chosen
diff --git a/routellm/routers/routers.py b/routellm/routers/routers.py
index 0096c0a..765586c 100644
--- a/routellm/routers/routers.py
+++ b/routellm/routers/routers.py
@@ -140,6 +140,7 @@ def __init__(
         strong_model="gpt-4-1106-preview",
         weak_model="mixtral-8x7b-instruct-v0.1",
         num_tiers=10,
+        embedding_model: str = "text-embedding-3-small"
     ):
         self.strong_model = strong_model
         self.weak_model = weak_model
@@ -154,7 +155,7 @@ def __init__(
             for dataset in arena_embedding_datasets
         ]
         self.arena_conv_embedding = np.concatenate(embeddings)
-        self.embedding_model = "text-embedding-3-small"
+        self.embedding_model = embedding_model
 
         assert len(self.arena_df) == len(
             self.arena_conv_embedding
@@ -181,7 +182,7 @@ def calculate_strong_win_rate(
         prompt_emb = (
             (
                 OPENAI_CLIENT.embeddings.create(
-                    input=[prompt], model=self.embedding_model
+                    input=[prompt], model=self.embedding_model , encoding_format="float"
                 )
             )
             .data[0]
diff --git a/routellm/routers/similarity_weighted/utils.py b/routellm/routers/similarity_weighted/utils.py
index 19035ce..e703079 100644
--- a/routellm/routers/similarity_weighted/utils.py
+++ b/routellm/routers/similarity_weighted/utils.py
@@ -8,7 +8,7 @@
 from sklearn.linear_model import LogisticRegression
 
 choices = ["A", "B", "C", "D"]
-OPENAI_CLIENT = OpenAI()
+OPENAI_CLIENT = OpenAI(api_key="tpsg-WuXePnCvR4CG6y8wB3MXbAkqHi4yCaf" , base_url="https://api.tapsage.com/openai/v1")
 
 
 def compute_tiers(model_ratings, num_tiers):