Update hf_client.py

hf_client.py

@@ -1,25 +1,36 @@
+# hf_client.py (REVISED)
+
 import os
+from typing import Optional
 
-from huggingface_hub import InferenceClient, HfApi
+from huggingface_hub import InferenceClient
 from tavily import TavilyClient
 
 # HF Inference Client
+# This is your token, used as a fallback for local testing or when a user token isn't available.
 HF_TOKEN = os.getenv('HF_TOKEN')
-if not HF_TOKEN:
-    raise RuntimeError("HF_TOKEN environment variable is not set. Please set it to your Hugging Face API token.")
 
-def get_inference_client(model_id, provider="auto"):
-    """Return an InferenceClient with provider based on model_id and user selection."""
+def get_inference_client(model_id: str, provider: str = "auto", user_token: Optional[str] = None):
+    """
+    Return an InferenceClient.
+    Prioritizes the user's token for billing if provided.
+    Falls back to the environment HF_TOKEN otherwise.
+    """
+    # Determine which token to use for the API call
+    token_to_use = user_token or HF_TOKEN
+    
+    if not token_to_use:
+        raise ValueError("An API token must be provided either by the user or as an environment variable (HF_TOKEN).")
+
     if model_id == "moonshotai/Kimi-K2-Instruct":
         provider = "groq"
     
-    # When using a third-party provider, simply specify the provider.
-    # Billing will automatically default to the account associated with the HF_TOKEN.
-    # The 'bill_to' parameter is NOT needed for personal billing.
+    # Initialize the client with the selected token.
+    # The Hugging Face Hub automatically bills the account associated with the token.
+    # We do NOT use the 'bill_to' parameter for this.
     return InferenceClient(
         provider=provider,
-        api_key=HF_TOKEN
-        # The `bill_to` parameter has been removed.
+        api_key=token_to_use
     )
 
 # Tavily Search Client