Use HF_TOKEN env var to authenticate for gated MedGemma model

models/medgemma_agent.py

@@ -203,9 +203,17 @@ class MedGemmaAgent:
 
         self._print("Initializing MedGemma agent...")
 
+        import os
         import torch
         from transformers import pipeline
 
+        # Authenticate with HF Hub if a token is provided (required for gated models)
+        hf_token = os.environ.get("HF_TOKEN")
+        if hf_token:
+            from huggingface_hub import login
+            login(token=hf_token, add_to_git_credential=False)
+            self._print("Authenticated with HF Hub")
+
         self._print(f"Loading model: {self.model_id}")
 
         if torch.cuda.is_available():