Upd HF token usage:

requirements.txt

@@ -1,5 +1,5 @@
 # requirements.txt
-# **Agents**
+# **LLMs**
 google-genai
 huggingface_hub
 # **RAG**
@@ -12,6 +12,9 @@ sentencepiece
 # **Environment**
 python-dotenv       # Not used in Streamlit deployment
 pymongo
+# **VLMs**
+transformers
+pillow
 # **Deployment**
 uvicorn
 fastapi

vlm.py

@@ -19,7 +19,7 @@ def load_vlm():
     global vlm_pipe
     if vlm_pipe is None:
         logger.info("⏳ Loading MedGEMMA model via Transformers pipeline...")
-        vlm_pipe = pipeline("image-to-text", model="google/medgemma-4b", device_map="auto")
+        vlm_pipe = pipeline("image-to-text", model="google/medgemma-4b", use_auth_token=HF_TOKEN, device_map="auto")
         logger.info("✅ MedGEMMA model ready.")
     return vlm_pipe