Update engine/responder.py

engine/responder.py

@@ -47,46 +47,38 @@ def generate_response(student_prompt, persona, conversation_history, force_mode=
 
 
 def generate_response_hf(student_prompt, persona, conversation_history, force_mode=None):
-    """
-    Generate response using Hugging Face Inference API (free, non-gated models).
-    """
+    """Generate response using Hugging Face Inference API (free, non-gated models)."""
     try:
         from huggingface_hub import InferenceClient
 
-        # Initialize state and mode
         state = persona.get("default_state", {}).copy()
         if force_mode:
             state["mode"] = force_mode
 
         mode = get_current_mode(state)
-
-        # Apply response effects
         state = apply_response_effects(state, student_prompt)
         mode = get_current_mode(state)
 
-        # Build prompt components
         system_prompt = build_system_prompt_for_ai(persona, state, mode)
         name = persona.get("persona_name", "Client")
 
-        # Format conversation history
         messages = [{"role": "system", "content": system_prompt}]
         for turn in conversation_history[-3:]:
             if "student" in turn:
                 messages.append({"role": "user", "content": turn["student"]})
             if "client" in turn:
                 messages.append({"role": "assistant", "content": turn["client"]})
-
-        # Add current student prompt
         messages.append({"role": "user", "content": student_prompt})
 
-        # Initialize client
-        client = InferenceClient(token=os.getenv("HF_TOKEN"))
+        print("[DEBUG] Prompt sent to model:")
+        import pprint
+        pprint.pprint(messages)
 
-        # Model priority list (fastest first)
+        client = InferenceClient(token=os.getenv("HF_TOKEN"))
         models = [
-            "microsoft/Phi-3-mini-4k-instruct",    # ✅ Fast and capable
-            "HuggingFaceH4/zephyr-7b-beta",        # Good instruction following
-            "mistralai/Mistral-7B-Instruct-v0.2",  # Higher quality, slower
+            "microsoft/Phi-3-mini-4k-instruct",
+            "HuggingFaceH4/zephyr-7b-beta",
+            "mistralai/Mistral-7B-Instruct-v0.2",
         ]
 
         response_text = None
@@ -100,17 +92,16 @@ def generate_response_hf(student_prompt, persona, conversation_history, force_mo
                     stream=False
                 )
                 response_text = response.choices[0].message.content.strip()
-                break  # Success
+                if response_text:
+                    break
             except Exception as model_error:
                 from engine.utils import safe_log
                 safe_log(f"HF model {model} failed", str(model_error))
-                continue  # Try next model
+                continue
 
-        # If all models failed
         if not response_text:
             raise Exception("All HF models failed")
 
-        # Update emotional memory
         if "emotional_memory" in state:
             if not isinstance(state["emotional_memory"], list):
                 state["emotional_memory"] = []
@@ -118,9 +109,8 @@ def generate_response_hf(student_prompt, persona, conversation_history, force_mo
             state["emotional_memory"].append(memory_tag)
             state["emotional_memory"] = state["emotional_memory"][-5:]
 
-        # Generate teaching note
         teaching_note = generate_teaching_note(state, student_prompt, mode)
-        teaching_note += "\n\n💡 Response generated using AI (Hugging Face)"
+        teaching_note += f"\n\n💡 Response generated using {model}"
 
         return response_text, state, teaching_note