Update inference.py

inference.py

@@ -18,7 +18,7 @@ tokenizer = AutoTokenizer.from_pretrained("bert-base-uncased")
 model = None
 last_mod_time = 0
 
-# 🔁 Load Evo model
+# 🔁 Load Evo model with auto-reload
 def load_model():
     global model, last_mod_time
     try:
@@ -34,14 +34,14 @@ def load_model():
         model = None
     return model
 
-# 🔮 Evo inference
+# 🔮 Evo inference core logic
 def evo_infer(query, options, user_context=""):
     model = load_model()
     if model is None:
         return "Model Error", 0.0, "Model not available", ""
 
-    context_str = "\n".join(web_search(query) + ([user_context] if user_context else []))
-    input_pairs = [f"{query} [SEP] {opt} [CTX] {context_str}" for opt in options]
+    context_text = "\n".join(web_search(query) + ([user_context] if user_context else []))
+    input_pairs = [f"{query} [SEP] {opt} [CTX] {context_text}" for opt in options]
 
     scores = []
     for pair in input_pairs:
@@ -56,10 +56,10 @@ def evo_infer(query, options, user_context=""):
         options[best_idx],
         max(scores),
         f"{options[0]}: {scores[0]:.3f} vs {options[1]}: {scores[1]:.3f}",
-        context_str
+        context_text
     )
 
-# 💬 GPT fallback (renamed properly to fix import)
+# 💬 GPT fallback (used for comparison only)
 def get_gpt_response(query, user_context=""):
     try:
         context_block = f"\n\nContext:\n{user_context}" if user_context else ""
@@ -72,9 +72,19 @@ def get_gpt_response(query, user_context=""):
     except Exception as e:
         return f"⚠️ GPT error:\n{str(e)}"
 
-# 🔄 UI integration
+# 🤖 Evo live chat prediction
 def evo_chat_predict(history, query, options):
-    context = "\n".join(history[-6:]) if history else ""
+    try:
+        # Support list or DataFrame
+        if isinstance(history, list):
+            context = "\n".join(history[-6:])
+        elif hasattr(history, "empty") and not history.empty:
+            context = "\n".join(history.tail(6).astype(str).tolist())
+        else:
+            context = ""
+    except Exception:
+        context = ""
+
     evo_ans, evo_score, evo_reason, evo_ctx = evo_infer(query, options, context)
     return {
         "answer": evo_ans,
@@ -83,7 +93,7 @@ def evo_chat_predict(history, query, options):
         "context_used": evo_ctx
     }
 
-# 📊 Evo config
+# 📊 Evo architecture stats
 def get_model_config():
     return {
         "num_layers": 6,
@@ -94,7 +104,7 @@ def get_model_config():
         "accuracy": "~64.5%"
     }
 
-# 💻 System info
+# 🖥️ System runtime stats
 def get_system_stats():
     gpu_info = torch.cuda.get_device_properties(0) if torch.cuda.is_available() else None
     memory = psutil.virtual_memory()