Update app.py

app.py

@@ -2,7 +2,6 @@ import gradio as gr
 from huggingface_hub import InferenceClient
 
 # --- Configuration: Model List ---
-# We use a Dictionary to map "Friendly Names" to "Model IDs"
 MODELS = {
     "Qwen 2.5 Coder 32B (Recommended)": "Qwen/Qwen2.5-Coder-32B-Instruct",
     "Llama 3.1 8B (Best Logic)": "meta-llama/Meta-Llama-3.1-8B-Instruct",
@@ -12,10 +11,10 @@ MODELS = {
 }
 
 # Configuration for Memory
-MAX_HISTORY = 5 
+MAX_HISTORY = 10
 
 def generate_abap(message, history, model_choice):
-    # 1. Get the Hugging Face Model ID from the dropdown selection
+    # 1. Get the Hugging Face Model ID
     model_id = MODELS.get(model_choice, "Qwen/Qwen2.5-Coder-32B-Instruct")
     
     client = InferenceClient()
@@ -23,16 +22,22 @@ def generate_abap(message, history, model_choice):
     system_prompt = "You are an expert SAP ABAP Developer. Write modern, efficient ABAP 7.4+ code. Always use inline declarations."
     messages = [{"role": "system", "content": system_prompt}]
     
-    # 2. Add History (Sliding Window)
-    recent_history = history[-MAX_HISTORY:] 
+    # 2. Add History (Robust Fix)
+    # We slice the history to keep memory usage low
+    recent_history = history[-MAX_HISTORY:]
     
     for turn in recent_history:
-        # Extract User and Bot messages safely
-        user_msg = turn[0]
-        bot_msg = turn[1]
-        messages.append({"role": "user", "content": str(user_msg)})
-        messages.append({"role": "assistant", "content": str(bot_msg)})
-    
+        # CASE 1: History is a List of Lists (Standard Gradio format: [[user, bot], ...])
+        if isinstance(turn, (list, tuple)):
+            messages.append({"role": "user", "content": str(turn[0])})
+            if len(turn) > 1 and turn[1] is not None:
+                messages.append({"role": "assistant", "content": str(turn[1])})
+        
+        # CASE 2: History is a List of Dictionaries (Newer format: [{'role': 'user', ...}])
+        elif isinstance(turn, dict):
+            # We can simply append the dictionary directly if it has 'role' and 'content'
+            messages.append(turn)
+
     # 3. Add Current Message
     messages.append({"role": "user", "content": str(message)})
 
@@ -41,7 +46,7 @@ def generate_abap(message, history, model_choice):
         stream = client.chat_completion(
             model=model_id,
             messages=messages,
-            max_tokens=2048, # Increased token limit for longer code
+            max_tokens=2048,
             temperature=0.1,
             top_p=0.9,
             stream=True
@@ -54,14 +59,13 @@ def generate_abap(message, history, model_choice):
                 yield partial_message
                 
     except Exception as e:
-        yield f"Error: The Free API is overloaded for {model_choice}. Try switching to Qwen or Llama. \n\nDetails: {str(e)}"
+        yield f"Error: The Free API is overloaded or model is too large. \nDetails: {str(e)}"
 
 # --- The UI ---
-with gr.Blocks( ) as demo:
+with gr.Blocks(theme="soft") as demo:
     gr.Markdown("# 🚀 ABAP Coder Multi-Model")
-    gr.Markdown("Select a model below. If one gives an error, try another!")
+    gr.Markdown("Select a model below. **Note:** Qwen 32B is large and may timeout on the free tier. If it fails, try Llama 3.1 8B.")
     
-    # Dropdown with all our new models
     model_selector = gr.Dropdown(
         choices=list(MODELS.keys()),
         value="Qwen 2.5 Coder 32B (Recommended)",