history fix

app.py

@@ -172,6 +172,16 @@ def chat_with_llm(messages: List[Dict[str, str]], max_tokens: int = 200, tempera
             )
             
             try:
+                # Limit conversation history to prevent context length issues
+                # Keep system message + last 10 message pairs (20 messages max)
+                if len(chat_messages) > 21:  # 1 system + 20 conversation messages
+                    # Keep system message and truncate to last 20 messages
+                    system_msg = chat_messages[0] if chat_messages[0].get("role") == "system" else None
+                    recent_messages = chat_messages[-20:] if not system_msg else [system_msg] + chat_messages[-20:]
+                    chat_messages = recent_messages
+                    print(f"DEBUG: Truncated conversation history from {len(chat_messages) + 1} to {len(chat_messages)} messages")
+                
+                print(f"DEBUG: Sending {len(chat_messages)} messages to Hugging Face API")
                 response = hf_client.chat_completion(
                     messages=chat_messages,
                     max_tokens=max_tokens,
@@ -200,14 +210,26 @@ def chat_with_llm(messages: List[Dict[str, str]], max_tokens: int = 200, tempera
                 # Otherwise convert to string
                 return str(response).strip()
             except Exception as chat_error:
-                # Chat completion failed - check if model only supports conversational
+                # Chat completion failed - log the full error for debugging
                 error_str = str(chat_error).lower()
+                error_full = str(chat_error)
+                print(f"ERROR: Hugging Face chat_completion failed: {error_full}")
+                print(f"ERROR: Error type: {type(chat_error)}")
+                print(f"ERROR: Number of messages sent: {len(chat_messages)}")
+                
+                # Check for specific error types
+                if "rate limit" in error_str or "429" in error_str or "quota" in error_str:
+                    raise Exception(f"Rate limit exceeded. Please wait a moment and try again. Error: {error_full}")
+                elif "timeout" in error_str or "timed out" in error_str:
+                    raise Exception(f"Request timed out. The conversation might be too long. Please try a shorter message. Error: {error_full}")
+                elif "context length" in error_str or "token" in error_str and "limit" in error_str:
+                    raise Exception(f"Conversation too long. Please start a new conversation. Error: {error_full}")
                 
                 # Never try text_generation for conversational-only models (Mistral, etc.)
                 if model_supports_only_conversational:
                     # Model only supports conversational/chat completion, don't try text generation
                     print(f"Chat completion failed for conversational-only model ({HF_MODEL}): {chat_error}")
-                    raise chat_error
+                    raise Exception(f"Hugging Face API error with {HF_MODEL}: {error_full}")
                 
                 # Also check error message for indicators that model only supports conversational
                 if "conversational" in error_str or ("text-generation" in error_str and "not supported" in error_str):