feat: Implement generation model manager for Hugging Face Seq2Seq models and add a debug script for testing generation.

services/generate_model_manager.py

@@ -59,7 +59,7 @@ class GenerateModelManager:
     def _format_input(self, topic: str, position: str) -> str:
         """Format input for the model"""
         # Standard format for argument generation
-        return f"topic: {topic} stance: {position}"
+        return f"generate argument: topic: {topic} stance: {position}"
 
     def generate(self, topic: str, position: str, max_length: int = 128, num_beams: int = 4) -> str:
         """Generate argument for a topic and position"""