import gradio as gr from src.core.engine import engine from src.utils.helpers import get_clean_text def bot_response(history, system_prompt, temperature, max_tokens): messages = [{"role": "system", "content": system_prompt}] for msg in history[-7:]: messages.append( {"role": msg["role"], "content": get_clean_text(msg["content"])} ) history.append({"role": "assistant", "content": ""}) try: stream = engine.generate_stream(messages, max_tokens, temperature) partial_text = "" for chunk in stream: delta = chunk["choices"][0]["delta"] if "content" in delta: partial_text += delta["content"] history[-1]["content"] = partial_text yield history except Exception as e: history[-1]["content"] += f"\n\n❌ Error: {str(e)}" yield history