A multi-phase context pipeline built with Python, Gradio, and LiteLLM. This engine mitigates Large Language Model (LLM) hallucinations by forcing the model through a structured, three-phase cognitive workflow (System-2 thinking) before generating a final calculation or answer.