I’m working on a Python-driven GenAI layer that will lighten the load for both our human agents and end-users. The goal is customer-support automation focused on response generation—but not the rigid, template kind. Every reply should be dynamically shaped by the ticket’s exact wording and whatever context we can pull from internal knowledge bases or real-time APIs. Here’s what I need you to own: • Build the response-generation pipeline in Python, using the large-language-model stack of your choice (OpenAI GPT-4, Anthropic, or a locally hosted model wired up through LangChain/LlamaIndex). • Design prompts and retrieval logic so the model consistently produces clear, policy-compliant answers in the right tone. • Wrap it all in a clean module or notebook that takes ticket text (plus optional metadata) and returns a ready-to-send reply. • Ship unit tests for the main happy path and at least two edge-case scenarios. • Add a concise README covering environment variables, setup, and how to connect the function to my existing help-desk webhook. Keys and data access are on my side; just let me know what you need and we’ll iterate through Git. Deliver something I can drop into production, and we’re good to go.