RAG Support Agent Pipeline

Implement support_pipeline(query, history, retrieve_fn, answer_fn). - query is the current user question (string). - history is a list of previous user messages (strings), oldest first. - If query has fewer than 5 words AND history is non-empty, rewrite the query by prepending the last history message: f"{history[-1]} {query}". Otherwise use the query as-is. - Call retrieve_fn(rewritten_query) to get a list of doc dicts. Each doc has {"id": str, "content": str, "score": float}. - Filter out docs where score < 0.5. - If no docs remain after filtering, return {"answer": "I cannot answer this from the available documentation.", "citations": [], "confidence": 0.0}. - Otherwise call answer_fn(rewritten_query, filtered_docs). It returns {"text": str, "citations": [str]}. - Compute confidence = min(1.0, max_score) using the highest score among the filtered docs. - Return {"answer": result["text"], "citations": result["citations"], "confidence": confidence}.