"""
Search Agent — Product search in the catalog.

Pipeline without LLM:
1. SEARCH: the resolved_query (in English, translated by the Router) is searched
   in the products FAISS index (k=12) with multilingual embeddings
2. RESPONSE: the FAISS results are returned as a structured list
   (the frontend shows the cards with thumbnails, the user filters visually)

Total: 0 LLM calls.

Uses the factory pattern: create_search_node(product_retriever) -> node function.
"""

from langchain_core.messages import AIMessage
from utils.i18n import t


def create_search_node(product_retriever):
    """Creates the Search Agent node with injected dependencies.

    Args:
        product_retriever: FAISS retriever of the product catalog (100k items)
    """
    def search_node(state):
        if not product_retriever:
            return {"messages": [AIMessage(content=t("search_catalog_unavailable"))]}

        # Use the query resolved by the Router (with context), fallback to the last message
        user_query = state.get("resolved_query", state["messages"][-1].content)

        # --- FAISS search (resolved_query in English from the Router) ---
        print("--- [SEARCH AGENT] Querying catalog... ---")
        print(f"   Query: '{user_query}'")
        results = product_retriever.invoke(user_query)

        if not results:
            return {
                "messages": [AIMessage(content=t("search_no_results"))],
                "products": [],
            }

        # Build structured product list
        products = []
        for doc in results:
            meta = doc.metadata
            asin = meta.get("asin", "")
            title = meta.get("title", t("unknown_product"))
            brand = meta.get("brand", "")
            price = meta.get("price", "")

            if brand == "Generic":
                brand = ""
            if price == "N/A":
                price = ""

            products.append({
                "asin": asin,
                "title": title,
                "brand": brand,
                "price": price,
                "similarity": None,
            })
            print(f"   -> {title[:60]}")

        msg = t("search_results")
        return {
            "messages": [AIMessage(content=msg)],
            "products": products,
        }

    return search_node