Answer a shopper's question about one specific product. The agent answers from that product's own data rather than by searching the catalog, which is what makes it suitable for the question box on a product page. Which sources it draws on is configured per index key, and typically covers the product's attributes and description, and its customer reviews where those are ingested. Frequently asked questions may be answered from a pre-generated set, so a common question returns faster than a novel one. An answer can cite the part of the product imagery it draws on, such as a size chart or a dimensions diagram, and can suggest related products where they help. page_context.item_id is the product the question is about, and shopper_input is the question itself. Pass thread_id to ask a follow-up in the same conversation. Results stream back as server-sent events (text/event-stream) so the agent's reply can be rendered as it is produced. To receive the finished reply as a single JSON document instead, send Accept: application/json.