Yes. Map the scale postmeta with values like 1/4, 1/6, 1/7, 1/8, 1/12, nendoroid, figma, and Q Posket. The system instruction tells the bot to always name the scale and the figure line when quoting, so a $74 nendoroid never gets confused with a $329 scale figure on the same character.

Add a licensed boolean and a recast_warning field. The bot reads both on every turn. If your shop is officially licensed and never stocks recasts, the bot says so plainly. If you also carry trusted third-party brands, the system instruction can be tuned to name them while flagging clearly unlicensed sellers as not stocked.

If you populate the manga_arc postmeta per volume, yes. The bot reads which arc each volume covers and quotes the volume range that maps to a named arc. For incomplete metadata, the system instruction tells it to point the reader to the volume index rather than guess and risk spoilers.

Yes. release_date is read on every turn. When the date is in the future, the bot says preorder explicitly, names the month and year, and warns if the window is longer than six months. Once a figure ships and stock arrives, _stock_quantity and the absence of a future release_date flip the bot into in-stock language.

Yes. Use a language postmeta with values like English Viz, English Yen Press, English Kodansha, and Japanese tankobon. The bot quotes the publisher and language explicitly on every manga answer so a customer asking for Chainsaw Man volume 14 in English gets the Viz price, not the Japanese import.

If you store photo galleries as ACF or postmeta image arrays, the bot can quote how many photos are available and offer a link. For loose secondhand figures, add a condition field and the bot will quote box condition, complete with accessories status, and whether the certificate of authenticity is included.

Yes. Multibot plus the series taxonomy lets each series archive have its own bot. A Demon Slayer bot on /series/demon-slayer/ uses character names confidently, while the manga bot on /shop/manga/ stays in publisher-and-volume territory. Display conditions handle the routing.

Every conversation is stored as a custom post in your own database with the model name, token usage, the originating product or series URL, and the full transcript. Nothing leaves your site except the model call itself, which goes to OpenAI, Anthropic, Google, or OpenRouter with your own API key.