MetaSlider models each slideshow as a term in the ml-slider taxonomy and each slide as an attachment post tagged with that term. Per-slide settings (caption, link URL, target, alt text, crop) live in the attachment's ml-slider_settings postmeta. Slideshow-level settings (type: Flex, Nivo, Coin, Responsive; autoplay, delay, theme) sit in term meta on the ml-slider term. SleekAI reads those structures, so the system message includes the live list of slides, their captions, link targets, and the slider's playback settings.

That gives the bot real ground when a visitor asks about the slideshow on the page. "What does the second image link to?" returns the actual URL from postmeta. "Is the slider on autoplay?" answers from the term meta. "What's the caption on the studio shot?" returns the exact text the editor typed. None of it has to be guessed or rephrased from marketing copy.

Display conditions scope the bot to the templates or pages that embed a slider, multibot keeps a homepage marketing prompt separate from documentation, and conversation logs capture every slideshow-related question. The bot reads MetaSlider, MetaSlider still owns every write, every slide reorder, and every theme switch.

MetaSlider is the slideshow plugin many editorial sites and shops have running quietly above the fold. Five rotating slides, a couple of links, an autoplay timer. Visitors absorb half of it, skim the rest, and click into chat with a half-remembered detail.

A generic chatbot stares back blankly because it has no idea what was on the slide a moment ago. SleekAI changes that by reading the same data MetaSlider already stores: the ml-slider taxonomy term, the attached attachments, the per-slide ml-slider_settings postmeta with caption, link, and alt text. With that context in the prompt, the bot can quote the actual caption, point to the actual destination URL, confirm whether the rotation pauses on hover, and tell a visitor whether the slide they liked has scrolled past or is still in view.

None of this requires letting the LLM touch the slideshow. SleekAI reads, MetaSlider owns the writes. Slide order, theme switching, autoplay settings, and revisioning stay in the plugin where they belong.

Combined with multibot, a marketing-aware bot can stay scoped to the slideshow's templates while a separate documentation bot answers help-center questions with no slide context at all. Add bring-your-own-key billing and conversation logging that surfaces the captions visitors actually ask about, and the slideshow stops being a one-shot piece of marketing real estate and becomes something a visitor can interrogate, in plain language, and get accurate answers from.