Story-in-Story, a Smarter AI Assistant & Framework v2: Icecat Studio – Sprint 106 Release Notes
Sprint 106 touched several parts of the platform. The AI Assistant got a new engine and a new model. Story-in-Story, the Publication Status Auditor and the new user system were demoed. Framework v2 moved closer to becoming the default, and the native Story Builder gained customization options. Search, import and language handling got smaller consistency upgrades.
Story-in-Story – Nested Stories for Targeted Campaigns
Story-in-Story lets you embed one story inside another and configure it directly in the builder. This allows a broader campaigns approach with retailer-specific banners.
Global Master with local overrides: a global master story is the default fallback. If nothing is set for a shop, nothing renders. You can add localized versions on top for individual shops or markets.
Live API configuration: the nested story is resolved per shop through the live API based on retailer request.
This makes campaign work easier. You can swap in a seasonal or targeted banner per market without duplicating the whole story.
AI Assistant – A New Engine, Tested on Real Stories
We reworked the assistant, and this is the biggest change in the sprint.
From fixed pipeline to an agent that reads and reasons
Previously a separate pipeline built stories from a rigid template. Now the assistant reads the product page, looks at the images, picks modules from the layout and assembles the story itself.
It creates stories from a website link or an attached file (XML, JSON, CSV, HTML) and respects instructions like “no carousels” or “no size block”.
After creation you can keep working in the same chat. It remembers the source page, its plan and what is already done. You can ask it to edit texts, change module styles (background, spacing, borders, rounding), or add, remove and reorder modules.
It also creates and finds assets, changes story statuses, and answers questions from the documentation. It can tell you what from the source page didn’t make it into the story.
Before every change it shows what it will do and waits for your confirmation.
It fixes small issues itself before creating the story, such as mobile crops, alternating sides and text/video pairs.
Quality
We evaluated 16 models and tested 9 on about 100 stories from websites, JSON, XML and HTML files, images, chat text and several languages. Almost all were created on the first attempt.
On the same pages, the stories are fuller, and the text is taken verbatim. Overall the result feels better than the previous version, and the comparison on identical pages supports that.
Framework v2 – Moving Toward the Default
Work this sprint covered the practical side of moving to the new rendering framework.
HTML Pack on v2: a separate bundle configuration for the HTML package, a rewritten archive script and the zip output switched to the v2 bundle. We check that stories render offline and on a partner page, including the shadow DOM embed.
Partner compatibility: the archive structure stays compatible with current HTML Pack consumers such as Flixmedia, or any changes are documented for partners.
v2 by default: the router is set to serve v2 for new and existing stories.
Migration checklist: every framework touchpoint (embed, HTML Pack, Story Output API, builder preview) is listed with its current state, required changes, owner and how it will be verified.
Builder performance: we measure preview load time, re-render on edit and bundle size, and remove unnecessary framework reloads.
Story Builder – Central Section
The native, iframe-free story editor keeps growing. This sprint added:
Visual cues: highlighting for margins, paddings and modules.
Customization: a UI to switch features on and off.
Story-in-Story readiness: module initialization reworked to work with nested stories.
Unified fallbacks: one consistent design for empty-content placeholders.