Oozor.ru, launched in 2018, is built for people who are looking for or creating attractive products for children across many categories: from toys to clothing and furniture.
Implement an FAQ bot that searches its own database, normalizes questions through AI, and is fully configurable from the store’s admin panel.
Inside the module, the client manages the greeting, fallback answer, FAQ topics, static answers, buttons, interface colors, and CSV exchange. On the storefront, the bot helps with common questions, and when needed it can transfer the user to support.
Client goal
The client wanted questions and answers to be managed directly in the CS-Cart admin panel. A potential buyer or seller asks in their own words, the system normalizes the question through AI, and then tries to match it against the existing FAQ base. At the same time, the Oozor team needed full control over the bot name, core prompt, texts, buttons, colors, and widget behavior.
What I did
I created a separate widget management interface: enable/disable, title, subtitle, greeting, default answer, and other basic settings are available without code changes.
The client maintains the FAQ table independently: entry status, topic, condition, buttons, and static answer. This removes the need for a developer for routine content updates.
Instead of a literal text search, the question is first converted into a more normalized form. That helps the bot find a relevant answer even when the user writes loosely or imprecisely.
I added settings for the main prompt, headings, buttons, color scheme, and bot/user message text. The widget can be adapted to the store’s style without editing the template.
On the storefront, the bot helps with products, comparisons, and order-related questions. If there is no answer, the flow can transfer the user to Oozor support.
For the team, I added FAQ import/export via CSV, as well as chat history export for a selected period.
Interfaces
The administrator works with the FAQ base and module settings, while the site visitor sees a fast branded interface that answers from a clear knowledge base.
The main result here is controllability. The team got not just a chat with a model, but its own support layer inside CS-Cart: the knowledge base, response logic, branding, and conversation history remained under the client’s control.
If your project has many repetitive questions and the answers need to stay controlled, this module gives you a hybrid flow: a managed FAQ base plus AI for matching wording.
When users regularly ask about listing terms, orders, returns, delivery, store rules, or product properties, part of the load should be handled automatically.
If the FAQ content changes often, every update should not require development. You need an interface where a manager can update questions and answers independently.
For support and marketplaces, it is important that the bot relies on verified wording, while AI is used for matching, not for improvisation.
The widget should look like part of the store: its own title, colors, texts, buttons, and conversation flow, without an external SaaS look.
If the base does not cover the question, the user should not hit a dead end. You need a fallback and a clear path to human support.
Chat history helps reveal FAQ gaps, understand repetitive questions, and improve support scenarios based on real interactions.
It is a chatbot for a CS-Cart store that does not answer by random generation, but from a managed database of questions and answers. This module helps reduce repetitive inquiries, answer product, order, and store-policy questions faster, and keep responses under business control.
Yes. In this case, the FAQ base is managed directly from the CS-Cart admin panel. A manager can add questions, enable or disable entries, and edit topics, conditions, buttons, and static answers without separate code changes.
First, the user’s question is normalized by AI so it does not depend on the exact wording. Then the system searches for the most relevant answer in the existing FAQ base by topic, conditions, and predefined scenarios. This makes the bot more accurate than a simple literal text match.
A regular widget usually responds with free generation and is harder to control for content quality. Here the logic is different: answers come from a managed knowledge base, and AI is used as a layer for matching the question to the right answer. That is safer for support, marketplace rules, and typical customer communication.
Yes. The module includes settings for the widget title and subtitle, welcome message, fallback answer, system prompt, interface colors, button text, and message text. This makes it possible to fit the AI assistant into the store’s style instead of leaving it as an external-looking block.
Yes. The team gets CSV exchange for the FAQ base and a separate export of chat history for a selected period. This is useful when the answer base needs to be updated in bulk, shared between staff members, or analyzed for real user requests.
Yes. If the bot cannot find a correct answer or the question goes beyond the knowledge base, you can show a fallback scenario and transfer the user to support. For an ecommerce project, this is important: the bot handles routine tasks, while complex requests stay with the human team.
Yes. Such a module is especially useful for marketplaces and Multi-Vendor projects that have repetitive questions from customers and sellers: about listing terms, products, orders, platform rules, and support. In these scenarios, a controlled knowledge base works much better than fully free-form generation.