01 / 01 screenshot-2026-08-24-at-12.05.55-2.png
Screen 1 of 1

Automated Book Listings for Ozon and Wildberries

 

For the bookstore, a system for automating the creation of product cards for the Ozon and Wildberries marketplaces was developed. The main objective of the — project is to reduce the manual work of operators when adding books to the market and accelerate the preparation of cards without loss of data quality.

Before automation, the process took a lot of time: the employee had to manually determine the book on the photo, find similar products, calculate the price, fill out the characteristics, prepare images, check the mandatory fields and only then send the goods to the marketplace.

The project aimed to make this process as automatic as possible: the operator uploads photos of the book and basic data, and the system itself recognizes the book, selects information, calculates the price, forms a card and prepares it for publication.

The challenge

It was necessary to create a service that takes photos of the book, article, size, weight and additional data, if any: ISBN, title, author, series, year of release and price.

Once loaded, the system shall:

  • identify a book by photo, title, author or ISBN
  • Find similar products on marketplaces and in book databases
  • price
  • fill out
  • prepare images and characteristics
  • Ship the goods to Ozon and Wildberries
  • Transfer disputed cases to manual inspection.

Special attention was paid to the fact that books are often used, old editions, without ISBN or with incomplete information. So the system didn’t have to depend on just one data source or just AI.

Implementation

The project was implemented as an internal CRM system for book processing and management of publications on marketplaces.

The logic of — is based on several processing steps:

1. Downloading a book

The operator uploads photos of the book: cover, root, additional pages and technical information. You can also specify the article, size, weight and other data, if they are known in advance.

After downloading the book falls into the processing queue. This allows the user not to wait for the completion of all operations on the page and allows you to process a large number of books gradually.

2. Recognition of a book with AI

The system analyzes the images and tries to determine the main data of the book:

  • name
  • writer
  • publisher
  • release
  • series
  • cover
  • page
  • description
  • other characteristics necessary for the product card.

AI is used not as the only source of truth, but as one of the stages of processing. If the data is insufficient or there are doubts, the system transfers the book to the status of manual verification.

3. Searching for similar products

After initial recognition, the system searches for the same or similar books in Ozon, Wildberries and book sources. Priority is given to exact matches: ISBN, title + author, cover match.

This is important because the same book can have different editions, different covers and different prices. To calculate the cost correctly, the system must find similar products, not just books with a similar name.

4. Price calculation

For the project, the logic of automatic price calculation was implemented. The system collects the prices of the found analogues, calculates the average or median cost and applies a given formula.

The formula is made in the settings so that it can be changed without reworking the code. This is convenient, because discounts, commissions and pricing policies of marketplaces can change.

If the system does not find enough matches or can not confidently calculate the price, the book is sent for manual verification.

5. Automatic filling of the card

After recognition and enrichment of data, the system forms a product card. It includes the main fields:

  • articular
  • name
  • barcode
  • category
  • product
  • price
  • discount
  • size
  • weight
  • picture
  • writer
  • publisher
  • series
  • release
  • tongue
  • type
  • cover
  • age restrictions
  • preservation
  • hashtag
  • abstract.

Separately, the requirements of marketplaces for card filling, including the content rating Ozon, were taken into account. The better the characteristics, images and description are filled, the higher the quality of the card and the less likely there are problems when publishing.

6. Publication on Ozon and Wildberries

After preparation, the card can be sent to marketplaces through API. For each site, its requirements for categories, characteristics, images, remains and publication statuses are taken into account.

Work with warehouse logic was also provided: shelves, residues, statuses of goods and further synchronization after sale.

Manual check and error control

It is impossible to fully automate such a process without checking, especially when it comes to old books, different publications and incomplete data. Therefore, the draft provides for the statuses and logic of manual verification.

The book is sent for verification if:

  • not found
  • Several similar options have been found.
  • failed to determine the price
  • Mandatory characteristics are not completed
  • The error was API.
  • The AI could not confidently identify the book.

The administrative panel displays processing statuses, queue of books, errors and logs. This allows the operator to quickly understand at what stage the problem occurred and what needs to be fixed.

Technical part

The project was developed at Laravel. For background processing, queues and individual tasks were used to ensure that the heavy operations of — recognition, search, price calculation and publication of — were performed stably and did not block the interface.

The project used:

  • Laravel for backend logic
  • queue
  • integration with API Ozon and Wildberries
  • AI for recognition and description formation
  • administration panel for book management
  • logging
  • settings for formulas, coefficients, templates and API keys.

Special attention was paid to the fact that the system was not just a « bot on AI», but a full-fledged working tool. If the AI fails, backup logic is turned on: search by ISBN, search by name, search by image and manual verification.

The result

As a result, a system was created that helps the bookstore to prepare goods for sale on marketplaces faster.

The project solves several business problems at once:

  • Reduces manual labor of operators
  • Accelerates the addition of new books
  • Reduces errors when filling out cards
  • Helps you calculate the price automatically
  • Improves the quality of product cards
  • Maintains control over disputed cases
  • Simplifies publication on Ozon and Wildberries.

The main value of the — project is not just in automatic unloading of goods, but in creating a full-fledged process: from a photo of a book to a finished card on the marketplace.

This system is especially useful for stores where there are many unique products, different states of books, incomplete data and a large amount of manual processing. Automation allows you to scale sales without proportionately increasing the burden on employees.

Related directions

AI/RAG portfolio ·