Product search

Jelou Shop ships with two product search engines. Both receive what the user types (for example, from the webview or the AI Agent tools) and return the most relevant products, but they solve the problem in different ways:

​

Hybrid search

Uses artificial intelligence to understand the meaning of what the user types and compare it with the meaning of each product in the catalog. In parallel it also searches by keywords, and merges both results into a single ranking. Technically, the meaning is represented with semantic vectors (embeddings).

​

How it works

By understanding the meaning of the query, it finds products even when the user doesn’t use the exact words in the catalog: “something for the cold” can return jackets and coats even if none of them mention “cold”.

​

Advantages

• Understands natural language and purchase intent, not just exact words.
• Finds products even when the user’s vocabulary differs from the catalog’s.
• Reduces empty results for general or descriptive queries.

​

Limitations

• It’s slower: each query goes through artificial intelligence services before returning results.
• It has an additional cost per query (AI services).
• It’s less precise with exact codes, such as SKUs.
• The ranking is less predictable: two similar queries may rank differently.

​

Text search

Searches directly over the catalog information, without artificial intelligence services. Before searching, it cleans and prepares the query, then matches it against each product’s name, SKU, categories, and description. Technically, it relies on full-text search (FTS).

​

How it works

Highlights:

• Ignores accents and casing: cafe matches café.
• Understands singular and plural: t-shirt matches t-shirts.
• All terms must match: searching red t-shirt doesn’t return everything red or every t-shirt — only red t-shirts.
• Category aware: searching t-shirt finds products in the “T-shirts” category even if the product name doesn’t mention it.
• Your store’s own synonyms: you can teach it that tv = television.
• Words to ignore: it discards words that only add noise to the search, such as “want” or “cheap”.
• Recognizes your store’s brands: you can register your brands so it detects them in the query and uses them to filter the results.
• Typo tolerance (optional): Xioami matches Xiaomi.

​

Advantages

• Very fast and low cost: it doesn’t depend on external services.
• Precise: excellent for SKUs, product names, and catalogs with technical vocabulary.
• Consistent results: the same search always returns the same products.
• Customizable per store: synonyms, words to ignore, and brands.

​

Limitations

• It only finds what is written: if the term doesn’t appear in the product’s name, SKU, category, or description (nor in the configured synonyms), there are no results.
• It doesn’t understand descriptive queries like “something for the cold” or “a gift for mom”.
• Your customers’ particular vocabulary may require setting up synonyms.

​

Comparison

​

Which one should you use?

If you’re not sure which one fits your catalog, contact the Jelou team.