FINDORA BY CLOUD ON · SMART SEARCH FOR E-SHOPS
Your customers type the way they talk. Your e‑shop understands.
Findora replaces your store’s search with one that understands Greek, Greeklish, typos and whole questions — with AI where it genuinely helps, and servers in the European Union.
Your store connects in 3 steps; Findora does the rest
You send your catalogue and orders and add one line of code. From then on, Findora answers every search, measures what sells and improves on its own — and you see it all in your portal.
Your e-shop
Findora servers in the EU
12 KPIs, reports, settings
Customers type however they like. They find what they’re looking for.
At its core runs a modern open-source search engine, Typesense — chosen after comparing it with OpenSearch and Meilisearch on a real pharmacy catalogue of 4,756 products and 2,215 real searches. On top of it, Findora adds its own Greek-language understanding layer:
Greeklish and typos
antiliako αντηλιακό (sunscreen)
Mistakes in accents and vowels, Latin characters — no problem.
Brands spelled the way they sound
λα ρος ποζε La Roche-Posay
And «σολγκαρ» → Solgar. Customers don’t need to know how it’s spelled.
Price and category in natural language
σαμπουάν κάτω από 10 € category + price range
“shampoo under €10” becomes filters, and customers see what the system understood.
Codes and barcodes
5201234567890 the exact product
EAN, a barcode from the phone’s camera, or the store’s own product code.
Your own rules
Synonyms, “pin this first”, hiding, redirecting to a page, and banners — per store, from the portal.
Correct results, even when categories are wrong
Our own product classification: at the pilot pharmacy, the correct product type in the top 24 results rose from 80% to 99%.
AI that only sells what you have
Findora’s AI is built on Anthropic’s Claude models. It always recommends products from your own catalogue — never made-up ones.
HYBRID SEARCH WITH MEANING
Customers type as if they were asking a person
“leg cramps at night” is not a product name. Findora handles it in three steps:
- The AI understands the need behind the sentence.
- The engine retrieves candidate products from the catalogue.
- The AI keeps only the ones that genuinely match.
Keywords give precision on names, codes and brands; the AI brings an understanding of meaning. Together, the best of both.
Shopping assistant
It chats with the visitor until they find the right product, with product cards and “add to basket”. If something isn’t available, it says so clearly and suggests alternatives that are in stock.
Product advisor
On every product page it answers “what is it for?” and “how do I use it?” from the description or the manufacturer’s official data sheet — citing its source.
Search by photo
The visitor takes a photo of a product and finds the same or a similar one in the store. The photo is not stored.
AI classification
Every new product is automatically given a type, an area and an audience, so results stay correct even when the store’s categories are wrong.
For pharmacies: the AI does not diagnose, does not give dosages, and refers customers to the pharmacist or doctor.
It finds its own mistakes — and fixes them
- 1
It picks up the signals
Every search and every assistant reply is checked: “nothing found”, the customer asked the same thing again, 👎, results nobody clicked.
- 2
A “reviewer” examines them overnight
A more powerful Claude model reviews those cases plus a sample of the rest, finds the cause and proposes a fix.
- 3
It fixes — with a human where needed
Synonyms that bring up products are added automatically. Every other change is approved by a person.
- 4
It tests before changing anything
Every change to the assistant is first run against a “golden set” of real questions with known correct answers. If even one gets worse, the change is not applied.
In the portal and the weekly email you can see how many were checked and what was improved.
Three steps, on any platform
For PrestaShop, WooCommerce, Magento 2 and OpenCart there are ready-made plugins that do all three for you — and the PrestaShop one also updates itself. Any other site connects with the same steps, manually.
STEP 1
Findora Feed
An XML file with your catalogue: names, prices, stock, categories, barcodes. It updates automatically, and every read produces a quality check — products without an image, wrong barcodes, categories that don’t exist.
STEP 2
One line of script
It brings the dropdown, the results, voice search, barcode scanning and the AI assistant. Lightweight — about 30 KB compressed — and served from the Cloudflare network.
STEP 3
Orders
One call on the order-confirmation page, so the revenue that search brings in can be measured.
Fast. And never in the way.
- In a load test with 500 requests per second and real searches, 95% of responses came back in 1.6 ms for the dropdown and 7.3 ms for the results page.
- Images are served as small thumbnails, so your site isn’t slowed down.
- Fail-open: if the service doesn’t respond within 800 ms or the subscription expires, the store automatically carries on with its own search. The site never “breaks”.
- Zero-downtime updates and regular backups.
Your data stays in Europe
- Servers in the EU: a data centre in Germany, with every service in its own isolated container. Designed for GDPR from the start.
- Cloudflare in front: HTTPS on every connection, attack protection, and per-IP and per-store limits against bots.
- Access: 2FA for administrators, user roles in the portal, and a log of every change.
- Domain verification: search is only activated once it’s proven that the store is yours.
- Visitors: searches are logged without IP addresses; in the assistant, emails, phone numbers, social-security numbers (AMKA) and IBANs are masked automatically before anything is stored.
- Payments: card details are handled directly by Revolut and never pass through our servers. Keys and secrets are stored encrypted.
08 — FOR TECHNICAL READERS
The infrastructure, piece by piece
Each piece runs in its own container with resource limits, so it can scale independently as traffic grows.
Questions we get asked
Which platforms does it work with?
There is a ready-made plugin for PrestaShop, WooCommerce, OpenCart and Magento 2 that makes the connection for you. For a custom site, the catalogue XML and one line of script are enough.
How much does it cost?
From €35 a month + VAT (Basic), €80 (Pro) and €180 (Enterprise), with 20% off an annual subscription. The AI add-ons are +€25 each and are included in Enterprise.
Can I try it?
Yes — 30 days free, with every feature and no card required.
What happens if the service doesn’t respond?
Your site carries on normally with its own search in under a second. The same applies if the subscription expires.
Do I need SoftOne or another of your products?
No. Findora is standalone and works with any e-shop. An ERP connection is only set up if you ask for it, on the Custom plan.
THE NEXT STEP
See Findora on your own store
On findora.tech you can try searches live, see prices from €35 a month and start a free 30-day trial — every feature, no card required.