Search that understands what people mean.

Relatiq Search turns products, articles and documents into vector embeddings and ranks results by meaning, not matching words. Connect your data, get an API key and ship natural-language search in days.

Try it on a food catalogue

Type a dish, a diet or an occasion. Switch profiles to see how the same query ranks differently.

Search profile
type to search

How it works

Three steps between your data and relevant results.

  1. Index

    Documents are converted into embeddings that capture meaning and context. New items are searchable in real time.

  2. Understand

    Language models read each query for intent, synonyms and context, in any language.

  3. Rank

    Results are ordered by semantic similarity, blended with keyword matching where precision matters.

Built for production

Everything you need to run search on a live catalogue or knowledge base.

Custom embeddings

Models tuned to your domain vocabulary, trained on our own GPU server.

Real-time indexing

Add, update and delete documents without reindexing.

Hybrid ranking

Semantic and keyword scores combined per search profile.

Mixed content

Text and documents in one index.

Filters and facets

Narrow results by any field alongside semantic ranking.

Scale

Index size grows with your reserved capacity, from small catalogues to tens of millions of documents.

Dedicated GPU infrastructure

Indexing and custom embeddings run on our own EU-hosted GPU server, not shared third-party capacity.

Multilingual by default

Queries and documents are understood for intent and meaning in any language, not just English.

Search profiles

Switch ranking strategy per use case — balanced, description-weighted, hybrid or name-focused.

query latency, Growth to Starter
60–150 ms
indexing throughput
~500/s per vCPU
NDCG@10, measured on your data
per catalogue
uptime target, Growth and up
99.5%

Where teams use it

E-commerce

Shoppers describe what they want and results match the intent.

"comfortable running shoes for beginners"

Knowledge bases

Employees and customers find answers regardless of phrasing.

"how do I reset two-factor login"

Enterprise data

One semantic layer across databases, documents and systems.

"contracts renewing next quarter with penalty clauses"

Index and query with one API.

REST API. Create indexes, define search profiles and query from any stack.

search · request
curl https://api.relatiq.io/v1/search \
  -H "Authorization: Bearer $RELATIQ_KEY" \
  -d '{"index": "products", "query": "tea with milk", "profile": "mix"}'
search · response
{
  "took_ms": 23,
  "results": [
    { "id": "sku-1042", "title": "Masala chai latte mix", "score": 0.91 },
    { "id": "sku-2210", "title": "Hong Kong milk tea", "score": 0.89 }
  ]
}

Plans that grow with usage

Each tier reserves a fixed share of compute and memory on your index. Outgrow it and we move you up, no surprise bills.

Starter
€29/mo
1 vCPU, 2 GB — prototypes and small catalogues
Growth
€49/mo
2 vCPU, 4 GB — production sites with steady traffic
Scale
from €79/mo
4 vCPU, 8 GB and up — custom volume, SLA and support
upgrades are prorated for the month you move tiersprices exclude VAT

Put semantic search on your site.

Talk to us about volume, custom models and SLA.