Don't just search. Understand.
XTAL combines large language models, vector search, and precision matching into one seamless experience. We call it Full-Spectrum Search™.
Finally, a search tool that doesn't create more work
Most 'powerful' search tools are powerful the way a fighter jet is powerful: technically impressive, requires years of training, crashes if you push the wrong button.
Manually map 'couch' → 'sofa', 'sneakers' → 'trainers', hundreds of variations
XTAL already knows they're the same thing
JSON configs, weight multipliers, A/B test every change
Tell it in English: 'prioritize new arrivals and in-stock items'
Adjust BM25 parameters, field weights, proximity scoring
It learns from your catalog structure automatically
Enrich every SKU with custom attributes for ML models
Works with your existing product data as-is
Faceted navigation setup, filter dependencies, UI configuration
Constraints extracted from natural language. Filters become optional
Null results pages, redirect rules, 'did you mean' logic
LLM understands intent even when queries are weird
Traditional search tools like Algolia, Searchspring, and Klevu are powerful if you have an engineer to configure them. XTAL gives you the results without the PhD.
One search bar. Any kind of query.
Your customers don't all search the same way. Some know exactly what they want. Some are exploring. Most are somewhere in between. Traditional search forces you to pick a lane: keyword OR semantic. XTAL does both, simultaneously.
Precision
Part numbers, SKUs, exact product names. When they know what they need, we get out of the way
Intent
Natural language, use-case descriptions. Semantic understanding finds products keyword search would miss
Hybrid
Brand + use case in one query. XTAL parses both signals and ranks accordingly
Constrained
Multiple constraints extracted and applied: color, occasion, price. No filters needed
Keyword precision meets semantic intelligence. No mode-switching required.
Three layers working together
Every query passes through a pipeline designed to maximize relevance without requiring configuration. Here's what happens in the ~200ms between keystroke and results.
Understand/ LLM Query Intelligence
A large language model analyzes the query in real-time, identifying:
- Intent type: browsing, researching, or ready to buy
- Explicit constraints: price ranges, colors, sizes, quantities
- Implicit context: occasion, recipient, use case
- Query structure: exact match signals vs. exploratory language
{
"input": "gift for mom who likes gardening under 50",
"output": {
"intent": "Gift purchase",
"recipient": "Mother",
"interest": "Gardening",
"priceMax": 50,
"occasion": "Likely birthday or Mother's Day"
}
}Match/ Full-Spectrum Retrieval
The parsed query hits multiple retrieval paths simultaneously:
- Vector similarity: finds products by meaning, not just words
- Keyword precision: exact matches for SKUs, brands, part numbers
- Attribute filtering: applies extracted constraints (price, color, etc.)
- Fuzzy matching: handles typos and spelling variations
We use a hybrid scoring model that dynamically weights each retrieval path based on query characteristics. SKU lookups prioritize keyword matching. Intent queries prioritize semantic understanding. Most queries blend both approaches.
Refine/ Business Logic Layer
A reasoning layer applies your merchandising priorities in plain English:
- Inventory awareness: deprioritize out-of-stock without hiding
- Margin optimization: boost higher-margin alternatives
- Seasonal relevance: surface timely products automatically
- Brand rules: handle exclusives, restrictions, preferences
In traditional search, this requires boost rules, custom scoring functions, and constant tuning. In XTAL, you just tell it what you want.
{
"rule": "For any search related to 'gifts' or 'presents', always prioritize the Holiday Bundle Collection",
"scope": "Global",
"status": "Active"
}You don't need to understand any of this to use XTAL. But if you're the type who likes knowing how things work, well, now you know.
See the difference on your own catalog
We'll connect to your store, run real searches side-by-side, and show you exactly what your customers are missing.