Fields & filters
The structure that lets your agent search with precision — typed fields and the filter vocabulary that narrows results to exactly the right ones.
Searching by meaning is powerful, but sometimes a customer wants something precise: black, size 38, under 500 TL, in stock. That precision comes from fields and filters — the structured side of a collection that lets the agent narrow thousands of items down to the handful that actually fit.
If you chose a ready-made template, the important fields and filters are already set up. This page explains what they are, so you understand what the agent can do — and so you can build your own with the Custom template.
Fields: the structure of an item
Each item in a collection is made of fields — named pieces of information. A product has a title, a price, a list of available sizes, a stock count. Each field has a type, and the type decides how the agent can search and filter on it:
- Text — names, codes, titles, URLs. The agent can match exactly, or by what a value starts with, ends with, or contains.
- Prose (rich text) — long written content. This is the part searched by meaning, and it's automatically broken into passages for precise retrieval.
- Lists (of text) — tags, sizes, colors, amenities. The agent can check whether an item has any or all of a set of values.
- Numbers — prices, stock counts, ratings, capacities. The agent can compare: more than, less than, between.
- True/false — flags like in stock. Simple yes/no filtering.
- Dates — for filtering by before/after a point in time.
- Location (coordinates) — for "within X km of here" searches.
Picking the right type pays off: store a price as a number and the agent can do "under 500"; store it as text and it can only match it exactly. The templates already make these choices well — you mostly need this when building Custom.
Filters: narrowing the results
A filter is a rule that narrows what comes back. The agent builds filters automatically from what the customer says — "black XL shirts under 50 TL" becomes color is black, size is XL, price under 50 — and applies them to the search.
The available checks, in plain terms:
| Filter | What it does | Example |
|---|---|---|
| Equal / Not equal | Matches (or excludes) an exact value | brand equals Nike |
| Greater / Less than (or or equal) | Compares numbers and dates | price less than 500 |
| Contains any | Has at least one of these values | size is any of S, M |
| Contains all | Has every one of these values | features include all of parking, garden |
| Starts with / Ends with / Contains | Partial text matching | SKU starts with 6Y284 |
| Is empty | The field has no value | image is empty |
| Within range (location) | Near a point on the map | within 5 km of a location |
The beauty is that the customer never sees any of this. They type naturally, and the agent translates their words into the right filters behind the scenes.
How filters and meaning work together
Filters don't replace meaning-based search — they sharpen it. The agent does both at once: it searches for what the customer means, and filters to what qualifies.
"a warm waterproof jacket for hiking, under 2,000 TL, in stock"
- Meaning finds jackets that are warm, waterproof, and hiking-appropriate.
- Filters keep only those under 2,000 TL and in stock.
The result is a short, relevant list — not a thousand jackets, and not the wrong three. This is the whole point of structuring your collection well.
For a store, the single biggest win is getting price, stock, and variant (size/color) fields right. They're what turn the agent from "here are some products" into "here are the three that match exactly what you asked for, and they're in stock."
Variants, handled
For products, Vivollo does something thoughtful with variants. A single shirt might come in five sizes and four colors — and the agent can filter on those combos intelligently. When a customer narrows by two or more attributes (say small and red), it filters to items that genuinely have that combination available, not just items that have a small somewhere and red somewhere. That's the difference between a confident "yes, in stock" and a disappointing "oh, not in that size."
Next, let's get content into your collection and keep it fresh.