Aether Search — Search & Relevance Playground

Live Query

Search Your Tenant Index

Searching live index

Type a query to run it across whichever Playground Mode is active. Default Search compares Exact / Autocomplete / Typo side by side, Faceted Navigation adds filters and facet distributions, Relevance Tuning lets you reorder results with attribute weights and ranking rules, and Hybrid Search (Vector) combines this query with a semantic vector using the HNSW nearest-neighbor index. The query box, geo context, and API key are shared across all of these.

Latitude Context geo

Reference point for distance ranking (decimal degrees)

Longitude Context geo

Pair with latitude to enable spatial distance scoring

User Token v18.7

When set, Default Search results are re-ranked to favour this user's previously clicked brands and categories.

Search Query / to focus · Esc to clear

Exact Match

What is Exact Match? Returns documents whose indexed terms contain the query token verbatim. Fastest mode — zero typo tolerance applied.

Type a query to see exact matches

Autocomplete

What is Autocomplete? Matches documents whose indexed terms start with the typed prefix. Ideal for search-as-you-type experiences.

Start typing to see prefix suggestions

Typo Correction

What is Typo Correction? Uses edit-distance (Levenshtein) to find documents even when the query contains spelling errors. Tolerance level is configurable in Relevance Tuning mode.

Intentional typos will still find results here

Exact Match

0ms

Waiting for query…

Autocomplete

0ms

Waiting for query…

Typo Correction

0ms

Waiting for query…

Developer Reference

Search Endpoints with Ranking Params

Every search endpoint accepts the same ranking parameters shown in the sandbox. Use typo_tolerance, custom_rank_field, and pinned_doc_ids to override defaults per request.

Exact match with ranking overrides

Relevance Sandbox

Tie-Breaking & Overrides

These apply to every Playground Mode since the search query is shared.

Typo Tolerance Levenshtein

Custom Rank Field

Numeric field used as secondary ranking signal

Pinned Document IDs boosted to top

Explain Ranking why did this rank here?

Schema Ranking Config

Weighted Rules & Attribute Ranking

Searchable attributes control which fields are tokenized and their relative weight. Custom ranking rules boost results based on numeric field values.

Searchable Attributes

field + weight

Custom Ranking Rules

field + direction + weight

Rules apply after text relevance. Direction: asc (lower wins), desc (higher wins).

Query Rules

pin + hide + block

Blocked Query Terms comma-separated

Ranking Formula v18.3

math expression

Functions: _val(field), _log(x), _sqrt(x), _days(x), _geoDistance(lat, lon). Example: _log(_val(popularity) + 1) / (_days(_val(created_at)) + 1). When set, this replaces Custom Ranking Rules.

Ranking Formula

Config JSON Preview

Experimentation

Ranking A/B Test

Splits live search traffic between two ranking configs. Config A serves the rest of traffic, Config B serves the fraction below. Use Copy Current Config to snapshot the sidebar's config above into either slot.

Traffic Fraction to Config B 0.0 – 1.0

Config A (JSON) majority traffic

Config B (JSON) experiment variant

A/B Test Result

Run an A/B test operation to see the server response.

Developer Reference

Additional API Endpoints

Examples below show every HTTP call this page makes. Copy any snippet directly into your project. Switch languages with the tabs — all examples use your live API key and current parameters.

Click Tracking & Analytics

Popular Query Suggestions

Co-item Recommendations

GET /hybrid_search/{query}

Hybrid Search — Keyword + Vector (HNSW)

Combines the keyword above with a semantic vector you paste in below, using an HNSW approximate-nearest-neighbor index server-side, then merges the two ranked lists with Reciprocal Rank Fusion (RRF).

Setup required: only documents that were saved with a vector embedding are matched by the vector side. Go to Tenant Ops → Operation 3 (Upsert a Document) and fill in the Vector Embedding field before testing here. If no documents have a vector yet, this mode still works but silently falls back to keyword-only results.

Query Vector comma or space separated numbers

Must be the same length as the vectors you attached to your documents.

Results to Return (k) max 100

RRF Constant default 60

Higher values flatten the boost that top-ranked keyword/vector hits get in the merged list.

Developer Reference

Hybrid Search Endpoint

The request reflects the query, vector, k, and RRF constant on the left.

Raw Response

Run a hybrid search to see the raw server JSON response.

Pro Tips

Geo ranking: Fill in the latitude/longitude fields above to enable distance-based scoring on geo documents.

Copy curl: Each column shows the exact HTTP request. Click Copy to get the ready-to-run curl command.

Switch modes: Faceted Navigation and Relevance Tuning reuse your current query — flip tabs to compare how each mode reorders the same results.

Authorization Gate

Search Your Tenant Index

Exact Match

Autocomplete

Typo Correction

Exact Match

Autocomplete

Typo Correction

Filter Format Reference

Faceted Search API

Exact Match

Autocomplete

Typo Correction

Search Endpoints with Ranking Params

Additional API Endpoints

Batch Query Testing

Hybrid Search — Keyword + Vector (HNSW)

Pro Tips