yexinru/doc-exports
1
0
Fork 0
forked from docs/doc-exports
Code Pull Requests Activity
Files
a2e28c76231c99d8bbac87147becbb2b7e6be32e
doc-exports/docs/css/umn/css_01_0119.html
zhengxiu 93d856d5c5 css umn 25.6.0 version 
Reviewed-by: Pruthi, Vineet <vineet.pruthi@t-systems.com>
Co-authored-by: zhengxiu <zhengxiu@huawei.com>
Co-committed-by: zhengxiu <zhengxiu@huawei.com>
2025-11-25 11:34:43 +00:00

497 lines
54 KiB
HTML
Raw Blame History

This file contains ambiguous Unicode characters

This file contains Unicode characters that might be confused with other characters. If you think that this is intentional, you can safely ignore this warning. Use the Escape button to reveal them.

<a name="EN-US_TOPIC_0000001955726494"></a><a name="EN-US_TOPIC_0000001955726494"></a>
<h1 class="topictitle1">Performing Vector Search</h1>
<div id="body0000001955726494"><div class="p" id="EN-US_TOPIC_0000001955726494__p2536314165316">The CSS vector database supports a variety of query methods, including standard query, hybrid query, script_score query, rescore query, and painless syntax extension. They enable efficient vector data search by accommodating diverse search needs.<ul id="EN-US_TOPIC_0000001955726494__ul1415819211585"><li id="EN-US_TOPIC_0000001955726494__li11581826586"><a href="#EN-US_TOPIC_0000001955726494__en-us_topic_0000001268154489_section344710865418">Standard Query</a>: retrieves documents that are most similar to the query vector.</li><li id="EN-US_TOPIC_0000001955726494__li1915815215814"><a href="#EN-US_TOPIC_0000001955726494__en-us_topic_0000001268154489_section102341611718">Hybrid Query</a>: combines vector search with traditional OpenSearch queries, such as pre-filtering and Boolean queries.</li><li id="EN-US_TOPIC_0000001955726494__li91581921580"><a href="#EN-US_TOPIC_0000001955726494__en-us_topic_0000001268154489_section4946246546">Script Score Query</a>: enables custom similarity calculations for vector searches by executing a custom script.</li><li id="EN-US_TOPIC_0000001955726494__li11158920581"><a href="#EN-US_TOPIC_0000001955726494__en-us_topic_0000001268154489_section107953451661">Rescore Query</a>: rescores and reranks the top results returned by an initial query to improve recall.</li><li id="EN-US_TOPIC_0000001955726494__li121593215820"><a href="#EN-US_TOPIC_0000001955726494__section102071126171611">Painless Syntax Extension</a>: allows the use of vector distance or similarity calculation functions in custom scripts.</li></ul>
</div>
<div class="section" id="EN-US_TOPIC_0000001955726494__en-us_topic_0000001268154489_section344710865418"><a name="EN-US_TOPIC_0000001955726494__en-us_topic_0000001268154489_section344710865418"></a><a name="en-us_topic_0000001268154489_section344710865418"></a><h4 class="sectiontitle">Standard Query</h4><p id="EN-US_TOPIC_0000001955726494__en-us_topic_0000002319719730_p116718237114">Standard query is used to retrieve documents that are most similar to the query vector.</p>
<p id="EN-US_TOPIC_0000001955726494__en-us_topic_0000002319719730_en-us_topic_0000001268154489_p10699133014548">The following command will return <em id="EN-US_TOPIC_0000001955726494__en-us_topic_0000002319719730_i15619747289542">k</em> (specified by <strong id="EN-US_TOPIC_0000001955726494__en-us_topic_0000002319719730_b19427032809542">size</strong>/<strong id="EN-US_TOPIC_0000001955726494__en-us_topic_0000002319719730_b15868144189542">topk</strong>) records that are the closest matches to the query vector.</p>
<pre class="screen" id="EN-US_TOPIC_0000001955726494__en-us_topic_0000002319719730_en-us_topic_0000001268154489_screen1466920229569">POST <i><span class="varname" id="EN-US_TOPIC_0000001955726494__en-us_topic_0000002319719730_varname47651948195211">my_index</span></i>/_search
{
"size":2,
"_source": false,
"query": {
"vector": {
"my_vector": {
"vector": [1, 1],
"topk":2
}
}
}
}</pre>
<div class="tablenoborder"><table cellpadding="4" cellspacing="0" summary="" id="EN-US_TOPIC_0000001955726494__en-us_topic_0000002319719730_en-us_topic_0000001268154489_table112016411577" frame="border" border="1" rules="all"><caption><b>Table 1 </b>Parameters for standard query</caption><thead align="left"><tr id="EN-US_TOPIC_0000001955726494__en-us_topic_0000002319719730_en-us_topic_0000001268154489_row120119417572"><th align="left" class="cellrowborder" valign="top" width="23%" id="mcps1.3.2.5.2.5.1.1"><p id="EN-US_TOPIC_0000001955726494__en-us_topic_0000002319719730_en-us_topic_0000001268154489_p152017415713">Parameter</p>
</th>
<th align="left" class="cellrowborder" valign="top" width="14.000000000000002%" id="mcps1.3.2.5.2.5.1.2"><p id="EN-US_TOPIC_0000001955726494__en-us_topic_0000002319719730_p1763712241312">Mandatory</p>
</th>
<th align="left" class="cellrowborder" valign="top" width="15%" id="mcps1.3.2.5.2.5.1.3"><p id="EN-US_TOPIC_0000001955726494__en-us_topic_0000002319719730_p166237264313">Type</p>
</th>
<th align="left" class="cellrowborder" valign="top" width="48%" id="mcps1.3.2.5.2.5.1.4"><p id="EN-US_TOPIC_0000001955726494__en-us_topic_0000002319719730_en-us_topic_0000001268154489_p02013445711">Description</p>
</th>
</tr>
</thead>
<tbody><tr id="EN-US_TOPIC_0000001955726494__en-us_topic_0000002319719730_row26533154"><td class="cellrowborder" valign="top" width="23%" headers="mcps1.3.2.5.2.5.1.1 "><p id="EN-US_TOPIC_0000001955726494__en-us_topic_0000002319719730_p1881333553">size</p>
</td>
<td class="cellrowborder" valign="top" width="14.000000000000002%" headers="mcps1.3.2.5.2.5.1.2 "><p id="EN-US_TOPIC_0000001955726494__en-us_topic_0000002319719730_p1482331954">Yes</p>
</td>
<td class="cellrowborder" valign="top" width="15%" headers="mcps1.3.2.5.2.5.1.3 "><p id="EN-US_TOPIC_0000001955726494__en-us_topic_0000002319719730_p69203313518">Integer</p>
</td>
<td class="cellrowborder" valign="top" width="48%" headers="mcps1.3.2.5.2.5.1.4 "><p id="EN-US_TOPIC_0000001955726494__en-us_topic_0000002319719730_p13915331450">Number of search results to return.</p>
<p id="EN-US_TOPIC_0000001955726494__en-us_topic_0000002319719730_p12477204662018">Default value: <strong id="EN-US_TOPIC_0000001955726494__en-us_topic_0000002319719730_b8755580319542">10</strong></p>
</td>
</tr>
<tr id="EN-US_TOPIC_0000001955726494__en-us_topic_0000002319719730_row18131348871"><td class="cellrowborder" valign="top" width="23%" headers="mcps1.3.2.5.2.5.1.1 "><p id="EN-US_TOPIC_0000001955726494__en-us_topic_0000002319719730_p31321448473">_source</p>
</td>
<td class="cellrowborder" valign="top" width="14.000000000000002%" headers="mcps1.3.2.5.2.5.1.2 "><p id="EN-US_TOPIC_0000001955726494__en-us_topic_0000002319719730_p513313481179">No</p>
</td>
<td class="cellrowborder" valign="top" width="15%" headers="mcps1.3.2.5.2.5.1.3 "><p id="EN-US_TOPIC_0000001955726494__en-us_topic_0000002319719730_p171331448370">Boolean</p>
</td>
<td class="cellrowborder" valign="top" width="48%" headers="mcps1.3.2.5.2.5.1.4 "><p id="EN-US_TOPIC_0000001955726494__en-us_topic_0000002319719730_p2013311481677">Whether to return the source text in documents. To reduce data transmission and improve query performance, set this parameter to <strong id="EN-US_TOPIC_0000001955726494__en-us_topic_0000002319719730_b139371755131414">false</strong>.</p>
<div class="p" id="EN-US_TOPIC_0000001955726494__en-us_topic_0000002319719730_p0836919144516">Value range:<ul id="EN-US_TOPIC_0000001955726494__en-us_topic_0000002319719730_ul74052714513"><li id="EN-US_TOPIC_0000001955726494__en-us_topic_0000002319719730_li1540927154519"><strong id="EN-US_TOPIC_0000001955726494__en-us_topic_0000002319719730_b36642032104917">true</strong> (default): Returns the source text.</li><li id="EN-US_TOPIC_0000001955726494__en-us_topic_0000002319719730_li940627184516"><strong id="EN-US_TOPIC_0000001955726494__en-us_topic_0000002319719730_b20329739135016">false</strong>: Not to return the source text.</li></ul>
</div>
</td>
</tr>
<tr id="EN-US_TOPIC_0000001955726494__en-us_topic_0000002319719730_row12407401962"><td class="cellrowborder" valign="top" width="23%" headers="mcps1.3.2.5.2.5.1.1 "><p id="EN-US_TOPIC_0000001955726494__en-us_topic_0000002319719730_p124014401260">query</p>
</td>
<td class="cellrowborder" valign="top" width="14.000000000000002%" headers="mcps1.3.2.5.2.5.1.2 "><p id="EN-US_TOPIC_0000001955726494__en-us_topic_0000002319719730_p22407408610">Yes</p>
</td>
<td class="cellrowborder" valign="top" width="15%" headers="mcps1.3.2.5.2.5.1.3 "><p id="EN-US_TOPIC_0000001955726494__en-us_topic_0000002319719730_p2024014018611">Map</p>
</td>
<td class="cellrowborder" valign="top" width="48%" headers="mcps1.3.2.5.2.5.1.4 "><p id="EN-US_TOPIC_0000001955726494__en-us_topic_0000002319719730_p569812262249">Specifies the query vector.</p>
<p id="EN-US_TOPIC_0000001955726494__en-us_topic_0000002319719730_p986113011243">Parameters:</p>
<p id="EN-US_TOPIC_0000001955726494__en-us_topic_0000002319719730_p16849539192517"><strong id="EN-US_TOPIC_0000001955726494__en-us_topic_0000002319719730_b18181755355">vector</strong> (mandatory): indicates a vector query (vector similarity-based search), including the vector field and query vector value.</p>
<p id="EN-US_TOPIC_0000001955726494__en-us_topic_0000002319719730_p816801552413"><strong id="EN-US_TOPIC_0000001955726494__en-us_topic_0000002319719730_b913132365619">my_vector</strong> (mandatory): queried vector field (for example, <span class="parmvalue" id="EN-US_TOPIC_0000001955726494__en-us_topic_0000002319719730_parmvalue159302219283"><b>my_vector</b></span>).</p>
</td>
</tr>
<tr id="EN-US_TOPIC_0000001955726494__en-us_topic_0000002319719730_row83814321817"><td class="cellrowborder" valign="top" width="23%" headers="mcps1.3.2.5.2.5.1.1 "><p id="EN-US_TOPIC_0000001955726494__en-us_topic_0000002319719730_p123913324819">vector (sub-parameter)</p>
</td>
<td class="cellrowborder" valign="top" width="14.000000000000002%" headers="mcps1.3.2.5.2.5.1.2 "><p id="EN-US_TOPIC_0000001955726494__en-us_topic_0000002319719730_p23983212812">Yes</p>
</td>
<td class="cellrowborder" valign="top" width="15%" headers="mcps1.3.2.5.2.5.1.3 "><p id="EN-US_TOPIC_0000001955726494__en-us_topic_0000002319719730_p73918321187">Array/String</p>
</td>
<td class="cellrowborder" valign="top" width="48%" headers="mcps1.3.2.5.2.5.1.4 "><p id="EN-US_TOPIC_0000001955726494__en-us_topic_0000002319719730_p647564453011">Query vector value. It is used to calculate the similarity between indexed vectors and the query vector.</p>
<p id="EN-US_TOPIC_0000001955726494__en-us_topic_0000002319719730_p88291013103120">The value can be an array (for example, [1, 1]) or Base64-encoded value (for example, <span class="parmvalue" id="EN-US_TOPIC_0000001955726494__en-us_topic_0000002319719730_parmvalue182541985329"><b>AAABAAACAAAD</b></span>).</p>
</td>
</tr>
<tr id="EN-US_TOPIC_0000001955726494__en-us_topic_0000002319719730_row16292516818"><td class="cellrowborder" valign="top" width="23%" headers="mcps1.3.2.5.2.5.1.1 "><p id="EN-US_TOPIC_0000001955726494__en-us_topic_0000002319719730_p56213291915">topk</p>
</td>
<td class="cellrowborder" valign="top" width="14.000000000000002%" headers="mcps1.3.2.5.2.5.1.2 "><p id="EN-US_TOPIC_0000001955726494__en-us_topic_0000002319719730_p1829125116816">Yes</p>
</td>
<td class="cellrowborder" valign="top" width="15%" headers="mcps1.3.2.5.2.5.1.3 "><p id="EN-US_TOPIC_0000001955726494__en-us_topic_0000002319719730_p42955110814">Integer</p>
</td>
<td class="cellrowborder" valign="top" width="48%" headers="mcps1.3.2.5.2.5.1.4 "><p id="EN-US_TOPIC_0000001955726494__en-us_topic_0000002319719730_p267231613315">The number of the most similar or relevant results to be returned.</p>
<p id="EN-US_TOPIC_0000001955726494__en-us_topic_0000002319719730_p8820452331">Default value: same as <span class="parmname" id="EN-US_TOPIC_0000001955726494__en-us_topic_0000002319719730_parmname1929142716339"><b>size</b></span>.</p>
</td>
</tr>
<tr id="EN-US_TOPIC_0000001955726494__en-us_topic_0000002319719730_row842795112810"><td class="cellrowborder" valign="top" width="23%" headers="mcps1.3.2.5.2.5.1.1 "><p id="EN-US_TOPIC_0000001955726494__en-us_topic_0000002319719730_p193995513415">ef</p>
</td>
<td class="cellrowborder" valign="top" width="14.000000000000002%" headers="mcps1.3.2.5.2.5.1.2 "><p id="EN-US_TOPIC_0000001955726494__en-us_topic_0000002319719730_p2428185118818">No</p>
</td>
<td class="cellrowborder" valign="top" width="15%" headers="mcps1.3.2.5.2.5.1.3 "><p id="EN-US_TOPIC_0000001955726494__en-us_topic_0000002319719730_p121017815399">Integer</p>
</td>
<td class="cellrowborder" valign="top" width="48%" headers="mcps1.3.2.5.2.5.1.4 "><p id="EN-US_TOPIC_0000001955726494__en-us_topic_0000002319719730_p1887713228362">How many nearest neighbors to explore when inserting a new vector into the graph. A larger value indicates a higher query accuracy yet slower query speed.</p>
<p id="EN-US_TOPIC_0000001955726494__en-us_topic_0000002319719730_p1225115180408">This parameter is available only when <span class="parmname" id="EN-US_TOPIC_0000001955726494__en-us_topic_0000002319719730_parmname13251218184015"><b>algorithm</b></span> is set to GRAPH, GRAPH_PQ, GRAPH_SQ8, or GRAPH_SQ4.</p>
<p id="EN-US_TOPIC_0000001955726494__en-us_topic_0000002319719730_p976082693617">Value range: 0–100000</p>
<p id="EN-US_TOPIC_0000001955726494__en-us_topic_0000002319719730_p2042813511386">Default value: <strong id="EN-US_TOPIC_0000001955726494__en-us_topic_0000002319719730_b1931737559542">200</strong></p>
</td>
</tr>
<tr id="EN-US_TOPIC_0000001955726494__en-us_topic_0000002319719730_en-us_topic_0000001268154489_row39015465716"><td class="cellrowborder" valign="top" width="23%" headers="mcps1.3.2.5.2.5.1.1 "><p id="EN-US_TOPIC_0000001955726494__en-us_topic_0000002319719730_p143905523418">max_scan_num</p>
</td>
<td class="cellrowborder" valign="top" width="14.000000000000002%" headers="mcps1.3.2.5.2.5.1.2 "><p id="EN-US_TOPIC_0000001955726494__en-us_topic_0000002319719730_p149563544119">No</p>
</td>
<td class="cellrowborder" valign="top" width="15%" headers="mcps1.3.2.5.2.5.1.3 "><p id="EN-US_TOPIC_0000001955726494__en-us_topic_0000002319719730_p1195614519418">Integer</p>
</td>
<td class="cellrowborder" valign="top" width="48%" headers="mcps1.3.2.5.2.5.1.4 "><p id="EN-US_TOPIC_0000001955726494__en-us_topic_0000002319719730_p144561915404">Maximum number of graph nodes to scan during search. A larger value indicates a higher query accuracy yet slower query speed.</p>
<p id="EN-US_TOPIC_0000001955726494__en-us_topic_0000002319719730_p635195504010">This parameter is available only when <span class="parmname" id="EN-US_TOPIC_0000001955726494__en-us_topic_0000002319719730_parmname17541532168"><b>algorithm</b></span> is set to GRAPH, GRAPH_PQ, GRAPH_SQ8, or GRAPH_SQ4.</p>
<p id="EN-US_TOPIC_0000001955726494__en-us_topic_0000002319719730_p1967361116406">Value range: 0–1000000</p>
<p id="EN-US_TOPIC_0000001955726494__en-us_topic_0000002319719730_p155710277406">Default value: <strong id="EN-US_TOPIC_0000001955726494__en-us_topic_0000002319719730_b18571901969542">10000</strong></p>
</td>
</tr>
<tr id="EN-US_TOPIC_0000001955726494__en-us_topic_0000002319719730_en-us_topic_0000001268154489_row0480125417579"><td class="cellrowborder" valign="top" width="23%" headers="mcps1.3.2.5.2.5.1.1 "><p id="EN-US_TOPIC_0000001955726494__en-us_topic_0000002319719730_p139165513419">nprobe</p>
</td>
<td class="cellrowborder" valign="top" width="14.000000000000002%" headers="mcps1.3.2.5.2.5.1.2 "><p id="EN-US_TOPIC_0000001955726494__en-us_topic_0000002319719730_p145620913410">No</p>
</td>
<td class="cellrowborder" valign="top" width="15%" headers="mcps1.3.2.5.2.5.1.3 "><p id="EN-US_TOPIC_0000001955726494__en-us_topic_0000002319719730_p05659104120">Integer</p>
</td>
<td class="cellrowborder" valign="top" width="48%" headers="mcps1.3.2.5.2.5.1.4 "><p id="EN-US_TOPIC_0000001955726494__en-us_topic_0000002319719730_p20228184614120">Number of centroids to explore during an IVF index query. A larger value indicates a higher query accuracy yet slower query speed.</p>
<p id="EN-US_TOPIC_0000001955726494__en-us_topic_0000002319719730_p12507171516424">This parameter is available only when <span class="parmname" id="EN-US_TOPIC_0000001955726494__en-us_topic_0000002319719730_parmname92321423544"><b>algorithm</b></span> is IVF_GRAPH or IVF_GRAPH_PQ.</p>
<p id="EN-US_TOPIC_0000001955726494__en-us_topic_0000002319719730_p065916559416">Value range: 0–100000</p>
<p id="EN-US_TOPIC_0000001955726494__en-us_topic_0000002319719730_p1250395714415">Default value: <strong id="EN-US_TOPIC_0000001955726494__en-us_topic_0000002319719730_b3918288929542">100</strong></p>
</td>
</tr>
</tbody>
</table>
</div>
</div>
<div class="section" id="EN-US_TOPIC_0000001955726494__en-us_topic_0000001268154489_section102341611718"><a name="EN-US_TOPIC_0000001955726494__en-us_topic_0000001268154489_section102341611718"></a><a name="en-us_topic_0000001268154489_section102341611718"></a><h4 class="sectiontitle">Hybrid Query</h4><p id="EN-US_TOPIC_0000001955726494__en-us_topic_0000001268154489_p364401311819">Hybrid query combines vector search with traditional OpenSearch queries, such as pre-filtering and Boolean queries.</p>
<div class="warning" id="EN-US_TOPIC_0000001955726494__note7315184563620"><span class="warningtitle"><img src="public_sys-resources/warning_3.0-en-us.png"> </span><div class="warningbody"><p id="EN-US_TOPIC_0000001955726494__p242820500615">Only OpenSearch 2.19.0 clusters support pre-filtering.</p>
</div></div>
<p id="EN-US_TOPIC_0000001955726494__p7992115142212">In the following example, the top 10 records whose <span class="parmname" id="EN-US_TOPIC_0000001955726494__parmname11430986972555"><b>my_label</b></span> value is <span class="parmvalue" id="EN-US_TOPIC_0000001955726494__parmvalue123669469372555"><b>red</b></span> are returned.</p>
<ul id="EN-US_TOPIC_0000001955726494__ul238620710614"><li id="EN-US_TOPIC_0000001955726494__li1583916257117"><strong id="EN-US_TOPIC_0000001955726494__b11651131413530">Pre-filtering query</strong><p id="EN-US_TOPIC_0000001955726494__p1325573345319">First, filters are applied to retrieve matching results. Then, vector search is performed on these results to retrieve the most relevant vectors based on similarity.</p>
<div class="p" id="EN-US_TOPIC_0000001955726494__p5328926115">The following is an example:<pre class="screen" id="EN-US_TOPIC_0000001955726494__screen2076214231683">POST <i><span class="varname" id="EN-US_TOPIC_0000001955726494__varname126175665219">my_index</span></i>/_search
{
"size": 10,
"query": {
"vector": {
"my_vector": {
"vector": [1, 2],
"topk": 10,
"filter": {
"term": { "my_label": "red" }
}
}
}
}
}</pre>
</div>
<div class="tablenoborder"><table cellpadding="4" cellspacing="0" summary="" id="EN-US_TOPIC_0000001955726494__table1727921614417" frame="border" border="1" rules="all"><caption><b>Table 2 </b>Parameters for pre-filtering query</caption><thead align="left"><tr id="EN-US_TOPIC_0000001955726494__row127941634111"><th align="left" class="cellrowborder" valign="top" width="23%" id="mcps1.3.3.5.1.4.2.5.1.1"><p id="EN-US_TOPIC_0000001955726494__p2279141644111">Parameter</p>
</th>
<th align="left" class="cellrowborder" valign="top" width="14.000000000000002%" id="mcps1.3.3.5.1.4.2.5.1.2"><p id="EN-US_TOPIC_0000001955726494__p727920162416">Mandatory</p>
</th>
<th align="left" class="cellrowborder" valign="top" width="15%" id="mcps1.3.3.5.1.4.2.5.1.3"><p id="EN-US_TOPIC_0000001955726494__p52791416144118">Type</p>
</th>
<th align="left" class="cellrowborder" valign="top" width="48%" id="mcps1.3.3.5.1.4.2.5.1.4"><p id="EN-US_TOPIC_0000001955726494__p627919166416">Description</p>
</th>
</tr>
</thead>
<tbody><tr id="EN-US_TOPIC_0000001955726494__row1727941624115"><td class="cellrowborder" valign="top" width="23%" headers="mcps1.3.3.5.1.4.2.5.1.1 "><p id="EN-US_TOPIC_0000001955726494__p116788306416">filter</p>
</td>
<td class="cellrowborder" valign="top" width="14.000000000000002%" headers="mcps1.3.3.5.1.4.2.5.1.2 "><p id="EN-US_TOPIC_0000001955726494__p06786301414">Yes</p>
</td>
<td class="cellrowborder" valign="top" width="15%" headers="mcps1.3.3.5.1.4.2.5.1.3 "><p id="EN-US_TOPIC_0000001955726494__p5678630124110">Map</p>
</td>
<td class="cellrowborder" valign="top" width="48%" headers="mcps1.3.3.5.1.4.2.5.1.4 "><p id="EN-US_TOPIC_0000001955726494__p19369920101116">Vector query filters. Standard OpenSearch query filters are supported, such as term and range.</p>
<p id="EN-US_TOPIC_0000001955726494__p3804104416172">If <span class="parmname" id="EN-US_TOPIC_0000001955726494__parmname861943373112542"><b>filter</b></span> is too restrictive, leading to a small intermediate result set, you can set the <span class="parmname" id="EN-US_TOPIC_0000001955726494__parmname1283754579112542"><b>index.vector.exact_search_threshold</b></span> parameter, so that when the intermediate result set is smaller than this threshold, pre-filtering query automatically switches over to brute-force query (FLAT algorithm), which ensures a high recall rate. For more information, see <a href="css_01_0121.html#EN-US_TOPIC_0000002319559898__en-us_topic_0000001309709789_section137344225249">Creating a Vector Index</a>.</p>
</td>
</tr>
<tr id="EN-US_TOPIC_0000001955726494__row127911160415"><td class="cellrowborder" valign="top" width="23%" headers="mcps1.3.3.5.1.4.2.5.1.1 "><p id="EN-US_TOPIC_0000001955726494__p967633017417">term</p>
</td>
<td class="cellrowborder" valign="top" width="14.000000000000002%" headers="mcps1.3.3.5.1.4.2.5.1.2 "><p id="EN-US_TOPIC_0000001955726494__p106769306410">No</p>
</td>
<td class="cellrowborder" valign="top" width="15%" headers="mcps1.3.3.5.1.4.2.5.1.3 "><p id="EN-US_TOPIC_0000001955726494__p667563010417">Map</p>
</td>
<td class="cellrowborder" valign="top" width="48%" headers="mcps1.3.3.5.1.4.2.5.1.4 "><p id="EN-US_TOPIC_0000001955726494__p8674193064111">Term query is a type of exact query. Documents that contain the exact term will be returned. For example, <span class="parmvalue" id="EN-US_TOPIC_0000001955726494__parmvalue1368864033112542"><b>{"term": {"my_label": "red"}}</b></span> means only to return documents whose <span class="parmname" id="EN-US_TOPIC_0000001955726494__parmname1913556673112542"><b>my_label</b></span> value is <span class="parmvalue" id="EN-US_TOPIC_0000001955726494__parmvalue1470353066112542"><b>red</b></span>.</p>
</td>
</tr>
</tbody>
</table>
</div>
</li><li id="EN-US_TOPIC_0000001955726494__li01282063320"><strong id="EN-US_TOPIC_0000001955726494__b1573843155318">Boolean query</strong><p id="EN-US_TOPIC_0000001955726494__p121371053132020">A Boolean query is in fact a post-filtering query method. Filtering and vector similarity-based search are performed separately. Then, the results of the two are combined using Boolean logic defined by clauses like must, should, and filter.</p>
<div class="p" id="EN-US_TOPIC_0000001955726494__p413711536208">The following is an example:<pre class="screen" id="EN-US_TOPIC_0000001955726494__screen1213875342014">POST <i><span class="varname" id="EN-US_TOPIC_0000001955726494__varname151381353172016">my_index</span></i>/_search
{
"size": 10,
"query": {
"bool": {
"must": {
"vector": {
"my_vector": {
"vector": [1, 2],
"topk": 10
}
}
},
"filter": {
"term": { "my_label": "red" }
}
}
}
}</pre>
</div>
<div class="tablenoborder"><table cellpadding="4" cellspacing="0" summary="" id="EN-US_TOPIC_0000001955726494__table8138145312208" frame="border" border="1" rules="all"><caption><b>Table 3 </b>Boolean query parameters</caption><thead align="left"><tr id="EN-US_TOPIC_0000001955726494__row813815311204"><th align="left" class="cellrowborder" valign="top" width="23%" id="mcps1.3.3.5.2.4.2.5.1.1"><p id="EN-US_TOPIC_0000001955726494__p13138125312011">Parameter</p>
</th>
<th align="left" class="cellrowborder" valign="top" width="14.000000000000002%" id="mcps1.3.3.5.2.4.2.5.1.2"><p id="EN-US_TOPIC_0000001955726494__p20138115342013">Mandatory</p>
</th>
<th align="left" class="cellrowborder" valign="top" width="15%" id="mcps1.3.3.5.2.4.2.5.1.3"><p id="EN-US_TOPIC_0000001955726494__p1213865314209">Type</p>
</th>
<th align="left" class="cellrowborder" valign="top" width="48%" id="mcps1.3.3.5.2.4.2.5.1.4"><p id="EN-US_TOPIC_0000001955726494__p0138105382019">Description</p>
</th>
</tr>
</thead>
<tbody><tr id="EN-US_TOPIC_0000001955726494__row51393531207"><td class="cellrowborder" valign="top" width="23%" headers="mcps1.3.3.5.2.4.2.5.1.1 "><p id="EN-US_TOPIC_0000001955726494__p813919536205">bool</p>
</td>
<td class="cellrowborder" valign="top" width="14.000000000000002%" headers="mcps1.3.3.5.2.4.2.5.1.2 "><p id="EN-US_TOPIC_0000001955726494__p181391553172011">Yes</p>
</td>
<td class="cellrowborder" valign="top" width="15%" headers="mcps1.3.3.5.2.4.2.5.1.3 "><p id="EN-US_TOPIC_0000001955726494__p3139135362019">Map</p>
</td>
<td class="cellrowborder" valign="top" width="48%" headers="mcps1.3.3.5.2.4.2.5.1.4 "><p id="EN-US_TOPIC_0000001955726494__p413995392019">A compound query clause that combines subqueries using configured Boolean logic.</p>
<div class="p" id="EN-US_TOPIC_0000001955726494__p6139195372010">Parameter description:<ul id="EN-US_TOPIC_0000001955726494__ul813985362010"><li id="EN-US_TOPIC_0000001955726494__li1413965372011"><strong id="EN-US_TOPIC_0000001955726494__b202353775172555">must</strong>: Clauses that must match for documents to be included in the results.</li><li id="EN-US_TOPIC_0000001955726494__li5139175310203"><strong id="EN-US_TOPIC_0000001955726494__b183099655672555">filter</strong>: It is similar to <strong id="EN-US_TOPIC_0000001955726494__b14110526572555">must</strong>, but do not contribute to the relevance score.</li><li id="EN-US_TOPIC_0000001955726494__li1813935317202"><strong id="EN-US_TOPIC_0000001955726494__b151803119572555">should</strong>: Clauses that should match, but are not required. They are like nice-to-haves.</li><li id="EN-US_TOPIC_0000001955726494__li1913925311208"><strong id="EN-US_TOPIC_0000001955726494__b188129203072555">must_not</strong>: Clauses that must not match for documents to be included in the results.</li></ul>
</div>
</td>
</tr>
<tr id="EN-US_TOPIC_0000001955726494__row1413975310201"><td class="cellrowborder" valign="top" width="23%" headers="mcps1.3.3.5.2.4.2.5.1.1 "><p id="EN-US_TOPIC_0000001955726494__p15139175362011">bool.must</p>
</td>
<td class="cellrowborder" valign="top" width="14.000000000000002%" headers="mcps1.3.3.5.2.4.2.5.1.2 "><p id="EN-US_TOPIC_0000001955726494__p121391953182013">Yes</p>
</td>
<td class="cellrowborder" valign="top" width="15%" headers="mcps1.3.3.5.2.4.2.5.1.3 "><p id="EN-US_TOPIC_0000001955726494__p191391753152020">Map</p>
</td>
<td class="cellrowborder" valign="top" width="48%" headers="mcps1.3.3.5.2.4.2.5.1.4 "><div class="p" id="EN-US_TOPIC_0000001955726494__p15139155316200">Clauses that must match for documents to be included in the results. Parameter description:<ul id="EN-US_TOPIC_0000001955726494__ul15139135372019"><li id="EN-US_TOPIC_0000001955726494__li151395537204"><strong id="EN-US_TOPIC_0000001955726494__b139861179972555">vector</strong>: query vector</li><li id="EN-US_TOPIC_0000001955726494__li114065382016"><strong id="EN-US_TOPIC_0000001955726494__b3244359572555">my_vector</strong>: vector field</li><li id="EN-US_TOPIC_0000001955726494__li114085316201"><strong id="EN-US_TOPIC_0000001955726494__b85620321872555">topk</strong>: number of results to return</li></ul>
</div>
</td>
</tr>
<tr id="EN-US_TOPIC_0000001955726494__row0140105317205"><td class="cellrowborder" valign="top" width="23%" headers="mcps1.3.3.5.2.4.2.5.1.1 "><p id="EN-US_TOPIC_0000001955726494__p1214020537201">bool.filter</p>
</td>
<td class="cellrowborder" valign="top" width="14.000000000000002%" headers="mcps1.3.3.5.2.4.2.5.1.2 "><p id="EN-US_TOPIC_0000001955726494__p914065362016">Yes</p>
</td>
<td class="cellrowborder" valign="top" width="15%" headers="mcps1.3.3.5.2.4.2.5.1.3 "><p id="EN-US_TOPIC_0000001955726494__p814055315207">Map</p>
</td>
<td class="cellrowborder" valign="top" width="48%" headers="mcps1.3.3.5.2.4.2.5.1.4 "><p id="EN-US_TOPIC_0000001955726494__p6140553142011">Clauses that must match, but do not contribute to the relevance score. Standard OpenSearch query filters are supported, such as term and range.</p>
</td>
</tr>
</tbody>
</table>
</div>
</li></ul>
</div>
<div class="section" id="EN-US_TOPIC_0000001955726494__en-us_topic_0000001268154489_section4946246546"><a name="EN-US_TOPIC_0000001955726494__en-us_topic_0000001268154489_section4946246546"></a><a name="en-us_topic_0000001268154489_section4946246546"></a><h4 class="sectiontitle">Script Score Query</h4><p id="EN-US_TOPIC_0000001955726494__en-us_topic_0000002319719730_en-us_topic_0000001268154489_p191338162050">Script_score query enables custom similarity calculations for vector searches by executing a user-defined script. It works as follows:</p>
<p id="EN-US_TOPIC_0000001955726494__en-us_topic_0000002319719730_en-us_topic_0000001268154489_p1512359042">Pre-filtering works with any query. <strong id="EN-US_TOPIC_0000001955726494__en-us_topic_0000002319719730_b13024816099542">script_score</strong> then calculates vector similarity on the pre-filtered results, and ranks the results. This query method does not use vector indexes. Its performance depends on the size of the intermediate result set after the pre-filtering. If the pre-filtering condition is set to <strong id="EN-US_TOPIC_0000001955726494__en-us_topic_0000002319719730_b560736449542">match_all</strong>, a brute-force search is performed on all data.</p>
<p id="EN-US_TOPIC_0000001955726494__en-us_topic_0000002319719730_p27367195718">The following is an example:</p>
<pre class="screen" id="EN-US_TOPIC_0000001955726494__en-us_topic_0000002319719730_en-us_topic_0000001268154489_screen3512185915411">POST <i><span class="varname" id="EN-US_TOPIC_0000001955726494__en-us_topic_0000002319719730_varname573812415551">my_index</span></i>/_search
{
"size":2,
"query": {
"script_score": {
"query": {
"match_all": {}
},
"script": {
"source": "vector_score",
"lang": "vector",
"params": {
"field": "my_vector",
"vector": [1.0, 2.0],
"metric": "euclidean"
}
}
}
}
}</pre>
<div class="tablenoborder"><table cellpadding="4" cellspacing="0" summary="" id="EN-US_TOPIC_0000001955726494__en-us_topic_0000002319719730_en-us_topic_0000001268154489_table95121659546" frame="border" border="1" rules="all"><caption><b>Table 4 </b>script_score query parameters</caption><thead align="left"><tr id="EN-US_TOPIC_0000001955726494__en-us_topic_0000002319719730_en-us_topic_0000001268154489_row1451214594419"><th align="left" class="cellrowborder" valign="top" width="23%" id="mcps1.3.4.6.2.5.1.1"><p id="EN-US_TOPIC_0000001955726494__en-us_topic_0000002319719730_en-us_topic_0000001268154489_p551216591840">Parameter</p>
</th>
<th align="left" class="cellrowborder" valign="top" width="14.000000000000002%" id="mcps1.3.4.6.2.5.1.2"><p id="EN-US_TOPIC_0000001955726494__en-us_topic_0000002319719730_p86491300215">Mandatory</p>
</th>
<th align="left" class="cellrowborder" valign="top" width="15%" id="mcps1.3.4.6.2.5.1.3"><p id="EN-US_TOPIC_0000001955726494__en-us_topic_0000002319719730_p435082812112">Type</p>
</th>
<th align="left" class="cellrowborder" valign="top" width="48%" id="mcps1.3.4.6.2.5.1.4"><p id="EN-US_TOPIC_0000001955726494__en-us_topic_0000002319719730_en-us_topic_0000001268154489_p105128591413">Description</p>
</th>
</tr>
</thead>
<tbody><tr id="EN-US_TOPIC_0000001955726494__en-us_topic_0000002319719730_row13141161515111"><td class="cellrowborder" valign="top" width="23%" headers="mcps1.3.4.6.2.5.1.1 "><p id="EN-US_TOPIC_0000001955726494__en-us_topic_0000002319719730_p101415158511">script_score</p>
</td>
<td class="cellrowborder" valign="top" width="14.000000000000002%" headers="mcps1.3.4.6.2.5.1.2 "><p id="EN-US_TOPIC_0000001955726494__en-us_topic_0000002319719730_p41414159512">Yes</p>
</td>
<td class="cellrowborder" valign="top" width="15%" headers="mcps1.3.4.6.2.5.1.3 "><p id="EN-US_TOPIC_0000001955726494__en-us_topic_0000002319719730_p121411415185118">Map</p>
</td>
<td class="cellrowborder" valign="top" width="48%" headers="mcps1.3.4.6.2.5.1.4 "><p id="EN-US_TOPIC_0000001955726494__en-us_topic_0000002319719730_p139226675218">A root parameter for the script_score query.</p>
<div class="p" id="EN-US_TOPIC_0000001955726494__en-us_topic_0000002319719730_p155441642526">Parameter description:<ul id="EN-US_TOPIC_0000001955726494__en-us_topic_0000002319719730_ul17141133935314"><li id="EN-US_TOPIC_0000001955726494__en-us_topic_0000002319719730_li81411039205320"><strong id="EN-US_TOPIC_0000001955726494__en-us_topic_0000002319719730_b1587905217328">query</strong>: pre-filtering criteria. When it is set to <span class="parmname" id="EN-US_TOPIC_0000001955726494__en-us_topic_0000002319719730_parmname94292115576"><b>match_all</b></span>, a brute-force search is performed on all data.</li><li id="EN-US_TOPIC_0000001955726494__en-us_topic_0000002319719730_li151411639165319"><strong id="EN-US_TOPIC_0000001955726494__en-us_topic_0000002319719730_b323565123315">script</strong>: custom script that calculates similarity scores.</li></ul>
</div>
</td>
</tr>
<tr id="EN-US_TOPIC_0000001955726494__en-us_topic_0000002319719730_en-us_topic_0000001268154489_row135121591243"><td class="cellrowborder" valign="top" width="23%" headers="mcps1.3.4.6.2.5.1.1 "><p id="EN-US_TOPIC_0000001955726494__en-us_topic_0000002319719730_en-us_topic_0000001268154489_p651218591747">source</p>
</td>
<td class="cellrowborder" valign="top" width="14.000000000000002%" headers="mcps1.3.4.6.2.5.1.2 "><p id="EN-US_TOPIC_0000001955726494__en-us_topic_0000002319719730_p106491830112116">Yes</p>
</td>
<td class="cellrowborder" valign="top" width="15%" headers="mcps1.3.4.6.2.5.1.3 "><p id="EN-US_TOPIC_0000001955726494__en-us_topic_0000002319719730_p11350182872118">String</p>
</td>
<td class="cellrowborder" valign="top" width="48%" headers="mcps1.3.4.6.2.5.1.4 "><p id="EN-US_TOPIC_0000001955726494__en-us_topic_0000002319719730_en-us_topic_0000001268154489_p1351225910417">Script name. The value is fixed to <span class="parmvalue" id="EN-US_TOPIC_0000001955726494__en-us_topic_0000002319719730_parmvalue8308162616239"><b>vector_score</b></span>, indicating that a built-in script is used for calculating similarity.</p>
</td>
</tr>
<tr id="EN-US_TOPIC_0000001955726494__en-us_topic_0000002319719730_en-us_topic_0000001268154489_row175123599411"><td class="cellrowborder" valign="top" width="23%" headers="mcps1.3.4.6.2.5.1.1 "><p id="EN-US_TOPIC_0000001955726494__en-us_topic_0000002319719730_en-us_topic_0000001268154489_p75121759143">lang</p>
</td>
<td class="cellrowborder" valign="top" width="14.000000000000002%" headers="mcps1.3.4.6.2.5.1.2 "><p id="EN-US_TOPIC_0000001955726494__en-us_topic_0000002319719730_p76493308212">Yes</p>
</td>
<td class="cellrowborder" valign="top" width="15%" headers="mcps1.3.4.6.2.5.1.3 "><p id="EN-US_TOPIC_0000001955726494__en-us_topic_0000002319719730_p1035018283215">String</p>
</td>
<td class="cellrowborder" valign="top" width="48%" headers="mcps1.3.4.6.2.5.1.4 "><p id="EN-US_TOPIC_0000001955726494__en-us_topic_0000002319719730_en-us_topic_0000001268154489_p1651235910417">Script language type. The value is fixed to <span class="parmvalue" id="EN-US_TOPIC_0000001955726494__en-us_topic_0000002319719730_parmvalue15555036162313"><b>vector</b></span>.</p>
</td>
</tr>
<tr id="EN-US_TOPIC_0000001955726494__en-us_topic_0000002319719730_en-us_topic_0000001268154489_row1651219599413"><td class="cellrowborder" valign="top" width="23%" headers="mcps1.3.4.6.2.5.1.1 "><p id="EN-US_TOPIC_0000001955726494__en-us_topic_0000002319719730_en-us_topic_0000001268154489_p1451211592047">field</p>
</td>
<td class="cellrowborder" valign="top" width="14.000000000000002%" headers="mcps1.3.4.6.2.5.1.2 "><p id="EN-US_TOPIC_0000001955726494__en-us_topic_0000002319719730_p76492030162114">Yes</p>
</td>
<td class="cellrowborder" valign="top" width="15%" headers="mcps1.3.4.6.2.5.1.3 "><p id="EN-US_TOPIC_0000001955726494__en-us_topic_0000002319719730_p1350428202120">String</p>
</td>
<td class="cellrowborder" valign="top" width="48%" headers="mcps1.3.4.6.2.5.1.4 "><p id="EN-US_TOPIC_0000001955726494__en-us_topic_0000002319719730_en-us_topic_0000001268154489_p1351218596418">Queried vector field, for example, <strong id="EN-US_TOPIC_0000001955726494__en-us_topic_0000002319719730_b1216171414377">my_vector</strong>.</p>
</td>
</tr>
<tr id="EN-US_TOPIC_0000001955726494__en-us_topic_0000002319719730_en-us_topic_0000001268154489_row1551285917417"><td class="cellrowborder" valign="top" width="23%" headers="mcps1.3.4.6.2.5.1.1 "><p id="EN-US_TOPIC_0000001955726494__en-us_topic_0000002319719730_en-us_topic_0000001268154489_p165129591342">vector</p>
</td>
<td class="cellrowborder" valign="top" width="14.000000000000002%" headers="mcps1.3.4.6.2.5.1.2 "><p id="EN-US_TOPIC_0000001955726494__en-us_topic_0000002319719730_p17936101012612">Yes</p>
</td>
<td class="cellrowborder" valign="top" width="15%" headers="mcps1.3.4.6.2.5.1.3 "><p id="EN-US_TOPIC_0000001955726494__en-us_topic_0000002319719730_p19936210463">Array/String</p>
</td>
<td class="cellrowborder" valign="top" width="48%" headers="mcps1.3.4.6.2.5.1.4 "><p id="EN-US_TOPIC_0000001955726494__en-us_topic_0000002319719730_p8936610961">Query vector value. It is used to calculate the similarity between indexed vectors and the query vector.</p>
<p id="EN-US_TOPIC_0000001955726494__en-us_topic_0000002319719730_p093613101067">The value can be an array (for example, [1, 1]) or Base64-encoded value (for example, <span class="parmvalue" id="EN-US_TOPIC_0000001955726494__en-us_topic_0000002319719730_parmvalue18217336145"><b>AAABAAACAAAD</b></span>).</p>
</td>
</tr>
<tr id="EN-US_TOPIC_0000001955726494__en-us_topic_0000002319719730_en-us_topic_0000001268154489_row751218591947"><td class="cellrowborder" valign="top" width="23%" headers="mcps1.3.4.6.2.5.1.1 "><p id="EN-US_TOPIC_0000001955726494__en-us_topic_0000002319719730_en-us_topic_0000001268154489_p18512659244">metric</p>
</td>
<td class="cellrowborder" valign="top" width="14.000000000000002%" headers="mcps1.3.4.6.2.5.1.2 "><p id="EN-US_TOPIC_0000001955726494__en-us_topic_0000002319719730_p917382645416">Yes</p>
</td>
<td class="cellrowborder" valign="top" width="15%" headers="mcps1.3.4.6.2.5.1.3 "><p id="EN-US_TOPIC_0000001955726494__en-us_topic_0000002319719730_p19173526175411">String</p>
</td>
<td class="cellrowborder" valign="top" width="48%" headers="mcps1.3.4.6.2.5.1.4 "><p id="EN-US_TOPIC_0000001955726494__en-us_topic_0000002319719730_p1417320262542">Vector distance metric, which measures the similarity or distance between vectors.</p>
<p id="EN-US_TOPIC_0000001955726494__en-us_topic_0000002319719730_p517362613541">Value range:</p>
<ul id="EN-US_TOPIC_0000001955726494__en-us_topic_0000002319719730_ul717332685413"><li id="EN-US_TOPIC_0000001955726494__en-us_topic_0000002319719730_li4173192616544"><strong id="EN-US_TOPIC_0000001955726494__en-us_topic_0000002319719730_b19047695919542">euclidean</strong> (default): Euclidean distance</li><li id="EN-US_TOPIC_0000001955726494__en-us_topic_0000002319719730_li1717392611548"><strong id="EN-US_TOPIC_0000001955726494__en-us_topic_0000002319719730_b10735702739542">inner_product</strong>: inner product distance</li><li id="EN-US_TOPIC_0000001955726494__en-us_topic_0000002319719730_li117317263542"><strong id="EN-US_TOPIC_0000001955726494__en-us_topic_0000002319719730_b19701872639542">cosine</strong>: cosine distance</li><li id="EN-US_TOPIC_0000001955726494__en-us_topic_0000002319719730_li8173102615546"><strong id="EN-US_TOPIC_0000001955726494__en-us_topic_0000002319719730_b4243944119542">hamming</strong>: Hamming distance, which can be used only when <strong id="EN-US_TOPIC_0000001955726494__en-us_topic_0000002319719730_b13693234849542">dim_type</strong> is set to <strong id="EN-US_TOPIC_0000001955726494__en-us_topic_0000002319719730_b3714137629542">binary</strong>.</li></ul>
</td>
</tr>
</tbody>
</table>
</div>
</div>
<div class="section" id="EN-US_TOPIC_0000001955726494__en-us_topic_0000001268154489_section107953451661"><a name="EN-US_TOPIC_0000001955726494__en-us_topic_0000001268154489_section107953451661"></a><a name="en-us_topic_0000001268154489_section107953451661"></a><h4 class="sectiontitle">Rescore Query</h4><p id="EN-US_TOPIC_0000001955726494__en-us_topic_0000002319719730_p176783181260">Rescore query rescores and reranks the top results returned by an initial query to improve recall.</p>
<p id="EN-US_TOPIC_0000001955726494__en-us_topic_0000002319719730_en-us_topic_0000001268154489_p147213301172">When the GRAPH_PQ or IVF_GRAPH_PQ indexing algorithm is used, query results are ranked based on the asymmetric distance calculated by PQ. Rescore query then rescores and reranks the initial search results to improve recall.</p>
<p id="EN-US_TOPIC_0000001955726494__en-us_topic_0000002319719730_en-us_topic_0000001268154489_p1567411217323">The following is an example of rescore query on a PQ index named <strong id="EN-US_TOPIC_0000001955726494__en-us_topic_0000002319719730_b4350133823010">my_index</strong>:</p>
<pre class="screen" id="EN-US_TOPIC_0000001955726494__en-us_topic_0000002319719730_en-us_topic_0000001268154489_screen1639814155510">GET <i><span class="varname" id="EN-US_TOPIC_0000001955726494__en-us_topic_0000002319719730_varname19581841103214">my_index</span></i>/_search
{
"size": 10,
"query": {
"vector": {
"my_vector": {
"vector": [1.0, 2.0],
"topk": 100
}
}
},
"rescore": {
"window_size": 100,
"vector_rescore": {
"field": "my_vector",
"vector": [1.0, 2.0],
"metric": "euclidean"
}
}
}</pre>
<div class="tablenoborder"><table cellpadding="4" cellspacing="0" summary="" id="EN-US_TOPIC_0000001955726494__en-us_topic_0000002319719730_en-us_topic_0000001268154489_table165271512103218" frame="border" border="1" rules="all"><caption><b>Table 5 </b>Rescore query parameters</caption><thead align="left"><tr id="EN-US_TOPIC_0000001955726494__en-us_topic_0000002319719730_en-us_topic_0000001268154489_row96748123320"><th align="left" class="cellrowborder" valign="top" width="23%" id="mcps1.3.5.6.2.5.1.1"><p id="EN-US_TOPIC_0000001955726494__en-us_topic_0000002319719730_en-us_topic_0000001268154489_p467481273214">Parameter</p>
</th>
<th align="left" class="cellrowborder" valign="top" width="14.000000000000002%" id="mcps1.3.5.6.2.5.1.2"><p id="EN-US_TOPIC_0000001955726494__en-us_topic_0000002319719730_p3133837113020">Mandatory</p>
</th>
<th align="left" class="cellrowborder" valign="top" width="15%" id="mcps1.3.5.6.2.5.1.3"><p id="EN-US_TOPIC_0000001955726494__en-us_topic_0000002319719730_p19858163823019">Type</p>
</th>
<th align="left" class="cellrowborder" valign="top" width="48%" id="mcps1.3.5.6.2.5.1.4"><p id="EN-US_TOPIC_0000001955726494__en-us_topic_0000002319719730_en-us_topic_0000001268154489_p7674111213212">Description</p>
</th>
</tr>
</thead>
<tbody><tr id="EN-US_TOPIC_0000001955726494__en-us_topic_0000002319719730_row168165751714"><td class="cellrowborder" valign="top" width="23%" headers="mcps1.3.5.6.2.5.1.1 "><p id="EN-US_TOPIC_0000001955726494__en-us_topic_0000002319719730_p681618741715">rescore</p>
</td>
<td class="cellrowborder" valign="top" width="14.000000000000002%" headers="mcps1.3.5.6.2.5.1.2 "><p id="EN-US_TOPIC_0000001955726494__en-us_topic_0000002319719730_p1381610711711">Yes</p>
</td>
<td class="cellrowborder" valign="top" width="15%" headers="mcps1.3.5.6.2.5.1.3 "><p id="EN-US_TOPIC_0000001955726494__en-us_topic_0000002319719730_p188166711177">Map</p>
</td>
<td class="cellrowborder" valign="top" width="48%" headers="mcps1.3.5.6.2.5.1.4 "><p id="EN-US_TOPIC_0000001955726494__en-us_topic_0000002319719730_p20222933111715">Defines rescoring parameters.</p>
<div class="p" id="EN-US_TOPIC_0000001955726494__en-us_topic_0000002319719730_p17550845141716">Key parameters:<ul id="EN-US_TOPIC_0000001955726494__en-us_topic_0000002319719730_ul4101105851713"><li id="EN-US_TOPIC_0000001955726494__en-us_topic_0000002319719730_li111011458141719">window_size: rescoring/reranking window size.</li><li id="EN-US_TOPIC_0000001955726494__en-us_topic_0000002319719730_li20101358161711">vector_rescore: other vector rescoring settings.</li></ul>
</div>
</td>
</tr>
<tr id="EN-US_TOPIC_0000001955726494__en-us_topic_0000002319719730_en-us_topic_0000001268154489_row4674161215326"><td class="cellrowborder" valign="top" width="23%" headers="mcps1.3.5.6.2.5.1.1 "><p id="EN-US_TOPIC_0000001955726494__en-us_topic_0000002319719730_en-us_topic_0000001268154489_p136741112123210">window_size</p>
</td>
<td class="cellrowborder" valign="top" width="14.000000000000002%" headers="mcps1.3.5.6.2.5.1.2 "><p id="EN-US_TOPIC_0000001955726494__en-us_topic_0000002319719730_p16133123716300">Yes</p>
</td>
<td class="cellrowborder" valign="top" width="15%" headers="mcps1.3.5.6.2.5.1.3 "><p id="EN-US_TOPIC_0000001955726494__en-us_topic_0000002319719730_p285810386306">Integer</p>
</td>
<td class="cellrowborder" valign="top" width="48%" headers="mcps1.3.5.6.2.5.1.4 "><p id="EN-US_TOPIC_0000001955726494__en-us_topic_0000002319719730_en-us_topic_0000001268154489_p116742129325">Rescoring/reranking window size. The vector search returns the top k results, but only the first <em id="EN-US_TOPIC_0000001955726494__en-us_topic_0000002319719730_i189817341712">window_size</em> results are rescored and reranked. A larger value indicates a larger reranking scope and hence a higher recall rate, but it also leads to higher computational overhead.</p>
<p id="EN-US_TOPIC_0000001955726494__en-us_topic_0000002319719730_p52661033192">Default value: <strong id="EN-US_TOPIC_0000001955726494__en-us_topic_0000002319719730_b7097594119542">100</strong></p>
</td>
</tr>
<tr id="EN-US_TOPIC_0000001955726494__en-us_topic_0000002319719730_en-us_topic_0000001268154489_row19674612133213"><td class="cellrowborder" valign="top" width="23%" headers="mcps1.3.5.6.2.5.1.1 "><p id="EN-US_TOPIC_0000001955726494__en-us_topic_0000002319719730_en-us_topic_0000001268154489_p16674171213212">field</p>
</td>
<td class="cellrowborder" valign="top" width="14.000000000000002%" headers="mcps1.3.5.6.2.5.1.2 "><p id="EN-US_TOPIC_0000001955726494__en-us_topic_0000002319719730_p1710022896">Yes</p>
</td>
<td class="cellrowborder" valign="top" width="15%" headers="mcps1.3.5.6.2.5.1.3 "><p id="EN-US_TOPIC_0000001955726494__en-us_topic_0000002319719730_p20100621913">String</p>
</td>
<td class="cellrowborder" valign="top" width="48%" headers="mcps1.3.5.6.2.5.1.4 "><p id="EN-US_TOPIC_0000001955726494__en-us_topic_0000002319719730_p6100142799">Queried vector field, for example, <strong id="EN-US_TOPIC_0000001955726494__en-us_topic_0000002319719730_b1631516241913">my_vector</strong>.</p>
</td>
</tr>
<tr id="EN-US_TOPIC_0000001955726494__en-us_topic_0000002319719730_en-us_topic_0000001268154489_row6674151212329"><td class="cellrowborder" valign="top" width="23%" headers="mcps1.3.5.6.2.5.1.1 "><p id="EN-US_TOPIC_0000001955726494__en-us_topic_0000002319719730_en-us_topic_0000001268154489_p14674171273210">vector</p>
</td>
<td class="cellrowborder" valign="top" width="14.000000000000002%" headers="mcps1.3.5.6.2.5.1.2 "><p id="EN-US_TOPIC_0000001955726494__en-us_topic_0000002319719730_p270335312817">Yes</p>
</td>
<td class="cellrowborder" valign="top" width="15%" headers="mcps1.3.5.6.2.5.1.3 "><p id="EN-US_TOPIC_0000001955726494__en-us_topic_0000002319719730_p17036534810">Array/String</p>
</td>
<td class="cellrowborder" valign="top" width="48%" headers="mcps1.3.5.6.2.5.1.4 "><p id="EN-US_TOPIC_0000001955726494__en-us_topic_0000002319719730_p1270414536812">Query vector value. It is used to calculate the similarity between indexed vectors and the query vector.</p>
<p id="EN-US_TOPIC_0000001955726494__en-us_topic_0000002319719730_p19704453986">The value can be an array (for example, [1, 1]) or Base64-encoded value (for example, <span class="parmvalue" id="EN-US_TOPIC_0000001955726494__en-us_topic_0000002319719730_parmvalue02201531147"><b>AAABAAACAAAD</b></span>).</p>
</td>
</tr>
<tr id="EN-US_TOPIC_0000001955726494__en-us_topic_0000002319719730_en-us_topic_0000001268154489_row7674121216327"><td class="cellrowborder" valign="top" width="23%" headers="mcps1.3.5.6.2.5.1.1 "><p id="EN-US_TOPIC_0000001955726494__en-us_topic_0000002319719730_en-us_topic_0000001268154489_p167419129328">metric</p>
</td>
<td class="cellrowborder" valign="top" width="14.000000000000002%" headers="mcps1.3.5.6.2.5.1.2 "><p id="EN-US_TOPIC_0000001955726494__en-us_topic_0000002319719730_p16988342781">Yes</p>
</td>
<td class="cellrowborder" valign="top" width="15%" headers="mcps1.3.5.6.2.5.1.3 "><p id="EN-US_TOPIC_0000001955726494__en-us_topic_0000002319719730_p198854213819">String</p>
</td>
<td class="cellrowborder" valign="top" width="48%" headers="mcps1.3.5.6.2.5.1.4 "><p id="EN-US_TOPIC_0000001955726494__en-us_topic_0000002319719730_p39889421985">Vector distance metric, which measures the similarity or distance between vectors.</p>
<p id="EN-US_TOPIC_0000001955726494__en-us_topic_0000002319719730_p1298864217818">Value range:</p>
<ul id="EN-US_TOPIC_0000001955726494__en-us_topic_0000002319719730_ul189881742384"><li id="EN-US_TOPIC_0000001955726494__en-us_topic_0000002319719730_li13988104218818"><strong id="EN-US_TOPIC_0000001955726494__en-us_topic_0000002319719730_b2261639249542">euclidean</strong> (default): Euclidean distance</li><li id="EN-US_TOPIC_0000001955726494__en-us_topic_0000002319719730_li159881042988"><strong id="EN-US_TOPIC_0000001955726494__en-us_topic_0000002319719730_b7131293759542">inner_product</strong>: inner product distance</li><li id="EN-US_TOPIC_0000001955726494__en-us_topic_0000002319719730_li39881442989"><strong id="EN-US_TOPIC_0000001955726494__en-us_topic_0000002319719730_b1061157679542">cosine</strong>: cosine distance</li><li id="EN-US_TOPIC_0000001955726494__en-us_topic_0000002319719730_li8988124220820"><strong id="EN-US_TOPIC_0000001955726494__en-us_topic_0000002319719730_b0103526801">hamming</strong>: Hamming distance, which can be used only when <strong id="EN-US_TOPIC_0000001955726494__en-us_topic_0000002319719730_b8103326302">dim_type</strong> is set to <strong id="EN-US_TOPIC_0000001955726494__en-us_topic_0000002319719730_b5103192611013">binary</strong>.</li></ul>
</td>
</tr>
</tbody>
</table>
</div>
</div>
<div class="section" id="EN-US_TOPIC_0000001955726494__section102071126171611"><a name="EN-US_TOPIC_0000001955726494__section102071126171611"></a><a name="section102071126171611"></a><h4 class="sectiontitle">Painless Syntax Extension</h4><p id="EN-US_TOPIC_0000001955726494__en-us_topic_0000002319719730_p177861335101615">Painless syntax extension allows the use of vector distance or similarity calculation functions in custom scripts. CSS extension supports several vector distance/similarity calculation functions, which users can use readily in custom Painless scripts to build flexible rescoring formulas.</p>
<p id="EN-US_TOPIC_0000001955726494__en-us_topic_0000002319719730_p18965620112713">The following is an example:</p>
<pre class="screen" id="EN-US_TOPIC_0000001955726494__en-us_topic_0000002319719730_screen11315195710305">POST <i><span class="varname" id="EN-US_TOPIC_0000001955726494__en-us_topic_0000002319719730_varname14606636193214">my_index</span></i>/_search
{
"size": 10,
"query": {
"script_score": {
"query": {
"match_all": {}
},
"script": {
"source": "1 / (1 + euclidean(params.vector, doc[params.field]))",
"params": {
"field": "my_vector",
"vector": [1, 2]
}
}
}
}
}</pre>
<div class="tablenoborder"><table cellpadding="4" cellspacing="0" summary="" id="EN-US_TOPIC_0000001955726494__en-us_topic_0000002319719730_table17610185023516" frame="border" border="1" rules="all"><caption><b>Table 6 </b>Supported vector distance/similarity calculation functions</caption><thead align="left"><tr id="EN-US_TOPIC_0000001955726494__en-us_topic_0000002319719730_row461012507354"><th align="left" class="cellrowborder" valign="top" width="34.32%" id="mcps1.3.6.5.2.3.1.1"><p id="EN-US_TOPIC_0000001955726494__en-us_topic_0000002319719730_p196101850203512">Function Signature</p>
</th>
<th align="left" class="cellrowborder" valign="top" width="65.68%" id="mcps1.3.6.5.2.3.1.2"><p id="EN-US_TOPIC_0000001955726494__en-us_topic_0000002319719730_p1061095020358">Description</p>
</th>
</tr>
</thead>
<tbody><tr id="EN-US_TOPIC_0000001955726494__en-us_topic_0000002319719730_row1261045014357"><td class="cellrowborder" valign="top" width="34.32%" headers="mcps1.3.6.5.2.3.1.1 "><p id="EN-US_TOPIC_0000001955726494__en-us_topic_0000002319719730_p13610450103520">euclidean(Float[], DocValues)</p>
</td>
<td class="cellrowborder" valign="top" width="65.68%" headers="mcps1.3.6.5.2.3.1.2 "><p id="EN-US_TOPIC_0000001955726494__en-us_topic_0000002319719730_p2610125033511">Euclidean distance</p>
</td>
</tr>
<tr id="EN-US_TOPIC_0000001955726494__en-us_topic_0000002319719730_row961085053513"><td class="cellrowborder" valign="top" width="34.32%" headers="mcps1.3.6.5.2.3.1.1 "><p id="EN-US_TOPIC_0000001955726494__en-us_topic_0000002319719730_p1610205014351">cosine(Float[], DocValues)</p>
</td>
<td class="cellrowborder" valign="top" width="65.68%" headers="mcps1.3.6.5.2.3.1.2 "><p id="EN-US_TOPIC_0000001955726494__en-us_topic_0000002319719730_p6610750143515">Cosine similarity</p>
</td>
</tr>
<tr id="EN-US_TOPIC_0000001955726494__en-us_topic_0000002319719730_row1961015501351"><td class="cellrowborder" valign="top" width="34.32%" headers="mcps1.3.6.5.2.3.1.1 "><p id="EN-US_TOPIC_0000001955726494__en-us_topic_0000002319719730_p46101950163517">innerproduct(Float[], DocValues)</p>
</td>
<td class="cellrowborder" valign="top" width="65.68%" headers="mcps1.3.6.5.2.3.1.2 "><p id="EN-US_TOPIC_0000001955726494__en-us_topic_0000002319719730_p461015012354">Inner product</p>
</td>
</tr>
<tr id="EN-US_TOPIC_0000001955726494__en-us_topic_0000002319719730_row98852044174013"><td class="cellrowborder" valign="top" width="34.32%" headers="mcps1.3.6.5.2.3.1.1 "><p id="EN-US_TOPIC_0000001955726494__en-us_topic_0000002319719730_p13885174412404">hamming(String, DocValues)</p>
</td>
<td class="cellrowborder" valign="top" width="65.68%" headers="mcps1.3.6.5.2.3.1.2 "><p id="EN-US_TOPIC_0000001955726494__en-us_topic_0000002319719730_p18852044114019">Hamming distance Only vectors whose <strong id="EN-US_TOPIC_0000001955726494__b898334319112542">dim_type</strong> is <strong id="EN-US_TOPIC_0000001955726494__b1865793899112542">binary</strong> are supported. The input query vector must be a Base64-encoded character string.</p>
<p id="EN-US_TOPIC_0000001955726494__p422195073115">Only OpenSearch 1.3.6 clusters support this function.</p>
</td>
</tr>
<tr id="EN-US_TOPIC_0000001955726494__row19222131811313"><td class="cellrowborder" valign="top" width="34.32%" headers="mcps1.3.6.5.2.3.1.1 "><p id="EN-US_TOPIC_0000001955726494__p1487113113111">hammings(String, DocValues)</p>
</td>
<td class="cellrowborder" valign="top" width="65.68%" headers="mcps1.3.6.5.2.3.1.2 "><p id="EN-US_TOPIC_0000001955726494__p6871183133117">Hamming distance Only vectors whose <strong id="EN-US_TOPIC_0000001955726494__b601122893112542">dim_type</strong> is <strong id="EN-US_TOPIC_0000001955726494__b1990165018112542">binary</strong> are supported. The input query vector must be a Base64-encoded character string.</p>
<p id="EN-US_TOPIC_0000001955726494__p231916518323">Only OpenSearch 2.19.0 clusters support this function.</p>
</td>
</tr>
</tbody>
</table>
</div>
</div>
</div>
<div>
<div class="familylinks">
<div class="parentlink"><strong>Parent topic:</strong> <a href="css_01_0101.html">Configuring Vector Search for OpenSearch Clusters</a></div>
</div>
</div>
View Git Blame Copy Permalink