Skip to content

armgit / xerj

Public Imported from GitHub
0

AI search engine and MCP server for bots and agents. One Rust binary indexes code, docs, logs, JSONL/CSV and PDFs for BM25 full-text, vector and hybrid search, code search without grep, agent memory and RAG. Elasticsearch-compatible API; Corpus Hub serves hundreds of reference memory corpora.

38.5 KB
Raw Blame
<!DOCTYPE html>
<html lang="en" data-theme="day">
<head>
<!-- ga4:begin — generated by scripts/seo/seo_head.py · do not hand-edit -->
<!-- Google tag (gtag.js) -->
<script async src="https://www.googletagmanager.com/gtag/js?id=G-Z77ETCDJYL"></script>
<script>
  window.dataLayer = window.dataLayer || [];
  function gtag(){dataLayer.push(arguments);}
  gtag('js', new Date());

  gtag('config', 'G-Z77ETCDJYL');
</script>
<!-- ga4:end -->
<meta charset="UTF-8">
<meta name="viewport" content="width=device-width, initial-scale=1.0">
<title>XERJ.ai · Docs · Vectors &amp; kNN</title>
<meta name="description" content="Dense vectors and kNN in XERJ — full-precision storage, a persisted HNSW graph, exact rescoring of candidates, and measured recall against ES 8.13.4.">
<link rel="preconnect" href="https://fonts.googleapis.com">
<link rel="preconnect" href="https://fonts.gstatic.com" crossorigin>
<link href="https://fonts.googleapis.com/css2?family=Big+Shoulders+Display:wght@400;700;800;900&family=Inter:wght@400;500;600;700&family=IBM+Plex+Sans:wght@400;500;600;700&family=JetBrains+Mono:wght@400;500;700&display=swap" rel="stylesheet">
<link rel="icon" href="data:image/svg+xml,<svg xmlns='http://www.w3.org/2000/svg' viewBox='0 0 100 100'><rect width='100' height='100' fill='%230b0b0d'/><text x='50' y='68' font-size='60' font-family='Big Shoulders Display,sans-serif' font-weight='900' fill='%23ffc400' text-anchor='middle'>X</text></svg>">
<script src="/theme.js"></script>
<link rel="stylesheet" href="/style.css?v=90f643e">
<!-- seo:begin — generated by scripts/seo/fix_heads.py · do not hand-edit -->
<link rel="canonical" href="https://xerj.org/docs/vectors">
<meta property="og:site_name" content="XERJ.ai">
<meta property="og:type" content="article">
<meta property="og:title" content="XERJ.ai · Docs · Vectors &amp; kNN">
<meta property="og:description" content="Dense vectors and kNN in XERJ — full-precision storage, a persisted HNSW graph, exact rescoring of candidates, and measured recall against ES 8.13.4.">
<meta property="og:url" content="https://xerj.org/docs/vectors">
<meta property="og:image" content="https://xerj.org/og/xerj-card.png">
<meta property="og:image:width" content="1200">
<meta property="og:image:height" content="630">
<meta property="og:image:alt" content="XERJ.ai — the search engine built for AI agents">
<meta property="og:locale" content="en_US">
<meta name="twitter:card" content="summary_large_image">
<meta name="twitter:title" content="XERJ.ai · Docs · Vectors &amp; kNN">
<meta name="twitter:description" content="Dense vectors and kNN in XERJ — full-precision storage, a persisted HNSW graph, exact rescoring of candidates, and measured recall against ES 8.13.4.">
<meta name="twitter:image" content="https://xerj.org/og/xerj-card.png">
<script type="application/ld+json">
{
  "@context": "https://schema.org",
  "@type": "Organization",
  "@id": "https://xerj.org/#organization",
  "name": "XERJ",
  "alternateName": "XERJ.ai",
  "url": "https://xerj.org/",
  "logo": "https://xerj.org/brandbook/xerj-mark.svg",
  "description": "A new way for AI to search for data: one local Rust binary that indexes your code, docs, specs and logs, plus Corpus Hub packs, and returns cited answers.",
  "sameAs": [
    "https://github.com/xerj-org/xerj",
    "https://x.com/xerj_org"
  ]
}
</script>
<script type="application/ld+json">
{
  "@context": "https://schema.org",
  "@type": "WebSite",
  "@id": "https://xerj.org/#website",
  "name": "XERJ.ai",
  "url": "https://xerj.org/",
  "inLanguage": "en",
  "publisher": {
    "@id": "https://xerj.org/#organization"
  }
}
</script>
<script type="application/ld+json">
{
  "@context": "https://schema.org",
  "@type": "TechArticle",
  "headline": "XERJ.ai · Docs · Vectors & kNN",
  "description": "Dense vectors and kNN in XERJ — full-precision storage, a persisted HNSW graph, exact rescoring of candidates, and measured recall against ES 8.13.4.",
  "url": "https://xerj.org/docs/vectors",
  "image": "https://xerj.org/og/xerj-card.png",
  "inLanguage": "en",
  "datePublished": "2026-06-29",
  "dateModified": "2026-09-08",
  "author": {
    "@id": "https://xerj.org/#organization"
  },
  "publisher": {
    "@id": "https://xerj.org/#organization"
  },
  "isPartOf": {
    "@id": "https://xerj.org/#website"
  },
  "mainEntityOfPage": {
    "@type": "WebPage",
    "@id": "https://xerj.org/docs/vectors"
  }
}
</script>
<script type="application/ld+json">
{
  "@context": "https://schema.org",
  "@type": "BreadcrumbList",
  "itemListElement": [
    {
      "@type": "ListItem",
      "position": 1,
      "name": "Home",
      "item": "https://xerj.org/"
    },
    {
      "@type": "ListItem",
      "position": 2,
      "name": "Docs",
      "item": "https://xerj.org/docs/"
    },
    {
      "@type": "ListItem",
      "position": 3,
      "name": "Vectors & kNN"
    }
  ]
}
</script>
<!-- seo:end -->
</head>
<body>


<nav class="nav" aria-label="Primary">
  <a href="/" class="brand">XERJ</a>
  <a href="/product">PRODUCT</a>
  <a href="/use-cases">USE CASES</a>
  <a href="/answers/">ANSWERS</a>
  <a href="/benchmarks/">BENCHMARKS</a>
  <a href="/blog/">BLOG</a>
  <span class="spacer"></span>
  <a href="/for-agents">FOR AGENTS</a>
  <a href="/docs/" class="active">DOCS</a>
  <a href="/playground">DASHBOARDS</a>
  <a href="https://hub.xerj.org" rel="external">HUB</a>
  <a class="nav-stars" href="https://github.com/xerj-org/xerj" rel="external" title="XERJ on GitHub" aria-label="XERJ on GitHub, 14,225 stars"><svg viewBox="0 0 16 16" width="14" height="14" aria-hidden="true"><path d="M8 0C3.58 0 0 3.58 0 8c0 3.54 2.29 6.53 5.47 7.59.4.07.55-.17.55-.38 0-.19-.01-.82-.01-1.49-2.01.37-2.53-.49-2.69-.94-.09-.23-.48-.94-.82-1.13-.28-.15-.68-.52-.01-.53.63-.01 1.08.58 1.23.82.72 1.21 1.87.87 2.33.66.07-.52.28-.87.51-1.07-1.78-.2-3.64-.89-3.64-3.95 0-.87.31-1.59.82-2.15-.08-.2-.36-1.02.08-2.12 0 0 .67-.21 2.2.82.64-.18 1.32-.27 2-.27.68 0 1.36.09 2 .27 1.53-1.04 2.2-.82 2.2-.82.44 1.1.16 1.92.08 2.12.51.56.82 1.27.82 2.15 0 3.07-1.87 3.75-3.65 3.95.29.25.54.73.54 1.48 0 1.07-.01 1.93-.01 2.2 0 .21.15.46.55.38A8.013 8.013 0 0016 8c0-4.42-3.58-8-8-8z"/></svg><span class="nav-stars-n" data-gh-stars="14225">★ 14.2k</span></a>
  <a href="https://x.com/xerj_org" rel="external" title="XERJ on X" style="display:inline-flex;align-items:center"><svg width="15" height="15" viewBox="0 0 24 24" fill="currentColor" aria-hidden="true"><path d="M18.244 2.25h3.308l-7.227 8.26 8.502 11.24H16.17l-5.214-6.817L4.99 21.75H1.68l7.73-8.835L1.254 2.25H8.08l4.713 6.231zm-1.161 17.52h1.833L7.084 4.126H5.117z"/></svg></a>
  <a href="https://discord.com/invite/yetXFusXfs" rel="external" title="XERJ on Discord" style="display:inline-flex;align-items:center"><svg width="15" height="15" viewBox="0 0 24 24" fill="currentColor" aria-hidden="true"><path d="M20.317 4.37a19.79 19.79 0 00-4.885-1.515.074.074 0 00-.079.037c-.21.375-.444.864-.608 1.25a18.27 18.27 0 00-5.487 0 12.64 12.64 0 00-.617-1.25.077.077 0 00-.079-.037A19.74 19.74 0 003.677 4.37a.07.07 0 00-.032.027C.533 9.046-.32 13.58.099 18.057a.082.082 0 00.031.057 19.9 19.9 0 005.993 3.03.078.078 0 00.084-.028c.462-.63.874-1.295 1.226-1.994a.076.076 0 00-.041-.106 13.1 13.1 0 01-1.872-.892.077.077 0 01-.008-.128 10.2 10.2 0 00.372-.292.074.074 0 01.077-.01c3.928 1.793 8.18 1.793 12.062 0a.074.074 0 01.078.01c.12.098.246.198.373.292a.077.077 0 01-.006.127 12.3 12.3 0 01-1.873.892.077.077 0 00-.041.107c.36.698.772 1.362 1.225 1.993a.076.076 0 00.084.028 19.84 19.84 0 006.002-3.03.077.077 0 00.032-.054c.5-5.177-.838-9.674-3.549-13.66a.061.061 0 00-.031-.03zM8.02 15.33c-1.183 0-2.157-1.085-2.157-2.419 0-1.333.956-2.419 2.157-2.419 1.21 0 2.176 1.096 2.157 2.42 0 1.333-.956 2.418-2.157 2.418zm7.975 0c-1.183 0-2.157-1.085-2.157-2.419 0-1.333.955-2.419 2.157-2.419 1.21 0 2.176 1.096 2.157 2.42 0 1.333-.946 2.418-2.157 2.418z"/></svg></a>
  <span class="theme">
    <button type="button" data-theme="day" class="active">DAY</button>
    <span class="dash">·</span>
    <button type="button" data-theme="night">NIGHT</button>
  </span>
</nav>

<div class="docs-search" id="docsSearch">
  <input type="text" id="docsSearchInput" placeholder="Search docs · keys, flags, endpoints, playbooks…" aria-label="Search docs" autocomplete="off" spellcheck="false">
  <span class="hint">⌘ K</span>
  <div class="results" id="docsSearchResults" aria-live="polite"></div>
</div>

<div class="docs-shell">
<nav class="docs-nav" aria-label="Docs">  <div class="group">
    <span class="label">START</span>
    <a href="/docs/" class=""><span class="eyebrow">01</span>Overview</a>
      <a href="/docs/quickstart" class=""><span class="eyebrow">02</span>Quickstart</a>
      <a href="/docs/install" class=""><span class="eyebrow">03</span>Install</a>
      <a href="/docs/migration-from-es" class=""><span class="eyebrow">04</span>Migrate from ES</a>
  </div>
  <div class="group">
    <span class="label">REFERENCE</span>
    <a href="/docs/cli" class=""><span class="eyebrow">01</span>CLI</a>
      <a href="/docs/config" class=""><span class="eyebrow">02</span>Config TOML</a>
      <a href="/docs/env" class=""><span class="eyebrow">03</span>Environment</a>
      <a href="/docs/api-native" class=""><span class="eyebrow">04</span>Native REST API</a>
      <a href="/docs/api-es-compat" class=""><span class="eyebrow">05</span>ES-compat API</a>
  </div>
  <div class="group">
    <span class="label">DATA MODEL</span>
    <a href="/docs/queries" class=""><span class="eyebrow">01</span>Query types</a>
      <a href="/docs/aggregations" class=""><span class="eyebrow">02</span>Aggregations</a>
      <a href="/docs/analyzers" class=""><span class="eyebrow">03</span>Analyzers</a>
      <a href="/docs/vectors" class="active"><span class="eyebrow">04</span>Vectors & kNN</a>
  </div>
  <div class="group">
    <span class="label">ENGINE</span>
    <a href="/docs/ingest" class=""><span class="eyebrow">01</span>Ingest pipelines</a>
      <a href="/docs/storage" class=""><span class="eyebrow">02</span>Storage & WAL</a>
      <a href="/docs/compression" class=""><span class="eyebrow">03</span>Compression</a>
      <a href="/docs/clustering" class=""><span class="eyebrow">04</span>Clustering</a>
      <a href="/docs/metrics" class=""><span class="eyebrow">05</span>Metrics</a>
      <a href="/docs/security" class=""><span class="eyebrow">06</span>Auth & TLS</a>
  </div>
  <div class="group">
    <span class="label">OPERATIONS</span>
    <a href="/docs/operations" class=""><span class="eyebrow">01</span>Running in production</a>
      <a href="/docs/backup-restore" class=""><span class="eyebrow">02</span>Backup & restore</a>
      <a href="/docs/upgrades" class=""><span class="eyebrow">03</span>Upgrades</a>
      <a href="/docs/troubleshooting" class=""><span class="eyebrow">04</span>Troubleshooting</a>
  </div>
  <div class="group">
    <span class="label">PLAYBOOKS</span>
    <a href="/docs/playbooks/siem" class=""><span class="eyebrow">01</span>SIEM</a>
      <a href="/docs/playbooks/log-analytics" class=""><span class="eyebrow">02</span>Log analytics</a>
      <a href="/docs/playbooks/vector-search" class=""><span class="eyebrow">03</span>Vector · RAG</a>
      <a href="/docs/playbooks/full-text" class=""><span class="eyebrow">04</span>Full-text search</a>
      <a href="/docs/playbooks/observability" class=""><span class="eyebrow">05</span>Observability</a>
  </div>
  <div class="group">
    <span class="label">RECIPES</span>
    <a href="/docs/recipes/" class=""><span class="eyebrow">00</span>All recipes</a>
    <a href="/docs/recipes/semantic-search-rag" class=""><span class="eyebrow">01</span>Semantic search &amp; RAG</a>
    <a href="/docs/recipes/agentic-memory" class=""><span class="eyebrow">02</span>Agent memory</a>
    <a href="/docs/recipes/passage-retrieval" class=""><span class="eyebrow">03</span>Passage retrieval</a>
    <a href="/docs/recipes/vector-search-knn" class=""><span class="eyebrow">04</span>Vector search &middot; kNN</a>
    <a href="/docs/recipes/vector-quantization" class=""><span class="eyebrow">05</span>Vector quantization</a>
    <a href="/docs/recipes/hybrid-search" class=""><span class="eyebrow">06</span>Hybrid search</a>
    <a href="/docs/recipes/log-analytics" class=""><span class="eyebrow">07</span>Log analytics</a>
    <a href="/docs/recipes/anomaly-detection" class=""><span class="eyebrow">08</span>Anomaly detection</a>
    <a href="/docs/recipes/continuous-anomaly-datafeeds" class=""><span class="eyebrow">09</span>Anomaly datafeeds</a>
    <a href="/docs/recipes/migrate-from-elasticsearch" class=""><span class="eyebrow">10</span>Migrate from ES</a>
  </div>
  <div class="group">
    <span class="label">CORPUS HUB</span>
    <a href="/docs/corpus-hub" class=""><span class="eyebrow">01</span>Ask, pull &amp; contribute</a>
  </div>
</nav>

<main class="docs-content">
  <!-- Print-only branded header (hidden on screen, visible in PDF) -->
  <div class="print-header" style="display:none;">XERJ</div>

  <div class="breadcrumb"><a href="/docs/">DOCS</a><span class="sep">·</span>DATA MODEL<span class="sep">·</span>VECTORS
    <button class="pdf-btn" onclick="window.print()" title="Save this page as PDF">
      <svg viewBox="0 0 24 24"><path d="M12 3v12M12 15l-4-4M12 15l4-4"/><path d="M4 17v2a2 2 0 002 2h12a2 2 0 002-2v-2"/></svg>
      PDF
    </button>
  </div>
  <div class="kicker"><span>04 · DATA MODEL</span></div>
  <h1>Vectors & kNN</h1>


  <p>XERJ stores your vectors in full at <strong>ingest</strong> time and maintains a <strong>persisted HNSW graph</strong> alongside them. An unfiltered <code class="inline">knn</code> on a full-precision cosine field (≥1,024 docs) is served by an approximate HNSW beam search whose candidates are <strong>exact-rescored</strong>, so returned scores match the exact path bit-for-bit — measured recall@10 1.00 on the official bench query, 100-probe mean 0.976 (ES 8.13.4 same protocol: 0.937). <code class="inline">num_candidates</code> sets the beam width (floored at 800 to match Elasticsearch's per-segment candidate semantics). A <code class="inline">filter</code>, non-cosine similarity, SQ8 quantization, a small index, or a stale graph falls back to the <strong>exact brute-force</strong> scan: every candidate is scored and the true nearest neighbours returned (recall 1.00; latency scales with vectors scanned). Distance metric and quantization are set in config. See <a href="#knn-query">KNN query</a> below.</p>

  <h2 id="embedders">Embedding backends — how vectors get made</h2>
  <p>Everything above assumes a vector already exists on the doc. XERJ can produce that vector for you from a <code class="inline">semantic_text</code> field, and you pick <em>how</em> once at the server with <code class="inline">--embed-mode</code> (or <code class="inline">embedding.mode</code> in config, or <code class="inline">XERJ_EMBED_MODE</code>). The mapping and the <code class="inline">semantic</code> query are identical across all three:</p>
  <div class="cfg" style="grid-template-columns: 1fr 1.2fr 3fr;">
    <div class="head">MODE</div><div class="head">SHIPS / NEEDS</div><div class="head">WHAT IT IS</div>
    <div class="cell k">lexical</div><div class="cell t">built-in · default</div><div class="cell dsc">Feature-hash embedder. Offline, deterministic, zero download. Maps overlapping and related vocabulary into vector space — strong when question and answer share terms, not a neural sentence encoder.</div>
    <div class="cell k">neural</div><div class="cell t">built-in · ~90 MB on first use</div><div class="cell dsc">BERT sentence encoder (<code class="inline">all-MiniLM-L6-v2</code>) running in-process via <code class="inline">candle</code> — pure Rust, no Python, no external service. The model auto-downloads (~90 MB) once, or point <code class="inline">embedding.local_model_dir</code> at a pre-staged copy for a fully <strong>air-gapped</strong> install. Captures meaning with no shared term (<em>rotate ≈ regenerate</em>).</div>
    <div class="cell k">proxy</div><div class="cell t">external endpoint</div><div class="cell dsc">Bring your own model: XERJ calls any OpenAI-compatible <code class="inline">/v1/embeddings</code> (OpenAI, Cohere, a local vLLM). Set <code class="inline">embedding.default_endpoint</code>. Embedding happens at ingest time.</div>
  </div>
  <p>Turn on the built-in neural embedder with a single server flag — no mapping or query change:</p>
<pre class="code">xerj --insecure --data-dir ./data --embed-mode neural</pre>
  <p>See the <a href="/docs/recipes/semantic-search-rag">Semantic search &amp; RAG recipe</a> for the end-to-end walkthrough.</p>

  <h2 id="hnsw">hnsw_* config keys — accepted, no effect on search</h2>
  <div class="cfg" style="grid-template-columns: 1.5fr 1fr 1fr 2.5fr;">
    <div class="head">KEY</div>
    <div class="head">TYPE</div>
    <div class="head">DEFAULT</div>
    <div class="head">DESCRIPTION</div>

    <div class="cell k">hnsw_m</div>
    <div class="cell t">u32</div>
    <div class="cell d">16</div>
    <div class="cell dsc">Accepted and validated; the built-in HNSW graph currently builds with fixed M=16 — this key is not yet wired to the build.</div>

    <div class="cell k">hnsw_ef_construction</div>
    <div class="cell t">u32</div>
    <div class="cell d">200</div>
    <div class="cell dsc">Accepted; must be ≥ hnsw_m. The graph currently builds with fixed ef_construction=200 — this key is not yet wired to the build.</div>

    <div class="cell k">hnsw_ef_search</div>
    <div class="cell t">u32</div>
    <div class="cell d">100</div>
    <div class="cell dsc">Accepted but not read by the serving path — the ANN beam width comes from the request's num_candidates (floored at 800).</div>

    <div class="cell k">default_metric</div>
    <div class="cell t">enum</div>
    <div class="cell d">"cosine"</div>
    <div class="cell dsc">"cosine" · "dot_product" · "euclidean". Accepted but not wired: similarity comes from the field mapping, and a mapping without one gets cosine.</div>
  </div>

  <h2 id="quantization">Quantization</h2>
  <p><strong>Quantization is off by default</strong> — <code class="inline">[vector] default_quantization = "none"</code>, full-precision f32. Opt in per field with <code class="inline">index_options.type: int8_hnsw</code> (or <code class="inline">int8_flat</code>) on the mapping. The node-wide key is not wired as a default: only its <code class="inline">"binary"</code> refusal reads it, so <code class="inline">"scalar8"</code> there quantizes nothing.</p>
  <ul>
    <li><strong>none</strong> — full-precision f32. <em>The default.</em> Use when recall matters more than RAM.</li>
    <li><strong>scalar8</strong> — 8-bit per dimension, 1–2% recall loss. The one quantizer the serving path reads. It buys <strong>precision, not memory</strong>: the scan reads the full-precision vector from <code class="inline">_source</code> and quantizes it per query, so nothing shrinks in RAM, and a <code class="inline">scalar8</code> <code class="inline">_score</code> (and the order it produces) depends on the candidate set — see the <a href="/docs/recipes/vector-quantization">quantization recipe</a> and <a href="https://github.com/xerj-org/xerj/issues/392">#392</a>.</li>
    <li><strong>binary</strong> — 1-bit per dimension. Not implemented: the server <strong>refuses to start</strong> on <code class="inline">default_quantization = "binary"</code> rather than silently storing full precision while claiming a 32× saving.</li>
  </ul>
  <p>A 4-bit quantizer exists in the vector crate but no config or mapping value reaches it, so <code class="inline">scalar4</code> is not a mode you can select in this build.</p>
  <p><strong>A mapping's <code class="inline">quantization</code> value is honoured or refused — never accepted and ignored</strong> (<a href="https://github.com/xerj-org/xerj/issues/275">#275</a>, fixed). <code class="inline">"scalar8"</code> (alias <code class="inline">"int8"</code>) and <code class="inline">"none"</code> are the values this build implements. Anything else — <code class="inline">"binary"</code>, <code class="inline">"scalar4"</code>, <code class="inline">"int4"</code>, or a typo such as <code class="inline">"sq8"</code> — is now a <code class="inline">400 mapper_parsing_exception</code> naming the value and the accepted set, on both <code class="inline">PUT /&lt;index&gt;</code> and <code class="inline">PUT /&lt;index&gt;/_mapping</code>, and the index is not created. Before the fix those values answered <code class="inline">200</code>, were echoed back verbatim by <code class="inline">GET /_mapping</code>, and were then ignored, so the field stored full-precision f32 while the mapping you read back to confirm it said otherwise. This is the same guard <code class="inline">Config::validate</code> applies to <code class="inline">[vector] default_quantization</code> at startup, now at the per-field surface that sets the same field.</p>
  <p>The strictness is confined to XERJ's own <code class="inline">quantization</code> key. Elasticsearch's <code class="inline">index_options.type</code> stays permissive — <code class="inline">hnsw</code>, <code class="inline">flat</code>, <code class="inline">int4_*</code> and <code class="inline">bbq_*</code> mappings still create, served by the exact f32 scan — because an ES mapping must not be rejected on a wire-compatibility surface.</p>

  <h2 id="knn-query">KNN query</h2>
<pre class="code">{
  "knn": {
    "field":      "embedding",
    "query_vector": [0.12, 0.08, -0.31, ...],
    "k":          20,
    "num_candidates": 200
  }
}</pre>
  <p><code class="inline">num_candidates</code> is the ANN beam width on the HNSW-served (unfiltered, ≥1,024-doc) path — larger values trade latency for recall, floored at 800 to match Elasticsearch's per-segment candidate semantics. On the exact scan (filtered kNN, small indexes, SQ8 fields) it has no effect on results.</p>

  <h2>Hybrid — BM25 + KNN in one planner pass</h2>
<pre class="code">{
  "hybrid": {
    "fusion": "rrf",
    "queries": [
      { "match": { "message": "kernel panic on reboot" } },
      { "knn":   { "field": "embedding", "query_vector": [...], "k": 50 } }
    ]
  }
}</pre>

  <p class="source-note">Source · engine/crates/xerj-engine/src/index.rs (run_knn_hnsw + exact fallback)</p>
  

<div class="next">
      <div class="step"><span class="dir">◀ PREV</span><a href="/docs/analyzers">Analyzers</a></div>
      <div class="step right"><span class="dir">NEXT ▶</span><a href="/docs/ingest">Ingest pipelines</a></div>
    </div>

  <!-- Print-only branded footer (hidden on screen, visible in PDF) -->
  <div class="print-footer" style="display:none;">
    <span>XERJ.AI · V1.0.0-RC.100 · 2026-08-11</span>
    <a href="https://xerj.org/docs/">https://xerj.org/docs/</a>
    <span>Vectors & kNN</span>
  </div>
</main>
</div>

<!-- seo:footer-begin — generated by scripts/seo/fix_links.py · do not hand-edit -->
<nav class="footer-map" aria-label="Site map">
  <div class="fm-col">
    <span class="fm-h">PRODUCT</span>
    <a href="/product">PRODUCT</a>
    <a href="/architecture/">ARCHITECTURE</a>
    <a href="/benchmarks/">BENCHMARKS</a>
    <a href="/pricing/">PRICING</a>
    <a href="/security/">SECURITY</a>
    <a href="/playground">DASHBOARDS</a>
    <a href="/demo/">REAL-DATA DEMO</a>
    <a href="/aise-demo">PRODUCT TOUR</a>
  </div>
  <div class="fm-col">
    <span class="fm-h">LEARN</span>
    <a href="/answers/">ANSWERS</a>
    <a href="/compare/">COMPARE</a>
    <a href="/use-cases">USE CASES</a>
    <a href="/case-studies">CASE STUDIES</a>
    <a href="/industries/">INDUSTRIES</a>
    <a href="/blog/">BLOG</a>
  </div>
  <div class="fm-col">
    <span class="fm-h">SOLUTIONS</span>
    <a href="/solutions/">SOLUTIONS</a>
    <a href="/industries/finserv">FINANCIAL SERVICES</a>
    <a href="/industries/public-sector">PUBLIC SECTOR</a>
    <a href="/industries/healthcare">HEALTHCARE</a>
    <a href="/industries/retail">RETAIL</a>
    <a href="/use-cases/code-security-audit">AI CODE-SECURITY AUDIT</a>
    <a href="/use-cases/semantic-analytics">SEMANTIC ANALYTICS</a>
    <a href="/resources/">SALES KIT</a>
  </div>
  <div class="fm-col">
    <span class="fm-h">DOCS</span>
    <a href="/docs/">DOCUMENTATION</a>
    <a href="/docs/quickstart">QUICKSTART</a>
    <a href="/docs/install">INSTALL</a>
    <a href="/docs/migration-from-es">MIGRATE FROM ELASTICSEARCH</a>
    <a href="/docs/recipes/">RECIPES</a>
    <a href="/docs/agents/">DOCS FOR AI</a>
    <a href="/for-agents">FOR AI AGENTS</a>
    <a href="/agent-search/">AGENTIC SEARCH DEMO</a>
    <a href="/llms.txt">LLMS.TXT</a>
    <a href="https://github.com/xerj-org/xerj" rel="external">GITHUB</a>
  </div>
</nav>
<!-- seo:footer-end -->
<footer class="footer">
  <span>XERJ.AI · DOCS · V1.0.0-RC.100</span>
  <span>READ THE SOURCE</span>
  <span><a href="/">HOME</a> · <a href="/product">PRODUCT</a> · <a href="/answers/">ANSWERS</a> · <a href="/playground">DASHBOARDS</a> &middot; <a href="/llms.txt">LLMS.TXT</a></span>
</footer>

<script type="application/json" id="docs-index">[{"title":"Corpus Hub","slug":"/docs/corpus-hub","kicker":"CORPUS HUB","snip":"Ask a pinned, signed, pre-indexed corpus: xerj ask, xerj hub pull/sync, the MCP hub tools, and how to contribute a corpus."},{"title":"Overview","slug":"/docs/","kicker":"START","snip":""},{"title":"Quickstart","slug":"/docs/quickstart","kicker":"START","snip":""},{"title":"Install","slug":"/docs/install","kicker":"START","snip":""},{"title":"Migrate from ES","slug":"/docs/migration-from-es","kicker":"START","snip":""},{"title":"CLI","slug":"/docs/cli","kicker":"REFERENCE","snip":""},{"title":"Config TOML","slug":"/docs/config","kicker":"REFERENCE","snip":""},{"title":"Environment","slug":"/docs/env","kicker":"REFERENCE","snip":""},{"title":"Native REST API","slug":"/docs/api-native","kicker":"REFERENCE","snip":""},{"title":"ES-compat API","slug":"/docs/api-es-compat","kicker":"REFERENCE","snip":""},{"title":"Query types","slug":"/docs/queries","kicker":"DATA MODEL","snip":""},{"title":"Aggregations","slug":"/docs/aggregations","kicker":"DATA MODEL","snip":""},{"title":"Analyzers","slug":"/docs/analyzers","kicker":"DATA MODEL","snip":""},{"title":"Vectors & kNN","slug":"/docs/vectors","kicker":"DATA MODEL","snip":""},{"title":"Ingest pipelines","slug":"/docs/ingest","kicker":"ENGINE","snip":""},{"title":"Storage & WAL","slug":"/docs/storage","kicker":"ENGINE","snip":""},{"title":"Compression","slug":"/docs/compression","kicker":"ENGINE","snip":""},{"title":"Clustering","slug":"/docs/clustering","kicker":"ENGINE","snip":""},{"title":"Metrics","slug":"/docs/metrics","kicker":"ENGINE","snip":""},{"title":"Auth & TLS","slug":"/docs/security","kicker":"ENGINE","snip":""},{"title":"Running in production","slug":"/docs/operations","kicker":"OPERATIONS","snip":""},{"title":"Backup & restore","slug":"/docs/backup-restore","kicker":"OPERATIONS","snip":""},{"title":"Upgrades","slug":"/docs/upgrades","kicker":"OPERATIONS","snip":""},{"title":"Troubleshooting","slug":"/docs/troubleshooting","kicker":"OPERATIONS","snip":""},{"title":"SIEM","slug":"/docs/playbooks/siem","kicker":"PLAYBOOKS","snip":""},{"title":"Log analytics","slug":"/docs/playbooks/log-analytics","kicker":"PLAYBOOKS","snip":""},{"title":"Vector · RAG","slug":"/docs/playbooks/vector-search","kicker":"PLAYBOOKS","snip":""},{"title":"Full-text search","slug":"/docs/playbooks/full-text","kicker":"PLAYBOOKS","snip":""},{"title":"Observability","slug":"/docs/playbooks/observability","kicker":"PLAYBOOKS","snip":""},{"title":"CLI · --config","slug":"/docs/cli#config-flag","kicker":"CLI","snip":"-c, --config <PATH>  Path to the TOML config file."},{"title":"CLI · --data-dir","slug":"/docs/cli#data-dir-flag","kicker":"CLI","snip":"-d, --data-dir <PATH>  Override [server] data_dir at the command line."},{"title":"CLI · --insecure","slug":"/docs/cli#insecure-flag","kicker":"CLI","snip":"-k, --insecure  Disable TLS and API-key auth. Dev only."},{"title":"[server] rest_port","slug":"/docs/config#server","kicker":"CONFIG","snip":"u16 · default 8080 · Native REST API listener port."},{"title":"[server] es_compat_port","slug":"/docs/config#server","kicker":"CONFIG","snip":"u16 · default 9200 · Elasticsearch-wire-format port. 0 = OS-assigned port."},{"title":"[storage] wal_sync","slug":"/docs/config#storage","kicker":"CONFIG","snip":"enum · default batched · sync · batched · async. Durability knob."},{"title":"[vector] hnsw_ef_search","slug":"/docs/config#vector","kicker":"CONFIG","snip":"u32 · default 100 · Accepted but not wired — the beam width comes from the request num_candidates."},{"title":"[vector] default_quantization","slug":"/docs/config#vector","kicker":"CONFIG","snip":"enum · none (default) · scalar8. binary is refused at startup; scalar4 is not accepted."},{"title":"GET /v1/health","slug":"/docs/api-native#health","kicker":"NATIVE API","snip":"Liveness + readiness. {status, uptime_s, version}."},{"title":"GET /v1/metrics","slug":"/docs/api-native#metrics","kicker":"NATIVE API","snip":"Prometheus text format. Scrape-friendly."},{"title":"POST /v1/indices/:name/turbo-ingest","slug":"/docs/api-native#turbo-ingest","kicker":"NATIVE API","snip":"NDJSON bulk path with parallel tokenization."},{"title":"POST /v1/indices/:name/search","slug":"/docs/api-native#search","kicker":"NATIVE API","snip":"Unified query endpoint — full-text, term, range, KNN, hybrid."},{"title":"POST /:index/_bulk","slug":"/docs/api-es-compat#bulk","kicker":"ES API","snip":"NDJSON bulk. Actions: index, create, update, delete."},{"title":"POST /:index/_search","slug":"/docs/api-es-compat#search","kicker":"ES API","snip":"ES query DSL. Drop-in migration path."},{"title":"match_all","slug":"/docs/queries#match-all","kicker":"QUERY","snip":"Everything in the index."},{"title":"bool","slug":"/docs/queries#bool","kicker":"QUERY","snip":"must / should / must_not / filter clauses."},{"title":"knn","slug":"/docs/queries#knn","kicker":"QUERY","snip":"Vector kNN — HNSW-served when unfiltered (exact-rescored); filtered runs the exact scan."},{"title":"semantic","slug":"/docs/queries#semantic","kicker":"QUERY","snip":"Embed the query text at query time and search the vector field."},{"title":"hybrid","slug":"/docs/queries#hybrid","kicker":"QUERY","snip":"RRF / linear fusion of sub-queries."},{"title":"terms aggregation","slug":"/docs/aggregations#terms","kicker":"AGG","snip":"Exact, pre-computed terms histograms over keyword fields."},{"title":"date_histogram","slug":"/docs/aggregations#date-histogram","kicker":"AGG","snip":"Time-bucketed counts."},{"title":"kNN serving","slug":"/docs/vectors#knn-query","kicker":"VECTORS","snip":"Unfiltered kNN via persisted HNSW + exact rescore; filtered kNN exact scan."},{"title":"SQ8 quantization","slug":"/docs/vectors#quantization","kicker":"VECTORS","snip":"8-bit scalar quantization. int8 scoring precision, not a memory saving (#392)."},{"title":"turbo-ingest","slug":"/docs/ingest#turbo","kicker":"INGEST","snip":"NDJSON bulk path. Parallel tokenization across cores."},{"title":"delta-of-delta","slug":"/docs/compression#delta-of-delta","kicker":"COMPRESSION","snip":"Timestamp encoding. Apache/Nginx/ISO auto-detect."},{"title":"dictionary encoding","slug":"/docs/compression#dictionary","kicker":"COMPRESSION","snip":"Up to 256 unique strings per column."},{"title":"api_key","slug":"/docs/security#api-key","kicker":"SECURITY","snip":"Static admin key. Auto-generated on first run."},{"title":"SIEM playbook","slug":"/docs/playbooks/siem","kicker":"PLAYBOOK","snip":"Terms agg over source IPs for SIEM, with detection rules."},{"title":"RAG playbook","slug":"/docs/playbooks/vector-search","kicker":"PLAYBOOK","snip":"Embeddings, kNN, hybrid search, BM25 + kNN fusion."},{"title":"[cluster] enabled","slug":"/docs/config#cluster","kicker":"CONFIG","snip":"bool · default false · Enable multi-node Raft cluster mode."},{"title":"[cluster] peers","slug":"/docs/config#cluster","kicker":"CONFIG","snip":"array · \"host:port=host:port\" · label must equal address (#1171)."},{"title":"[cluster] port","slug":"/docs/config#cluster","kicker":"CONFIG","snip":"u16 · default 9300 · Intra-cluster Raft + search port."},{"title":"[embedding] default_endpoint","slug":"/docs/config#embedding","kicker":"CONFIG","snip":"string · OpenAI-compatible embeddings URL for auto-embedding on ingest."},{"title":"[embedding] default_model","slug":"/docs/config#embedding","kicker":"CONFIG","snip":"string · Model name (e.g. text-embedding-3-small, nomic-embed-text)."},{"title":"Raft","slug":"/docs/clustering","kicker":"CLUSTER","snip":"Embedded Raft ring — leader election, heartbeats, authenticated transport."},{"title":"Leader failover","slug":"/docs/clustering","kicker":"CLUSTER","snip":"Kill the leader of a 3-node ring: survivors elect a new leader in ~250 ms."},{"title":"Cluster honesty","slug":"/docs/clustering","kicker":"CLUSTER","snip":"Create-time number_of_replicas copies acknowledged document writes. Bulk, replica promotion, and snapshot catch-up are still open (#1170)."},{"title":"Systemd unit","slug":"/docs/operations","kicker":"OPS","snip":"Production unit file with hardening directives."},{"title":"Readiness probe","slug":"/docs/operations","kicker":"OPS","snip":"GET /v1/health/ready — only 200 once WAL replay is complete."},{"title":"Config reload","slug":"/docs/operations","kicker":"OPS","snip":"SIGHUP reloads most keys without restart."},{"title":"Hot backup","slug":"/docs/backup-restore","kicker":"OPS","snip":"rsync the data dir after POST /v1/admin/flush."},{"title":"S3 backup","slug":"/docs/backup-restore","kicker":"OPS","snip":"POST /v1/admin/backup → any S3-compatible endpoint."},{"title":"Cluster upgrade","slug":"/docs/upgrades","kicker":"OPS","snip":"The cluster wire has no version negotiation — full stop/start, not rolling."},{"title":"Merge pressure","slug":"/docs/troubleshooting","kicker":"OPS","snip":"Segment count climbing → raise merge.max_concurrent + io_rate."},{"title":"Cluster flapping","slug":"/docs/troubleshooting","kicker":"OPS","snip":"Sustained leader churn is a defect — capture RUST_LOG=xerj_cluster=debug and file it."},{"title":"WAL warnings after a crash","slug":"/docs/troubleshooting","kicker":"OPS","snip":"Repaired on startup; no manual repair command."},{"title": "Recipes", "slug": "/docs/recipes/", "kicker": "RECIPES", "snip": "Task-oriented AI & search recipes, each verified end-to-end against a live XERJ."},{"title": "Semantic search & RAG", "slug": "/docs/recipes/semantic-search-rag", "kicker": "RECIPE", "snip": "Retrieval by meaning with semantic_text — auto-embed on ingest, no separate vector DB."},{"title": "Agent memory", "slug": "/docs/recipes/agentic-memory", "kicker": "RECIPE", "snip": "Long-term memory for agents via /_memory — store, recall, forget, per-agent isolation."},{"title": "Passage retrieval", "slug": "/docs/recipes/passage-retrieval", "kicker": "RECIPE", "snip": "Auto-chunk long docs; best-passage (max-sim) scoring — 98% top-3."},{"title": "Vector search · kNN", "slug": "/docs/recipes/vector-search-knn", "kicker": "RECIPE", "snip": "Nearest-neighbor over dense_vector — HNSW-served unfiltered, exact with filters."},{"title": "Vector quantization", "slug": "/docs/recipes/vector-quantization", "kicker": "RECIPE", "snip": "scalar8 (accepts ES int8_hnsw; exact code scan) — int8 scoring at recall@10 = 0.998. Precision, not memory."},{"title": "Hybrid search", "slug": "/docs/recipes/hybrid-search", "kicker": "RECIPE", "snip": "Keyword + vector in one query — RRF fusion."},{"title": "Log analytics", "slug": "/docs/recipes/log-analytics", "kicker": "RECIPE", "snip": "Raw logs to dashboards — error rates, p95 latency, top services."},{"title": "Anomaly detection", "slug": "/docs/recipes/anomaly-detection", "kicker": "RECIPE", "snip": "Statistical _ml detectors that flag spikes in metrics/logs."},{"title": "Anomaly datafeeds", "slug": "/docs/recipes/continuous-anomaly-datafeeds", "kicker": "RECIPE", "snip": "Live _ml datafeed re-scores an index on a timer, stores anomaly records."},{"title": "Migrate from ES", "slug": "/docs/recipes/migrate-from-elasticsearch", "kicker": "RECIPE", "snip": "Point your existing ES client at XERJ — same wire, change the URL."}]</script>

<script>
  // IIFE — js-auth navigates between decrypted pages in the same
  // JS realm, so const bindings must be lexically scoped.
  (function () {
    const root = document.documentElement;

    // Client-side docs search — filters the inline JSON index.
    const searchEl    = document.getElementById('docsSearch');
    const inputEl     = document.getElementById('docsSearchInput');
    const resultsEl   = document.getElementById('docsSearchResults');
    const indexJson   = document.getElementById('docs-index');
    if (searchEl && inputEl && resultsEl && indexJson) {
      let index = [];
      try { index = JSON.parse(indexJson.textContent || '[]'); } catch {}
      const render = (q) => {
        const query = q.trim().toLowerCase();
        if (!query) { searchEl.classList.remove('active'); resultsEl.innerHTML = ''; return; }
        const scored = [];
        for (const e of index) {
          const hay = (e.title + ' ' + (e.kicker || '') + ' ' + (e.snip || '')).toLowerCase();
          if (hay.includes(query)) {
            let score = 0;
            if (e.title.toLowerCase().includes(query)) score += 10;
            if ((e.kicker || '').toLowerCase() === query.toUpperCase()) score += 5;
            if ((e.snip || '').toLowerCase().includes(query)) score += 1;
            scored.push({ e, score });
          }
        }
        scored.sort((a, b) => b.score - a.score);
        const top = scored.slice(0, 20);
        if (top.length === 0) {
          resultsEl.innerHTML = '<div class="no-results">NO MATCHES — TRY A SHORTER QUERY</div>';
        } else {
          resultsEl.innerHTML = top.map(({e}) => {
            const kicker = e.kicker || '';
            const snip = e.snip ? '<div class="snip">' + escHtml(e.snip) + '</div>' : '';
            return '<a href="' + e.slug + '">' +
              '<span class="kicker">' + escHtml(kicker) + '</span>' +
              '<span class="title">' + escHtml(e.title) + '</span>' +
              snip + '</a>';
          }).join('');
        }
        searchEl.classList.add('active');
      };
      const escHtml = (s) => String(s).replace(/&/g, '&amp;').replace(/</g, '&lt;').replace(/>/g, '&gt;').replace(/"/g, '&quot;');
      inputEl.addEventListener('input', (e) => render(e.target.value));
      inputEl.addEventListener('focus', () => { if (inputEl.value) render(inputEl.value); });
      document.addEventListener('keydown', (e) => {
        if ((e.key === 'k' || e.key === 'K') && (e.metaKey || e.ctrlKey)) {
          e.preventDefault();
          inputEl.focus();
          inputEl.select();
        }
        if (e.key === 'Escape' && document.activeElement === inputEl) {
          inputEl.value = '';
          searchEl.classList.remove('active');
          resultsEl.innerHTML = '';
          inputEl.blur();
        }
      });
    }
  })();
</script>

</body>
</html>