Vector Similarity Search

<h2 id="vector-similarity-search-in-geode" class="position-relative d-flex align-items-center group"> Vector Similarity Search in Geode <button type="button" class="h-share btn btn-link p-0 text-decoration-none link-secondary opacity-50 hover-opacity-100 transition-all ms-1" data-share-target="vector-similarity-search-in-geode" aria-haspopup="dialog" aria-label="Share link: Vector Similarity Search in Geode"> Share link </button> </h2><div id="headingShareModal" class="heading-share-modal" role="dialog" aria-modal="true" aria-labelledby="headingShareTitle" hidden> <div class="hsm-dialog" role="document"> <div class="hsm-header"> <h2 id="headingShareTitle" class="h6 mb-0 fw-bold">Share this section</h2> <button type="button" class="hsm-close" aria-label="Close"> </button> </div> <div class="hsm-body"> <label for="headingShareInput" class="form-label small text-muted mb-1 text-uppercase fw-bold" style="font-size: 0.7rem; letter-spacing: 0.5px;">Permalink</label> <div class="input-group mb-4 hsm-url-group"> <input id="headingShareInput" type="text" class="form-control font-monospace" readonly aria-readonly="true" style="font-size: 0.85rem;" /> <button class="btn btn-primary hsm-copy" type="button" aria-label="Copy" title="Copy"> </button> </div> <div class="small fw-bold mb-2 text-muted text-uppercase" style="font-size: 0.7rem; letter-spacing: 0.5px;">Share via</div> <div class="hsm-share-grid"> <a id="share-twitter" class="btn btn-outline-secondary w-100" target="_blank" rel="noopener noreferrer"> Twitter </a> <a id="share-linkedin" class="btn btn-outline-secondary w-100" target="_blank" rel="noopener noreferrer"> LinkedIn </a> <a id="share-facebook" class="btn btn-outline-secondary w-100" target="_blank" rel="noopener noreferrer"> Facebook </a> </div> </div> </div> </div> <style> .heading-share-modal { position: fixed; inset: 0; display: flex; justify-content: center; align-items: center; background: rgba(0, 0, 0, 0.6); z-index: 1050; padding: 1rem; backdrop-filter: blur(4px); -webkit-backdrop-filter: blur(4px); } .heading-share-modal[hidden] { display: none !important; } .hsm-dialog { max-width: 420px; width: 100%; background: var(--bs-body-bg, #fff); color: var(--bs-body-color, #212529); border: 1px solid var(--bs-border-color, rgba(0,0,0,0.1)); border-radius: 1rem; box-shadow: 0 25px 50px -12px rgba(0, 0, 0, 0.25); overflow: hidden; animation: hsm-fade-in 0.2s ease-out; } @keyframes hsm-fade-in { from { opacity: 0; transform: scale(0.95); } to { opacity: 1; transform: scale(1); } } [data-bs-theme="dark"] .hsm-dialog { background: #1e293b; border-color: rgba(255,255,255,0.1); color: #f8f9fa; } .hsm-header { display: flex; justify-content: space-between; align-items: center; padding: 1rem 1.5rem; border-bottom: 1px solid var(--bs-border-color, rgba(0,0,0,0.1)); background: rgba(0,0,0,0.02); } [data-bs-theme="dark"] .hsm-header { background: rgba(255,255,255,0.02); border-color: rgba(255,255,255,0.1); } .hsm-close { background: transparent; border: none; color: inherit; opacity: 0.5; padding: 0.25rem 0.5rem; border-radius: 0.25rem; font-size: 1.2rem; line-height: 1; transition: opacity 0.2s; } .hsm-close:hover { opacity: 1; } .hsm-body { padding: 1.5rem; } .hsm-url-group { display: flex !important; align-items: stretch; } .hsm-url-group .form-control { flex: 1; min-width: 0; margin: 0; background: var(--bs-secondary-bg, #f8f9fa); border-color: var(--bs-border-color, #dee2e6); border-top-right-radius: 0; border-bottom-right-radius: 0; height: 42px; } .hsm-url-group .btn { flex: 0 0 auto; margin: 0; margin-left: -1px; border-top-left-radius: 0; border-bottom-left-radius: 0; height: 42px; display: flex; align-items: center; justify-content: center; padding: 0 1.25rem; z-index: 2; } [data-bs-theme="dark"] .hsm-url-group .form-control { background: #0f172a; border-color: #334155; color: #e2e8f0; } .hsm-share-grid { display: flex; flex-direction: column; gap: 0.5rem; } .hsm-share-grid .btn { display: flex; align-items: center; justify-content: center; font-size: 0.9rem; padding: 0.6rem; border-color: var(--bs-border-color); width: 100%; } [data-bs-theme="dark"] .hsm-share-grid .btn { color: #e2e8f0; border-color: #475569; } [data-bs-theme="dark"] .hsm-share-grid .btn:hover { background: #334155; border-color: #cbd5e1; } </style> <script> (function(){ const modal = document.getElementById('headingShareModal'); if(!modal) return; const input = modal.querySelector('#headingShareInput'); const copyBtn = modal.querySelector('.hsm-copy'); const twitter = modal.querySelector('#share-twitter'); const linkedin = modal.querySelector('#share-linkedin'); const facebook = modal.querySelector('#share-facebook'); const closeBtn = modal.querySelector('.hsm-close'); let lastFocus=null; let trapBound=false; function buildUrl(id){ return window.location.origin + window.location.pathname + '#' + id; } function isOpen(){ return !modal.hasAttribute('hidden'); } function hydrate(id){ const url=buildUrl(id); input.value=url; const enc=encodeURIComponent(url); const text=encodeURIComponent(document.title); if(twitter) twitter.href=`https://twitter.com/intent/tweet?url=${enc}&text=${text}`; if(linkedin) linkedin.href=`https://www.linkedin.com/sharing/share-offsite/?url=${enc}`; if(facebook) facebook.href=`https://www.facebook.com/sharer/sharer.php?u=${enc}`; } function openModal(id){ lastFocus=document.activeElement; hydrate(id); if(!isOpen()){ modal.removeAttribute('hidden'); } requestAnimationFrame(()=>{ input.focus(); }); trapFocus(); } function closeModal(){ if(!isOpen()) return; modal.setAttribute('hidden',''); if(lastFocus && typeof lastFocus.focus==='function') lastFocus.focus(); } function copyCurrent(){ try{ navigator.clipboard.writeText(input.value).then(()=>feedback(true),()=>fallback()); } catch(e){ fallback(); } } function fallback(){ input.select(); try{ document.execCommand('copy'); feedback(true);}catch(e){ feedback(false);} } function feedback(ok){ if(!copyBtn) return; const icon=copyBtn.querySelector('i'); if(!icon) return; const prev=copyBtn.getAttribute('data-prev')||icon.className; if(!copyBtn.getAttribute('data-prev')) copyBtn.setAttribute('data-prev',prev); icon.className= ok ? 'fa-duotone fa-clipboard-check':'fa-duotone fa-circle-exclamation'; setTimeout(()=>{ icon.className=prev; },1800); } function handleShareClick(e){ e.preventDefault(); const btn=e.currentTarget; const id=btn.getAttribute('data-share-target'); if(id) openModal(id); } function bindShareButtons(){ document.querySelectorAll('.h-share').forEach(btn=>{ if(!btn.dataset.hShareBound){ btn.addEventListener('click', handleShareClick); btn.dataset.hShareBound='1'; } }); } bindShareButtons(); if(document.readyState==='loading'){ document.addEventListener('DOMContentLoaded', bindShareButtons); } else { requestAnimationFrame(bindShareButtons); } document.addEventListener('click', function(e){ const shareBtn=e.target.closest && e.target.closest('.h-share'); if(shareBtn && !shareBtn.dataset.hShareBound){ handleShareClick.call(shareBtn, e); } }, true); document.addEventListener('click', e=>{ if(e.target===modal) closeModal(); if(e.target.closest && e.target.closest('.hsm-close')){ e.preventDefault(); closeModal(); } if(copyBtn && (e.target===copyBtn || (e.target.closest && e.target.closest('.hsm-copy')))) { e.preventDefault(); copyCurrent(); } }); document.addEventListener('keydown', e=>{ if(e.key==='Escape' && isOpen()) closeModal(); }); function trapFocus(){ if(trapBound) return; trapBound=true; modal.addEventListener('keydown', f=>{ if(f.key==='Tab' && isOpen()){ const focusable=[...modal.querySelectorAll('a[href],button,input,textarea,select,[tabindex]:not([tabindex="-1"])')].filter(el=>!el.hasAttribute('disabled')); if(!focusable.length) return; const first=focusable[0]; const last=focusable[focusable.length-1]; if(f.shiftKey && document.activeElement===first){ f.preventDefault(); last.focus(); } else if(!f.shiftKey && document.activeElement===last){ f.preventDefault(); first.focus(); } } }); } if(closeBtn) closeBtn.addEventListener('click', e=>{ e.preventDefault(); closeModal(); }); })(); </script>Vector similarity search is a powerful feature in Geode that enables efficient nearest-neighbor queries over high-dimensional vector embeddings stored directly in graph properties. This capability is essential for modern machine learning applications including semantic search, recommendation systems, image similarity, and retrieval-augmented generation (RAG) workloads. <h3 id="introduction-to-vector-search" class="position-relative d-flex align-items-center group"> Introduction to Vector Search <button type="button" class="h-share btn btn-link p-0 text-decoration-none link-secondary opacity-50 hover-opacity-100 transition-all ms-1" data-share-target="introduction-to-vector-search" aria-haspopup="dialog" aria-label="Share link: Introduction to Vector Search"> Share link </button> </h3>Vector search addresses the challenge of finding similar items in high-dimensional space. Instead of exact matching, vector search uses distance metrics (cosine similarity, Euclidean distance, dot product) to find the k-nearest neighbors to a query vector. This technology powers applications like: <ul> <li>Semantic Search: Finding documents or content with similar meaning, not just matching keywords</li> <li>Recommendation Engines: Identifying items similar to user preferences</li> <li>Image and Video Search: Finding visually similar media by comparing embedding vectors</li> <li>Anomaly Detection: Identifying outliers by measuring distance from normal patterns</li> <li>Question Answering: Retrieving relevant context for large language models (LLMs)</li> </ul> Traditional exact nearest-neighbor search has O(n) complexity, making it impractical for large datasets. Geode uses Hierarchical Navigable Small World (HNSW) graphs to achieve approximate nearest-neighbor (ANN) search with logarithmic complexity. <h3 id="geodes-vector-search-implementation" class="position-relative d-flex align-items-center group"> Geode&rsquo;s Vector Search Implementation <button type="button" class="h-share btn btn-link p-0 text-decoration-none link-secondary opacity-50 hover-opacity-100 transition-all ms-1" data-share-target="geodes-vector-search-implementation" aria-haspopup="dialog" aria-label="Share link: Geodes Vector Search Implementation"> Share link </button> </h3>Geode implements vector search as native graph capabilities through several components: <h4 id="hnsw-index-integration" class="position-relative d-flex align-items-center group"> HNSW Index Integration <button type="button" class="h-share btn btn-link p-0 text-decoration-none link-secondary opacity-50 hover-opacity-100 transition-all ms-1" data-share-target="hnsw-index-integration" aria-haspopup="dialog" aria-label="Share link: HNSW Index Integration"> Share link </button> </h4>HNSW indexes are stored alongside graph data, allowing seamless integration of vector search with graph traversals. Properties containing vector data can be indexed using: <div class="highlight"><pre tabindex="0" class="chroma"><code class="language-gql" data-lang="gql">CREATE VECTOR INDEX product_embeddings ON Product(embedding) WITH ( metric = 'cosine', dimensions = 768, ef_construction = 200, m = 16 ); </code></pre></div>Parameters explained: <ul> <li><code>metric</code>: Distance function (cosine, euclidean, dot_product)</li> <li><code>dimensions</code>: Vector dimensionality (must match your embeddings)</li> <li><code>ef_construction</code>: Build-time accuracy parameter (higher = more accurate, slower build)</li> <li><code>m</code>: Maximum connections per node (higher = better recall, more memory)</li> </ul> <h4 id="native-gql-vector-functions" class="position-relative d-flex align-items-center group"> Native GQL Vector Functions <button type="button" class="h-share btn btn-link p-0 text-decoration-none link-secondary opacity-50 hover-opacity-100 transition-all ms-1" data-share-target="native-gql-vector-functions" aria-haspopup="dialog" aria-label="Share link: Native GQL Vector Functions"> Share link </button> </h4>Geode extends GQL with vector search functions that integrate naturally with pattern matching: <div class="highlight"><pre tabindex="0" class="chroma"><code class="language-gql" data-lang="gql">MATCH (p:Product) WHERE vector_similarity(p.embedding, $query_vector, 'cosine') > 0.8 RETURN p.name, p.description ORDER BY vector_similarity(p.embedding, $query_vector, 'cosine') DESC LIMIT 10; </code></pre></div> <h4 id="hybrid-search-combining-graph-and-vector-queries" class="position-relative d-flex align-items-center group"> Hybrid Search: Combining Graph and Vector Queries <button type="button" class="h-share btn btn-link p-0 text-decoration-none link-secondary opacity-50 hover-opacity-100 transition-all ms-1" data-share-target="hybrid-search-combining-graph-and-vector-queries" aria-haspopup="dialog" aria-label="Share link: Hybrid Search: Combining Graph and Vector Queries"> Share link </button> </h4>Geode’s unique advantage is combining graph topology with vector similarity: <div class="highlight"><pre tabindex="0" class="chroma"><code class="language-gql" data-lang="gql">-- Find similar products in the same category MATCH (category:Category {name: 'Electronics'})-[:CONTAINS]->(p:Product) WITH p, vector_similarity(p.embedding, $query_vector, 'cosine') AS similarity WHERE similarity > 0.75 RETURN p.name, similarity ORDER BY similarity DESC LIMIT 5; </code></pre></div><div class="highlight"><pre tabindex="0" class="chroma"><code class="language-gql" data-lang="gql">-- Collaborative filtering with vector search MATCH (user:User {id: $user_id})-[:PURCHASED]->(past:Product) WITH collect(past.embedding) AS user_history MATCH (candidate:Product) WHERE NOT (user)-[:PURCHASED]->(candidate) WITH candidate, avg([emb IN user_history | vector_similarity(candidate.embedding, emb, 'cosine')]) AS avg_similarity WHERE avg_similarity > 0.7 RETURN candidate.name, avg_similarity ORDER BY avg_similarity DESC LIMIT 10; </code></pre></div> <h3 id="use-cases-and-code-examples" class="position-relative d-flex align-items-center group"> Use Cases and Code Examples <button type="button" class="h-share btn btn-link p-0 text-decoration-none link-secondary opacity-50 hover-opacity-100 transition-all ms-1" data-share-target="use-cases-and-code-examples" aria-haspopup="dialog" aria-label="Share link: Use Cases and Code Examples"> Share link </button> </h3> <h4 id="use-case-1-semantic-document-search" class="position-relative d-flex align-items-center group"> Use Case 1: Semantic Document Search <button type="button" class="h-share btn btn-link p-0 text-decoration-none link-secondary opacity-50 hover-opacity-100 transition-all ms-1" data-share-target="use-case-1-semantic-document-search" aria-haspopup="dialog" aria-label="Share link: Use Case 1: Semantic Document Search"> Share link </button> </h4>Store document embeddings generated from sentence transformers or OpenAI models: <div class="highlight"><pre tabindex="0" class="chroma"><code class="language-python" data-lang="python">from geode_client import Client import asyncio async def create_document_index(): client = Client(host="localhost", port=3141) async with client.connection() as conn: # Create schema with vector index await conn.execute(""" CREATE VECTOR INDEX doc_embeddings ON Document(embedding) WITH (metric = 'cosine', dimensions = 384, m = 16); """) # Insert documents with embeddings await conn.execute(""" CREATE (d:Document { title: 'Introduction to Graph Databases', content: 'Graph databases model data as nodes and relationships...', embedding: $embedding }) """, {"embedding": generate_embedding("Graph databases model...")}) async def semantic_search(query_text): client = Client(host="localhost", port=3141) async with client.connection() as conn: query_embedding = generate_embedding(query_text) result, _ = await conn.query(""" MATCH (d:Document) WITH d, vector_similarity(d.embedding, $query_emb, 'cosine') AS score WHERE score > 0.6 RETURN d.title, d.content, score ORDER BY score DESC LIMIT 5 """, {"query_emb": query_embedding}) for row in result.rows: print(f"{row['score']:.3f} - {row['title']}") </code></pre></div> <h4 id="use-case-2-product-recommendations-with-knowledge-graph" class="position-relative d-flex align-items-center group"> Use Case 2: Product Recommendations with Knowledge Graph <button type="button" class="h-share btn btn-link p-0 text-decoration-none link-secondary opacity-50 hover-opacity-100 transition-all ms-1" data-share-target="use-case-2-product-recommendations-with-knowledge-graph" aria-haspopup="dialog" aria-label="Share link: Use Case 2: Product Recommendations with Knowledge Graph"> Share link </button> </h4>Combine product similarity with graph relationships: <div class="highlight"><pre tabindex="0" class="chroma"><code class="language-gql" data-lang="gql">-- Find products similar to items in cart, considering brand preferences MATCH (user:User {id: $user_id})-[:PREFERS]->(brand:Brand) MATCH (brand)-[:MANUFACTURES]->(product:Product) MATCH (cart_item:Product {id: $cart_item_id}) WITH product, cart_item, vector_similarity(product.embedding, cart_item.embedding, 'cosine') AS similarity WHERE similarity > 0.7 AND product.id <> cart_item.id RETURN product.name, product.price, similarity ORDER BY similarity DESC LIMIT 5; </code></pre></div> <h4 id="use-case-3-image-similarity-search" class="position-relative d-flex align-items-center group"> Use Case 3: Image Similarity Search <button type="button" class="h-share btn btn-link p-0 text-decoration-none link-secondary opacity-50 hover-opacity-100 transition-all ms-1" data-share-target="use-case-3-image-similarity-search" aria-haspopup="dialog" aria-label="Share link: Use Case 3: Image Similarity Search"> Share link </button> </h4>Use image embeddings from models like CLIP or ResNet: <div class="highlight"><pre tabindex="0" class="chroma"><code class="language-python" data-lang="python">async def find_similar_images(image_path, limit=10): embedding = image_encoder.encode(image_path) # Generate embedding client = Client(host="localhost", port=3141) async with client.connection() as conn: result, _ = await conn.query(""" MATCH (img:Image) WITH img, vector_similarity(img.embedding, $query_emb, 'euclidean') AS distance WHERE distance < 0.5 RETURN img.url, img.tags, distance ORDER BY distance ASC LIMIT $limit """, {"query_emb": embedding, "limit": limit}) return result.bindings </code></pre></div> <h3 id="best-practices" class="position-relative d-flex align-items-center group"> Best Practices <button type="button" class="h-share btn btn-link p-0 text-decoration-none link-secondary opacity-50 hover-opacity-100 transition-all ms-1" data-share-target="best-practices" aria-haspopup="dialog" aria-label="Share link: Best Practices"> Share link </button> </h3> <h4 id="choosing-index-parameters" class="position-relative d-flex align-items-center group"> Choosing Index Parameters <button type="button" class="h-share btn btn-link p-0 text-decoration-none link-secondary opacity-50 hover-opacity-100 transition-all ms-1" data-share-target="choosing-index-parameters" aria-haspopup="dialog" aria-label="Share link: Choosing Index Parameters"> Share link </button> </h4>Dimensions: Match your embedding model exactly: <ul> <li>Sentence transformers: 384, 768, 1024</li> <li>OpenAI ada-002: 1536</li> <li>CLIP: 512 or 768</li> <li>Custom models: verify output shape</li> </ul> Metric selection: <ul> <li>Cosine: Best for normalized embeddings (most common)</li> <li>Euclidean: When magnitude matters</li> <li>Dot product: For sparse vectors or specific models</li> </ul> HNSW tuning: <ul> <li><code>m = 16</code> (default): Good balance for most cases</li> <li><code>m = 32</code>: Higher recall, 2x memory usage</li> <li><code>ef_construction = 200</code>: Production default</li> <li><code>ef_construction = 400</code>: Higher quality index, slower build</li> </ul> <h4 id="embedding-generation" class="position-relative d-flex align-items-center group"> Embedding Generation <button type="button" class="h-share btn btn-link p-0 text-decoration-none link-secondary opacity-50 hover-opacity-100 transition-all ms-1" data-share-target="embedding-generation" aria-haspopup="dialog" aria-label="Share link: Embedding Generation"> Share link </button> </h4>Consistency is critical: <div class="highlight"><pre tabindex="0" class="chroma"><code class="language-python" data-lang="python"># WRONG: Different models or preprocessing doc_embedding = model_v1.encode(text) query_embedding = model_v2.encode(query) # Won't match! # RIGHT: Same model and preprocessing def generate_embedding(text): normalized = text.lower().strip() return sentence_transformer.encode(normalized) </code></pre></div>Batch processing for efficiency: <div class="highlight"><pre tabindex="0" class="chroma"><code class="language-python" data-lang="python">async def index_documents_batch(documents, batch_size=100): client = Client(host="localhost", port=3141) async with client.connection() as conn: for i in range(0, len(documents), batch_size): batch = documents[i:i + batch_size] embeddings = model.encode([d.text for d in batch]) for doc, emb in zip(batch, embeddings): await conn.execute(""" CREATE (d:Document { id: $id, text: $text, embedding: $emb }) """, {"id": doc.id, "text": doc.text, "emb": emb.tolist()}) </code></pre></div> <h4 id="query-optimization" class="position-relative d-flex align-items-center group"> Query Optimization <button type="button" class="h-share btn btn-link p-0 text-decoration-none link-secondary opacity-50 hover-opacity-100 transition-all ms-1" data-share-target="query-optimization" aria-haspopup="dialog" aria-label="Share link: Query Optimization"> Share link </button> </h4>Use appropriate similarity thresholds: <div class="highlight"><pre tabindex="0" class="chroma"><code class="language-gql" data-lang="gql">-- Too restrictive: May return no results WHERE vector_similarity(n.emb, $query, 'cosine') > 0.95 -- Too permissive: Returns irrelevant results WHERE vector_similarity(n.emb, $query, 'cosine') > 0.3 -- Just right: Adjust based on your data WHERE vector_similarity(n.emb, $query, 'cosine') > 0.7 </code></pre></div>Limit result sets: <div class="highlight"><pre tabindex="0" class="chroma"><code class="language-gql" data-lang="gql">-- HNSW is optimized for top-k queries MATCH (d:Document) WITH d, vector_similarity(d.embedding, $query, 'cosine') AS score ORDER BY score DESC LIMIT 20 -- HNSW explores only as needed </code></pre></div> <h3 id="performance-considerations" class="position-relative d-flex align-items-center group"> Performance Considerations <button type="button" class="h-share btn btn-link p-0 text-decoration-none link-secondary opacity-50 hover-opacity-100 transition-all ms-1" data-share-target="performance-considerations" aria-haspopup="dialog" aria-label="Share link: Performance Considerations"> Share link </button> </h3> <h4 id="indexing-performance" class="position-relative d-flex align-items-center group"> Indexing Performance <button type="button" class="h-share btn btn-link p-0 text-decoration-none link-secondary opacity-50 hover-opacity-100 transition-all ms-1" data-share-target="indexing-performance" aria-haspopup="dialog" aria-label="Share link: Indexing Performance"> Share link </button> </h4>Build time scales with dataset size: <ul> <li>100K vectors: ~1-2 minutes</li> <li>1M vectors: ~15-30 minutes</li> <li>10M vectors: ~3-5 hours</li> </ul> Memory requirements: <ul> <li>Base: <code>num_vectors * dimensions * 4 bytes</code> (float32)</li> <li>HNSW overhead: <code>num_vectors * m * 16 * 4 bytes</code></li> <li>Example: 1M vectors × 768D × 16M = ~50GB RAM</li> </ul> Incremental indexing: <div class="highlight"><pre tabindex="0" class="chroma"><code class="language-gql" data-lang="gql">-- Create index first CREATE VECTOR INDEX CONCURRENTLY product_embeddings ON Product(embedding) WITH (metric = 'cosine', dimensions = 768); -- Insert nodes normally; index updates incrementally CREATE (p:Product {name: 'New Item', embedding: $emb}); </code></pre></div> <h4 id="query-performance" class="position-relative d-flex align-items-center group"> Query Performance <button type="button" class="h-share btn btn-link p-0 text-decoration-none link-secondary opacity-50 hover-opacity-100 transition-all ms-1" data-share-target="query-performance" aria-haspopup="dialog" aria-label="Share link: Query Performance"> Share link </button> </h4>Typical latency (10k vectors, 10-NN): <ul> <li>Single vector search: 1-5ms at ~90% recall</li> <li>Combined graph + vector: workload-dependent (varies by traversal and filters)</li> <li>Batch queries: throughput depends on workload and hardware</li> </ul> Tuning runtime accuracy (not yet exposed, coming soon): <div class="highlight"><pre tabindex="0" class="chroma"><code class="language-gql" data-lang="gql">-- Higher ef_search = more accurate, slower SET vector_search_ef = 100; -- Default: 50 MATCH (d:Document) WITH d, vector_similarity(d.embedding, $query, 'cosine') AS score ORDER BY score DESC LIMIT 10; </code></pre></div> <h4 id="scaling-vector-search" class="position-relative d-flex align-items-center group"> Scaling Vector Search <button type="button" class="h-share btn btn-link p-0 text-decoration-none link-secondary opacity-50 hover-opacity-100 transition-all ms-1" data-share-target="scaling-vector-search" aria-haspopup="dialog" aria-label="Share link: Scaling Vector Search"> Share link </button> </h4>Horizontal scaling: <ul> <li>Partition large datasets by category or domain</li> <li>Use graph structure to route queries to relevant partitions</li> <li>Combine results from distributed searches</li> </ul> Caching strategies: <div class="highlight"><pre tabindex="0" class="chroma"><code class="language-python" data-lang="python"># Cache frequently queried embeddings embedding_cache = {} async def cached_search(query_text): cache_key = hash(query_text) if cache_key not in embedding_cache: embedding_cache[cache_key] = generate_embedding(query_text) return await search_by_vector(embedding_cache[cache_key]) </code></pre></div> <h3 id="troubleshooting" class="position-relative d-flex align-items-center group"> Troubleshooting <button type="button" class="h-share btn btn-link p-0 text-decoration-none link-secondary opacity-50 hover-opacity-100 transition-all ms-1" data-share-target="troubleshooting" aria-haspopup="dialog" aria-label="Share link: Troubleshooting"> Share link </button> </h3> <h4 id="poor-search-quality" class="position-relative d-flex align-items-center group"> Poor Search Quality <button type="button" class="h-share btn btn-link p-0 text-decoration-none link-secondary opacity-50 hover-opacity-100 transition-all ms-1" data-share-target="poor-search-quality" aria-haspopup="dialog" aria-label="Share link: Poor Search Quality"> Share link </button> </h4>Problem: Results aren’t relevant Solutions: <ol> <li>Verify embedding model consistency</li> <li>Check vector normalization (cosine requires normalized vectors)</li> <li>Adjust similarity threshold</li> <li>Retrain or upgrade embedding model</li> </ol> Problem: Slow query performance Solutions: <ol> <li>Increase <code>m</code> parameter (rebuild index)</li> <li>Add filters before vector search to reduce candidate set</li> <li>Use EXPLAIN to identify bottlenecks</li> <li>Consider partitioning large datasets</li> </ol> Problem: High memory usage Solutions: <ol> <li>Reduce <code>m</code> parameter (less accuracy, less memory)</li> <li>Use lower-dimensional embeddings if possible</li> <li>Partition data across multiple nodes</li> <li>Use dimensionality reduction (PCA, UMAP)</li> </ol> <h4 id="index-maintenance" class="position-relative d-flex align-items-center group"> Index Maintenance <button type="button" class="h-share btn btn-link p-0 text-decoration-none link-secondary opacity-50 hover-opacity-100 transition-all ms-1" data-share-target="index-maintenance" aria-haspopup="dialog" aria-label="Share link: Index Maintenance"> Share link </button> </h4>Monitoring index health: <div class="highlight"><pre tabindex="0" class="chroma"><code class="language-gql" data-lang="gql">SHOW INDEXES WHERE name = 'product_embeddings'; -- Returns: size, num_vectors, build_status </code></pre></div>Rebuilding indexes: <div class="highlight"><pre tabindex="0" class="chroma"><code class="language-gql" data-lang="gql">-- If index becomes corrupted or parameters need changing DROP INDEX product_embeddings; CREATE VECTOR INDEX product_embeddings ON Product(embedding) WITH (metric = 'cosine', dimensions = 768, m = 32); </code></pre></div> <h3 id="related-topics" class="position-relative d-flex align-items-center group"> Related Topics <button type="button" class="h-share btn btn-link p-0 text-decoration-none link-secondary opacity-50 hover-opacity-100 transition-all ms-1" data-share-target="related-topics" aria-haspopup="dialog" aria-label="Share link: Related Topics"> Share link </button> </h3><ul> <li><a href="/tags/hnsw/" >HNSW</a> : Deep dive into Hierarchical Navigable Small World algorithm</li> <li><a href="/tags/machine-learning/" >Machine Learning</a> : ML integration patterns with Geode</li> <li><a href="/tags/embeddings/" >Embeddings</a> : Best practices for generating and storing embeddings</li> <li><a href="/tags/performance/" >Performance</a> : General performance optimization techniques</li> <li><a href="/tags/indexing/" >Indexing</a> : Overview of all index types in Geode</li> <li><a href="/tags/recommendations/" >Recommendations</a> : Building recommendation systems</li> </ul> <h3 id="further-reading" class="position-relative d-flex align-items-center group"> Further Reading <button type="button" class="h-share btn btn-link p-0 text-decoration-none link-secondary opacity-50 hover-opacity-100 transition-all ms-1" data-share-target="further-reading" aria-haspopup="dialog" aria-label="Share link: Further Reading"> Share link </button> </h3><ul> <li>HNSW Paper: “Efficient and robust approximate nearest neighbor search using Hierarchical Navigable Small World graphs” (Malkov & Yashunin, 2018)</li> <li>Sentence Transformers: <a href="https://www.sbert.net/" aria-label="https://www.sbert.net/ – opens in new window" target="_blank" rel="noopener noreferrer" >https://www.sbert.net/ ↗ </a> - Popular embedding models</li> <li>OpenAI Embeddings: <a href="https://platform.openai.com/docs/guides/embeddings" aria-label="https://platform.openai.com/docs/guides/embeddings – opens in new window" target="_blank" rel="noopener noreferrer" >https://platform.openai.com/docs/guides/embeddings ↗ </a> </li> <li>Geode Vector Search Guide: <code>/docs/advanced-features/vector-search/</code></li> <li>Performance Tuning: <code>/docs/performance/vector-optimization/</code></li> </ul> <h3 id="advanced-vector-search-techniques" class="position-relative d-flex align-items-center group"> Advanced Vector Search Techniques <button type="button" class="h-share btn btn-link p-0 text-decoration-none link-secondary opacity-50 hover-opacity-100 transition-all ms-1" data-share-target="advanced-vector-search-techniques" aria-haspopup="dialog" aria-label="Share link: Advanced Vector Search Techniques"> Share link </button> </h3> <h4 id="hybrid-dense-sparse-search" class="position-relative d-flex align-items-center group"> Hybrid Dense-Sparse Search <button type="button" class="h-share btn btn-link p-0 text-decoration-none link-secondary opacity-50 hover-opacity-100 transition-all ms-1" data-share-target="hybrid-dense-sparse-search" aria-haspopup="dialog" aria-label="Share link: Hybrid Dense-Sparse Search"> Share link </button> </h4>Combine vector similarity with keyword matching: <div class="highlight"><pre tabindex="0" class="chroma"><code class="language-gql" data-lang="gql">-- Hybrid search: HNSW + BM25 MATCH (d:Document) WHERE text_search(d.content, $keyword_query) AND vector_similarity(d.embedding, $vector_query, 'cosine') > 0.6 WITH d, text_score(d, $keyword_query) AS bm25_score, vector_similarity(d.embedding, $vector_query) AS vector_score RETURN d.doc_id, d.title, bm25_score, vector_score, 0.5 * bm25_score + 0.5 * vector_score AS hybrid_score ORDER BY hybrid_score DESC LIMIT 20; </code></pre></div> <h4 id="multi-vector-search" class="position-relative d-flex align-items-center group"> Multi-Vector Search <button type="button" class="h-share btn btn-link p-0 text-decoration-none link-secondary opacity-50 hover-opacity-100 transition-all ms-1" data-share-target="multi-vector-search" aria-haspopup="dialog" aria-label="Share link: Multi-Vector Search"> Share link </button> </h4>Search across multiple embedding spaces: <div class="highlight"><pre tabindex="0" class="chroma"><code class="language-gql" data-lang="gql">-- Search using both content and title embeddings MATCH (d:Document) WITH d, vector_similarity(d.content_embedding, $content_query_emb, 'cosine') AS content_sim, vector_similarity(d.title_embedding, $title_query_emb, 'cosine') AS title_sim WITH d, 0.7 * content_sim + 0.3 * title_sim AS combined_similarity WHERE combined_similarity > 0.75 RETURN d.doc_id, d.title, combined_similarity ORDER BY combined_similarity DESC; </code></pre></div> <h3 id="query-time-optimizations" class="position-relative d-flex align-items-center group"> Query-Time Optimizations <button type="button" class="h-share btn btn-link p-0 text-decoration-none link-secondary opacity-50 hover-opacity-100 transition-all ms-1" data-share-target="query-time-optimizations" aria-haspopup="dialog" aria-label="Share link: Query-Time Optimizations"> Share link </button> </h3> <h4 id="pre-filtering-vs-post-filtering" class="position-relative d-flex align-items-center group"> Pre-Filtering vs Post-Filtering <button type="button" class="h-share btn btn-link p-0 text-decoration-none link-secondary opacity-50 hover-opacity-100 transition-all ms-1" data-share-target="pre-filtering-vs-post-filtering" aria-haspopup="dialog" aria-label="Share link: Pre-Filtering vs Post-Filtering"> Share link </button> </h4><div class="highlight"><pre tabindex="0" class="chroma"><code class="language-gql" data-lang="gql">-- Efficient: Pre-filter then vector search MATCH (d:Document) WHERE d.category = 'technical' AND d.publish_date > date('2024-01-01') AND d.language = 'en' WITH d WHERE vector_similarity(d.embedding, $query, 'cosine') > 0.7 RETURN d ORDER BY vector_similarity(d.embedding, $query) DESC LIMIT 10; -- Less efficient: Vector search then filter CALL vector.search({index: 'docs', query: $query, k: 1000}) YIELD node WHERE node.category = 'technical' // Post-filter loses HNSW efficiency RETURN node LIMIT 10; </code></pre></div> <h4 id="cascaded-search" class="position-relative d-flex align-items-center group"> Cascaded Search <button type="button" class="h-share btn btn-link p-0 text-decoration-none link-secondary opacity-50 hover-opacity-100 transition-all ms-1" data-share-target="cascaded-search" aria-haspopup="dialog" aria-label="Share link: Cascaded Search"> Share link </button> </h4>Fast approximate search followed by reranking: <div class="highlight"><pre tabindex="0" class="chroma"><code class="language-gql" data-lang="gql">-- Stage 1: Fast approximate retrieval (top 100) CALL vector.search({ index: 'products', query: $query_embedding, k: 100, ef: 50 // Lower ef for speed }) YIELD node AS candidate, similarity AS approx_score -- Stage 2: Precise reranking (top 20) WITH candidate, vector.similarity(candidate.high_quality_embedding, $query_embedding, 'cosine') AS precise_score ORDER BY precise_score DESC LIMIT 20 RETURN candidate, precise_score; </code></pre></div> <h3 id="approximate-nearest-neighbors-ann-tuning" class="position-relative d-flex align-items-center group"> Approximate Nearest Neighbors (ANN) Tuning <button type="button" class="h-share btn btn-link p-0 text-decoration-none link-secondary opacity-50 hover-opacity-100 transition-all ms-1" data-share-target="approximate-nearest-neighbors-ann-tuning" aria-haspopup="dialog" aria-label="Share link: Approximate Nearest Neighbors (ANN) Tuning"> Share link </button> </h3> <h4 id="hnsw-parameter-impact" class="position-relative d-flex align-items-center group"> HNSW Parameter Impact <button type="button" class="h-share btn btn-link p-0 text-decoration-none link-secondary opacity-50 hover-opacity-100 transition-all ms-1" data-share-target="hnsw-parameter-impact" aria-haspopup="dialog" aria-label="Share link: HNSW Parameter Impact"> Share link </button> </h4>M (connections per layer): <ul> <li>M=4: ~10MB/million vectors, 85% recall</li> <li>M=16: ~40MB/million vectors, 95% recall</li> <li>M=32: ~80MB/million vectors, 98% recall</li> </ul> ef_construction: <ul> <li>ef_construction=100: Fast index build, 90% quality</li> <li>ef_construction=200: Balanced (recommended)</li> <li>ef_construction=400: Slow build, 98% quality</li> </ul> ef_search (query-time): <ul> <li>ef_search=16: <1ms latency, 85% recall</li> <li>ef_search=64: ~2ms latency, 95% recall</li> <li>ef_search=256: ~10ms latency, 99% recall</li> </ul> <h4 id="dynamic-ef_search-tuning" class="position-relative d-flex align-items-center group"> Dynamic ef_search Tuning <button type="button" class="h-share btn btn-link p-0 text-decoration-none link-secondary opacity-50 hover-opacity-100 transition-all ms-1" data-share-target="dynamic-ef_search-tuning" aria-haspopup="dialog" aria-label="Share link: Dynamic ef_search Tuning"> Share link </button> </h4><div class="highlight"><pre tabindex="0" class="chroma"><code class="language-gql" data-lang="gql">-- Adjust ef_search based on query importance CALL vector.search({ index: 'embeddings', query: $query, k: 10, ef: CASE WHEN $user_tier = 'premium' THEN 200 ELSE 50 END }) YIELD node, similarity RETURN node, similarity; </code></pre></div> <h3 id="quantization-and-compression" class="position-relative d-flex align-items-center group"> Quantization and Compression <button type="button" class="h-share btn btn-link p-0 text-decoration-none link-secondary opacity-50 hover-opacity-100 transition-all ms-1" data-share-target="quantization-and-compression" aria-haspopup="dialog" aria-label="Share link: Quantization and Compression"> Share link </button> </h3> <h4 id="scalar-quantization" class="position-relative d-flex align-items-center group"> Scalar Quantization <button type="button" class="h-share btn btn-link p-0 text-decoration-none link-secondary opacity-50 hover-opacity-100 transition-all ms-1" data-share-target="scalar-quantization" aria-haspopup="dialog" aria-label="Share link: Scalar Quantization"> Share link </button> </h4>Reduce memory by 4x with minimal accuracy loss: <div class="highlight"><pre tabindex="0" class="chroma"><code class="language-python" data-lang="python"># Quantize float32 to int8 def quantize_embeddings(embeddings): # Find min/max for normalization min_val, max_val = embeddings.min(), embeddings.max() # Scale to [0, 255] quantized = ((embeddings - min_val) / (max_val - min_val) * 255).astype(np.uint8) return quantized, min_val, max_val # Store quantized embeddings await client.execute(""" MATCH (d:Document {doc_id: $id}) SET d.embedding_quantized = $quantized, d.quantization_min = $min_val, d.quantization_max = $max_val """, {"id": doc_id, "quantized": quantized.tolist(), "min_val": min_val, "max_val": max_val}) </code></pre></div> <h4 id="product-quantization-pq" class="position-relative d-flex align-items-center group"> Product Quantization (PQ) <button type="button" class="h-share btn btn-link p-0 text-decoration-none link-secondary opacity-50 hover-opacity-100 transition-all ms-1" data-share-target="product-quantization-pq" aria-haspopup="dialog" aria-label="Share link: Product Quantization (PQ)"> Share link </button> </h4>Compress 1536d to ~96 bytes: <div class="highlight"><pre tabindex="0" class="chroma"><code class="language-python" data-lang="python"># Use Faiss for product quantization import faiss # Train PQ codec d = 1536 # Original dimension m = 48 # Number of subquantizers nbits = 8 # Bits per code pq = faiss.IndexPQ(d, m, nbits) pq.train(training_embeddings) # Encode embeddings codes = pq.sa_encode(embeddings) # Store compressed codes await client.execute(""" MATCH (d:Document {doc_id: $id}) SET d.embedding_pq = $codes """, {"id": doc_id, "codes": codes.tolist()}) </code></pre></div> <h3 id="further-reading-1" class="position-relative d-flex align-items-center group"> Further Reading <button type="button" class="h-share btn btn-link p-0 text-decoration-none link-secondary opacity-50 hover-opacity-100 transition-all ms-1" data-share-target="further-reading-1" aria-haspopup="dialog" aria-label="Share link: Further Reading"> Share link </button> </h3><ul> <li>Vector Search: HNSW, LSH, and IVF Algorithms</li> <li>Hybrid Search: Combining Dense and Sparse Retrieval</li> <li>Quantization: Scalar, Product, and Binary Quantization</li> <li>Performance: Benchmarking and Optimization Techniques</li> </ul> Browse tagged content for comprehensive vector search documentation.

Popular

Related Articles

Graph Algorithms and Analytics

Recommendation Systems