Skip to main content

Documentation Index

Fetch the complete documentation index at: https://docs.cyborg.co/llms.txt

Use this file to discover all available pages before exploring further.

Searches for nearest neighbors in the encrypted index using vector similarity search. Supports both single vector queries and batch queries with multiple vectors, as well as content-based search. 
async query({
    queryVectors?: number[] | number[][] | Float32Array,  // optional, required if `queryContents` not provided
    queryContents?: string,                 // optional, required if `queryVectors` not provided
    topK?: number,                          // optional, default: 100
    nProbes?: number,                       // optional, default: undefined (auto)
    filters?: FilterExpression,             // optional, default: {}
    include?: string[],                     // optional, default: [] (only `id` is returned)
    greedy?: boolean,                       // optional, default: false
    dimension?: number                      // optional, required when queryVectors is Float32Array
}): Promise<QueryResponse>

​
Parameters

ParameterTypeDefaultDescription
queryVectorsnumber[] | number[][] | Float32Array-Single vector [0.1, 0.2], batch vectors [[0.1, 0.2], [0.3, 0.4]], or typed array
queryContentsstring-(Optional) Text content to embed and search (alternative to queryVectors)
topKnumberundefined (server default: 100)(Optional) Number of nearest neighbors to return per query
nProbesnumberundefined (auto)(Optional) Number of cluster centers to search (higher = better recall). Auto-determined when undefined.
filtersFilterExpressionundefined(Optional) Metadata filters to apply to the search (no filtering when undefined)
includestring[]undefined (server default: [] — only id is returned)(Optional) Fields to include: "distance", "metadata", "vector", "contents"
greedybooleanfalse(Optional) Use faster approximate search
dimensionnumber-(Optional) Required when queryVectors is a Float32Array. Specifies vector dimensionality so the SDK can reshape the flat typed array.
Passing a Float32Array as queryVectors uses an optimized binary transfer format for better performance. When using Float32Array, the dimension parameter is required so the SDK can correctly reshape the flat array into individual query vectors.

​
Returns

Promise<QueryResponse>: A Promise that resolves to search results. The response format depends on the query type:
  • Single vector query: results is a flat array of QueryResultItem[]
  • Batch query: results is a nested array of QueryResultItem[][] (one array per input vector)

​
Exceptions

  • Throws if the API request fails due to network connectivity issues.
  • Throws if authentication fails (invalid API key).
  • Throws if the encryption key is invalid for the specified index.
  • Throws if there are internal server errors during the search.
  • Throws if neither queryVectors nor queryContents is provided.
  • Throws if vector dimensions don’t match the index configuration.
  • Throws if parameter values are out of valid ranges.
  • Throws if the include parameter contains invalid field names.

​
Example Usage

​
Basic Single Vector Query

import { Client } from 'cyborgdb';

const client = new Client({ baseUrl: 'http://localhost:8000', apiKey: 'your-api-key' });

// Load an existing index
const indexKey = new Uint8Array(Buffer.from('your-stored-hex-key', 'hex'));
const index = await client.loadIndex({ indexName: 'my-vector-index', indexKey });

// Search for similar vectors
const queryVector = [0.15, 0.25, 0.35, 0.45];

try {
    const results = await index.query({
        queryVectors: queryVector,
        topK: 5,
        include: ['distance', 'metadata']  // include defaults to [] (only `id`); request distance and metadata
    });
    
    // Single vector query returns flat array
    console.log(`Found ${results.results.length} similar vectors:`);
    
    results.results.forEach((item, i) => {
        console.log(`${i + 1}. ID: ${item.id}, Distance: ${item.distance}`);
        if (item.metadata) {
            console.log(`   Metadata: ${JSON.stringify(item.metadata)}`);
        }
    });
    
} catch (error: any) {
    console.error('Query failed:', error.message);
}

​
Content-Based Search (Text Search)

// Search by text content (requires embedding model configured on index)
try {
    const contentResults = await index.query({
        queryContents: "machine learning tutorial",  // content to embed and search
        topK: 10,                                    // return top 10 results
        include: ['distance', 'metadata']            // include defaults to [] (only `id`); request distance and metadata
    });
    
    console.log('Content-based search results:');
    contentResults.results.forEach((item, i) => {
        console.log(`${i + 1}. ${item.id}: ${item.metadata?.title}`);
        console.log(`   Distance: ${item.distance}`);
    });
    
} catch (error: any) {
    console.error('Content search failed:', error.message);
    // May fail if no embedding model is configured for the index
}

​
Advanced Single Query with Filters

const queryVector = [0.1, 0.2, 0.3, 0.4];

try {
    const results = await index.query({
        queryVectors: queryVector,
        topK: 10,                                       // return top 10 results
        nProbes: 5,                                     // probe 5 clusters for better accuracy
        filters: { category: 'research', published: true },  // only research documents
        include: ['distance', 'metadata', 'contents'],       // return all fields
        greedy: false                                        // use exact search
    });
    
    console.log('Advanced query results:');
    results.results.forEach((item, i) => {
        console.log(`${i + 1}. ${item.id}`);
        console.log(`   Distance: ${item.distance}`);
        console.log(`   Category: ${item.metadata?.category}`);
        console.log(`   Content Preview: ${item.contents?.toString().substring(0, 100)}...`);
    });
    
} catch (error: any) {
    console.error('Advanced query failed:', error.message);
}

​
Batch Vector Query

// Query multiple vectors at once
const queryVectors = [
    [0.1, 0.2, 0.3, 0.4],  // First query vector
    [0.5, 0.6, 0.7, 0.8],  // Second query vector
    [0.9, 1.0, 1.1, 1.2]   // Third query vector
];

try {
    const batchResults = await index.query({
        queryVectors,
        topK: 3,
        include: ['distance']  // include defaults to [] (only `id`); request distance
    });

    // Batch query returns nested arrays - one per input vector
    console.log(`Batch query completed for ${queryVectors.length} vectors:`);
    
    batchResults.results.forEach((queryResults, queryIndex) => {
        console.log(`\nResults for query vector ${queryIndex + 1}:`);
        queryResults.forEach((item, resultIndex) => {
            console.log(`  ${resultIndex + 1}. ${item.id} (distance: ${item.distance})`);
        });
    });
    
} catch (error: any) {
    console.error('Batch query failed:', error.message);
}

​
Complex Metadata Filtering

// Simple equality filter
const simpleFilter = { category: 'research' };

// Complex nested filter with operators
const complexFilter = {
    "$and": [
        { "author.name": "Dr. Smith" },
        { "metrics.score": { "$gte": 0.8 } },
        { "tags": { "$in": ["ai", "ml", "neural-networks"] } },
        { "published_date": { "$gte": "2024-01-01" } }
    ]
};

// Range filter example
const dateRangeFilter = {
    "created_date": {
        "$gte": "2024-01-01",
        "$lte": "2024-12-31"
    },
    "status": "published"
};

try {
    const filteredResults = await index.query({
        queryVectors: queryVector,
        topK: 15,
        nProbes: 3,
        filters: complexFilter,  // Use complex filter
        include: ['distance', 'metadata']
    });

    console.log(`Found ${filteredResults.results.length} filtered results`);
    
} catch (error: any) {
    console.error('Filtered query failed:', error.message);
}

​
Response Format

​
Single Query Response

{
    "results": [
        {
            "id": "doc1",
            "distance": 0.234,
            "metadata": { "title": "Document 1", "category": "research" },
        }
    ]
}

​
Batch Query Response

{
    "results": [
        [ // Results for first query vector
            { "id": "doc1", "distance": 0.234, "metadata": {...} },
            { "id": "doc2", "distance": 0.456, "metadata": {...} }
        ],
        [ // Results for second query vector
            { "id": "doc3", "distance": 0.123, "metadata": {...} },
            { "id": "doc4", "distance": 0.567, "metadata": {...} }
        ]
    ]
}