A powerful graph & vector data platform for AI applications across any environment

✨ Overview

Brainy combines the power of vector search with graph relationships in a lightweight, cross-platform database. Whether you're building AI applications, recommendation systems, or knowledge graphs, Brainy provides the tools you need to store, connect, and retrieve your data intelligently.

What makes Brainy special? It intelligently adapts to your environment! Brainy automatically detects your platform, adjusts its storage strategy, and optimizes performance based on your usage patterns. The more you use it, the smarter it gets - learning from your data to provide increasingly relevant results and connections.

🚀 Key Features

• Run Everywhere - Works in browsers, Node.js, serverless functions, and containers
• Vector Search - Find semantically similar content using embeddings
• Advanced JSON Document Search - Search within specific fields of JSON documents with field prioritization and service-based field standardization
• Graph Relationships - Connect data with meaningful relationships
• Streaming Pipeline - Process data in real-time as it flows through the system
• Extensible Augmentations - Customize and extend functionality with pluggable components
• Built-in Conduits - Sync and scale across instances with WebSocket and WebRTC
• TensorFlow Integration - Use TensorFlow.js for high-quality embeddings
• Adaptive Intelligence - Automatically optimizes for your environment and usage patterns
• Persistent Storage - Data persists across sessions and scales to any size
• TypeScript Support - Fully typed API with generics
• CLI Tools & Web Service - Command-line interface and REST API web service for data management
• Model Control Protocol (MCP) - Allow external AI models to access Brainy data and use augmentation pipeline as tools

🚀 Live Demo

Try the live demo - Check out the interactive demo on GitHub Pages that showcases Brainy's main features.

📊 What Can You Build?

• Semantic Search Engines - Find content based on meaning, not just keywords
• Recommendation Systems - Suggest similar items based on vector similarity
• Knowledge Graphs - Build connected data structures with relationships
• AI Applications - Store and retrieve embeddings for machine learning models
• AI-Enhanced Applications - Build applications that leverage vector embeddings for intelligent data processing
• Data Organization Tools - Automatically categorize and connect related information
• Adaptive Experiences - Create applications that learn and evolve with your users
• Model-Integrated Systems - Connect external AI models to Brainy data and tools using MCP

🔧 Installation

npm install @soulcraft/brainy

TensorFlow.js packages are included as bundled dependencies and will be automatically installed without any additional configuration.

Additional Packages

Brainy offers specialized packages for different use cases:

CLI Package

npm install -g @soulcraft/brainy-cli

Command-line interface for data management, bulk operations, and database administration.

Web Service Package

npm install @soulcraft/brainy-web-service

REST API web service wrapper that provides HTTP endpoints for search operations and database queries.

🏁 Quick Start

Brainy uses a unified build that automatically adapts to your environment (Node.js, browser, or serverless):

import { BrainyData, NounType, VerbType } from '@soulcraft/brainy'

// Create and initialize the database
const db = new BrainyData()
await db.init()

// Add data (automatically converted to vectors)
const catId = await db.add("Cats are independent pets", {
  noun: NounType.Thing,
  category: 'animal'
})

const dogId = await db.add("Dogs are loyal companions", {
  noun: NounType.Thing,
  category: 'animal'
})

// Search for similar items
const results = await db.searchText("feline pets", 2)
console.log(results)

// Add a relationship between items
await db.addVerb(catId, dogId, {
  verb: VerbType.RelatedTo,
  description: 'Both are common household pets'
})

Import Options

// Standard import - automatically adapts to any environment
import { BrainyData } from '@soulcraft/brainy'

// Minified version for production
import { BrainyData } from '@soulcraft/brainy/min'

Note

: The CLI functionality is available as a separate package @soulcraft/brainy-cli to reduce the bundle size of the main package. Install it globally with npm install -g @soulcraft/brainy-cli to use the command-line interface.

Browser Usage

<script type="module">
  // Use local files instead of CDN
  import { BrainyData } from './dist/unified.js'

  // Or minified version
  // import { BrainyData } from './dist/unified.min.js'

  const db = new BrainyData()
  await db.init()
  // ...
</script>

Modern bundlers like Webpack, Rollup, and Vite will automatically use the unified build which adapts to any environment.

🧩 How It Works

Brainy combines four key technologies to create its adaptive intelligence:

1. Vector Embeddings - Converts data (text, images, etc.) into numerical vectors that capture semantic meaning
2. HNSW Algorithm - Enables fast similarity search through a hierarchical graph structure
3. Adaptive Environment Detection - Automatically senses your platform and optimizes accordingly:
   • Detects browser, Node.js, and serverless environments
   • Adjusts performance parameters based on available resources
   • Learns from query patterns to optimize future searches
   • Tunes itself for your specific use cases
4. Intelligent Storage Selection - Uses the best available storage option for your environment:
   • Browser: Origin Private File System (OPFS)
   • Node.js: File system
   • Server: S3-compatible storage (optional)
   • Serverless: In-memory storage with optional cloud persistence
   • Fallback: In-memory storage
   • Automatically migrates between storage types as needed
   • Uses a simplified, consolidated storage structure for all noun types

🚀 The Brainy Pipeline

Brainy's data processing pipeline transforms raw data into searchable, connected knowledge that gets smarter over time:

Raw Data → Embedding → Vector Storage → Graph Connections → Adaptive Learning → Query & Retrieval

Each time data flows through this pipeline, Brainy learns more about your usage patterns and environment, making future operations faster and more relevant.

Pipeline Stages

1. Data Ingestion

   • Raw text or pre-computed vectors enter the pipeline
   • Data is validated and prepared for processing

2. Embedding Generation

   • Text is transformed into numerical vectors using embedding models
   • Uses TensorFlow Universal Sentence Encoder for high-quality text embeddings
   • Custom embedding functions can be plugged in for specialized domains

3. Vector Indexing

   • Vectors are indexed using the HNSW algorithm
   • Hierarchical structure enables fast similarity search
   • Configurable parameters for precision vs. performance tradeoffs

4. Graph Construction

   • Nouns (entities) become nodes in the knowledge graph
   • Verbs (relationships) connect related entities
   • Typed relationships add semantic meaning to connections

5. Adaptive Learning

   • Analyzes usage patterns to optimize future operations
   • Tunes performance parameters based on your environment
   • Adjusts search strategies based on query history
   • Becomes more efficient and relevant the more you use it

6. Intelligent Storage

   • Data is saved using the optimal storage for your environment
   • Automatic selection between OPFS, filesystem, S3, or memory
   • Migrates between storage types as your application's needs evolve
   • Scales from tiny datasets to massive data collections
   • Configurable storage adapters for custom persistence needs

Augmentation Types

Brainy uses a powerful augmentation system to extend functionality. Augmentations are processed in the following order:

1. SENSE

   • Ingests and processes raw, unstructured data into nouns and verbs
   • Handles text, images, audio streams, and other input formats
   • Example: Converting raw text into structured entities

2. MEMORY

   • Provides storage capabilities for data in different formats
   • Manages persistence across sessions
   • Example: Storing vectors in OPFS or filesystem

3. COGNITION

   • Enables advanced reasoning, inference, and logical operations
   • Analyzes relationships between entities
   • Examples:
     • Inferring new connections between existing data
     • Deriving insights from graph relationships

4. CONDUIT

   • Establishes channels for structured data exchange
   • Connects with external systems and syncs between Brainy instances
   • Two built-in iConduit augmentations for scaling out and syncing:
     • WebSocket iConduit - Syncs data between browsers and servers
     • WebRTC iConduit - Direct peer-to-peer syncing between browsers
   • Examples:
     • Integrating with third-party APIs
     • Syncing Brainy instances between browsers using WebSockets
     • Peer-to-peer syncing between browsers using WebRTC

5. ACTIVATION

   • Initiates actions, responses, or data manipulations
   • Triggers events based on data changes
   • Example: Sending notifications when new data is processed

6. PERCEPTION

   • Interprets, contextualizes, and visualizes identified nouns and verbs
   • Creates meaningful representations of data
   • Example: Generating visualizations of graph relationships

7. DIALOG

   • Facilitates natural language understanding and generation
   • Enables conversational interactions
   • Example: Processing user queries and generating responses

8. WEBSOCKET

   • Enables real-time communication via WebSockets
   • Can be combined with other augmentation types
   • Example: Streaming data processing in real-time

Streaming Data Support

Brainy's pipeline is designed to handle streaming data efficiently:

1. WebSocket Integration

   • Built-in support for WebSocket connections
   • Process data as it arrives without blocking
   • Example: setupWebSocketPipeline(url, dataType, options)

2. Asynchronous Processing

   • Non-blocking architecture for real-time data handling
   • Parallel processing of incoming streams
   • Example: createWebSocketHandler(connection, dataType, options)

3. Event-Based Architecture

   • Augmentations can listen to data feeds and streams
   • Real-time updates propagate through the pipeline
   • Example: listenToFeed(feedUrl, callback)

4. Threaded Execution

   • Comprehensive multi-threading for high-performance operations
   • Parallel processing for batch operations, vector calculations, and embedding generation
   • Configurable execution modes (SEQUENTIAL, PARALLEL, THREADED)
   • Automatic thread management based on environment capabilities
   • Example: executeTypedPipeline(augmentations, method, args, { mode: ExecutionMode.THREADED })

Running the Pipeline

The pipeline runs automatically when you:

// Add data (runs embedding → indexing → storage)
const id = await db.add("Your text data here", { metadata })

// Search (runs embedding → similarity search)
const results = await db.searchText("Your query here", 5)

// Connect entities (runs graph construction → storage)
await db.addVerb(sourceId, targetId, { verb: VerbType.RelatedTo })

Using the CLI:

# Add data through the CLI pipeline
brainy add "Your text data here" '{"noun":"Thing"}'

# Search through the CLI pipeline
brainy search "Your query here" --limit 5

# Connect entities through the CLI
brainy addVerb <sourceId> <targetId> RelatedTo

Extending the Pipeline

Brainy's pipeline is designed for extensibility at every stage:

1. Custom Embedding

   // Create your own embedding function
   const myEmbedder = async (text) => {
     // Your custom embedding logic here
     return [0.1, 0.2, 0.3, ...] // Return a vector
   }
   
   // Use it in Brainy
   const db = new BrainyData({
     embeddingFunction: myEmbedder
   })
   

2. Custom Distance Functions

   // Define your own distance function
   const myDistance = (a, b) => {
     // Your custom distance calculation
     return Math.sqrt(a.reduce((sum, val, i) => sum + Math.pow(val - b[i], 2), 0))
   }
   
   // Use it in Brainy
   const db = new BrainyData({
     distanceFunction: myDistance
   })
   

3. Custom Storage Adapters

   // Implement the StorageAdapter interface
   class MyStorage implements StorageAdapter {
     // Your storage implementation
   }
   
   // Use it in Brainy
   const db = new BrainyData({
     storageAdapter: new MyStorage()
   })
   

4. Augmentations System

   // Create custom augmentations to extend functionality
   const myAugmentation = {
     type: 'memory',
     name: 'my-custom-storage',
     // Implementation details
   }
   
   // Register with Brainy
   db.registerAugmentation(myAugmentation)
   

Data Model

Brainy uses a graph-based data model with two primary concepts:

Nouns (Entities)

The main entities in your data (nodes in the graph):

• Each noun has a unique ID, vector representation, and metadata
• Nouns can be categorized by type (Person, Place, Thing, Event, Concept, etc.)
• Nouns are automatically vectorized for similarity search

Verbs (Relationships)

Connections between nouns (edges in the graph):

• Each verb connects a source noun to a target noun
• Verbs have types that define the relationship (RelatedTo, Controls, Contains, etc.)
• Verbs can have their own metadata to describe the relationship

Type Utilities

Brainy provides utility functions to access lists of noun and verb types:

import {
  NounType,
  VerbType,
  getNounTypes,
  getVerbTypes,
  getNounTypeMap,
  getVerbTypeMap
} from '@soulcraft/brainy'

// At development time:
// Access specific types directly from the NounType and VerbType objects
console.log(NounType.Person)        // 'person'
console.log(VerbType.Contains)      // 'contains'

// At runtime:
// Get a list of all noun types
const nounTypes = getNounTypes()    // ['person', 'organization', 'location', ...]

// Get a list of all verb types
const verbTypes = getVerbTypes()    // ['relatedTo', 'contains', 'partOf', ...]

// Get a map of noun type keys to values
const nounTypeMap = getNounTypeMap() // { Person: 'person', Organization: 'organization', ... }

// Get a map of verb type keys to values
const verbTypeMap = getVerbTypeMap() // { RelatedTo: 'relatedTo', Contains: 'contains', ... }

These utility functions make it easy to:

• Get a complete list of available noun and verb types
• Validate user input against valid types
• Create dynamic UI components that display or select from available types
• Map between type keys and their string values

Command Line Interface

Brainy includes a powerful CLI for managing your data. The CLI is available as a separate package @soulcraft/brainy-cli to reduce the bundle size of the main package.

Installing and Using the CLI

# Install the CLI globally
npm install -g @soulcraft/brainy-cli

# Initialize a database
brainy init

# Add some data
brainy add "Cats are independent pets" '{"noun":"Thing","category":"animal"}'
brainy add "Dogs are loyal companions" '{"noun":"Thing","category":"animal"}'

# Search for similar items
brainy search "feline pets" 5

# Add relationships between items
brainy addVerb <sourceId> <targetId> RelatedTo '{"description":"Both are pets"}'

# Visualize the graph structure
brainy visualize
brainy visualize --root <id> --depth 3

Using the CLI in Your Code

The CLI functionality is available as a separate package @soulcraft/brainy-cli. If you need CLI functionality in your application, install the CLI package:

npm install @soulcraft/brainy-cli

Then you can use the CLI commands programmatically or through the command line interface.

Available Commands

Basic Database Operations:

• init - Initialize a new database
• add <text> [metadata] - Add a new noun with text and optional metadata
• search <query> [limit] - Search for nouns similar to the query
• get <id> - Get a noun by ID
• delete <id> - Delete a noun by ID
• addVerb <sourceId> <targetId> <verbType> [metadata] - Add a relationship
• getVerbs <id> - Get all relationships for a noun
• status - Show database status
• clear - Clear all data from the database
• generate-random-graph - Generate test data
• visualize - Visualize the graph structure
• completion-setup - Setup shell autocomplete

Pipeline and Augmentation Commands:

• list-augmentations - List all available augmentation types and registered augmentations
• augmentation-info <type> - Get detailed information about a specific augmentation type
• test-pipeline [text] - Test the sequential pipeline with sample data
  • -t, --data-type <type> - Type of data to process (default: 'text')
  • -m, --mode <mode> - Execution mode: sequential, parallel, threaded (default: 'sequential')
  • -s, --stop-on-error - Stop execution if an error occurs
  • -v, --verbose - Show detailed output
• stream-test - Test streaming data through the pipeline (simulated)
  • -c, --count <number> - Number of data items to stream (default: 5)
  • -i, --interval <ms> - Interval between data items in milliseconds (default: 1000)
  • -t, --data-type <type> - Type of data to process (default: 'text')
  • -v, --verbose - Show detailed output

API Reference

Database Management

// Initialize the database
await db.init()

// Clear all data
await db.clear()

// Get database status
const status = await db.status()

// Backup all data from the database
const backupData = await db.backup()

// Restore data into the database
const restoreResult = await db.restore(backupData, { clearExisting: true })

Database Statistics

Brainy provides a way to get statistics about the current state of the database. For detailed information about the statistics system, including implementation details, scalability improvements, and usage examples, see our Statistics Guide.

import { BrainyData, getStatistics } from '@soulcraft/brainy'

// Create and initialize the database
const db = new BrainyData()
await db.init()

// Get statistics using the instance method
const stats = await db.getStatistics()
console.log(stats)
// Output: { nounCount: 0, verbCount: 0, metadataCount: 0, hnswIndexSize: 0, serviceBreakdown: {...} }

Working with Nouns (Entities)

// Add a noun (automatically vectorized)
const id = await db.add(textOrVector, {
  noun: NounType.Thing,
  // other metadata...
})

// Add multiple nouns in parallel (with multithreading and batch embedding)
const ids = await db.addBatch([
  {
    vectorOrData: "First item to add",
    metadata: { noun: NounType.Thing, category: 'example' }
  },
  {
    vectorOrData: "Second item to add",
    metadata: { noun: NounType.Thing, category: 'example' }
  },
  // More items...
], {
  forceEmbed: false,
  concurrency: 4, // Control the level of parallelism (default: 4)
  batchSize: 50   // Control the number of items to process in a single batch (default: 50)
})

// Retrieve a noun
const noun = await db.get(id)

// Update noun metadata
await db.updateMetadata(id, {
  noun: NounType.Thing,
  // updated metadata...
})

// Delete a noun
await db.delete(id)

// Search for similar nouns
const results = await db.search(vectorOrText, numResults)
const textResults = await db.searchText("query text", numResults)

// Search by noun type
const thingNouns = await db.searchByNounTypes([NounType.Thing], numResults)

// Search within specific fields of JSON documents
const fieldResults = await db.search("Acme Corporation", 10, {
  searchField: "company"
})

// Search using standard field names across different services
const titleResults = await db.searchByStandardField("title", "climate change", 10)
const authorResults = await db.searchByStandardField("author", "johndoe", 10, {
  services: ["github", "reddit"]
})

Field Standardization and Service Tracking

Brainy automatically tracks field names from JSON documents and associates them with the service that inserted the data. This enables powerful cross-service search capabilities:

// Get all available field names organized by service
const fieldNames = await db.getAvailableFieldNames()
// Example output: { "github": ["repository.name", "issue.title"], "reddit": ["title", "selftext"] }

// Get standard field mappings
const standardMappings = await db.getStandardFieldMappings()
// Example output: { "title": { "github": ["repository.name"], "reddit": ["title"] } }

When adding data, specify the service name to ensure proper field tracking:

// Add data with service name
await db.add(jsonData, metadata, { service: "github" })

Working with Verbs (Relationships)

// Add a relationship between nouns
await db.addVerb(sourceId, targetId, {
  verb: VerbType.RelatedTo,
  // other metadata...
})

// Add a relationship with auto-creation of missing nouns
// This is useful when the target noun might not exist yet
await db.addVerb(sourceId, targetId, {
  verb: VerbType.RelatedTo,
  // Enable auto-creation of missing nouns
  autoCreateMissingNouns: true,
  // Optional metadata for auto-created nouns
  missingNounMetadata: {
    noun: NounType.Concept,
    description: 'Auto-created noun'
  }
})

// Get all relationships
const verbs = await db.getAllVerbs()

// Get relationships by source noun
const outgoingVerbs = await db.getVerbsBySource(sourceId)

// Get relationships by target noun
const incomingVerbs = await db.getVerbsByTarget(targetId)

// Get relationships by type
const containsVerbs = await db.getVerbsByType(VerbType.Contains)

// Get a specific relationship
const verb = await db.getVerb(verbId)

// Delete a relationship
await db.deleteVerb(verbId)

Advanced Configuration

Database Modes

Brainy supports special operational modes that restrict certain operations:

import { BrainyData } from '@soulcraft/brainy'

// Create and initialize the database
const db = new BrainyData()
await db.init()

// Set the database to read-only mode (prevents write operations)
db.setReadOnly(true)

// Check if the database is in read-only mode
const isReadOnly = db.isReadOnly() // Returns true

// Set the database to write-only mode (prevents search operations)
db.setWriteOnly(true)

// Check if the database is in write-only mode
const isWriteOnly = db.isWriteOnly() // Returns true

// Reset to normal mode (allows both read and write operations)
db.setReadOnly(false)
db.setWriteOnly(false)

• Read-Only Mode: When enabled, prevents all write operations (add, update, delete). Useful for deployment scenarios where you want to prevent modifications to the database.
• Write-Only Mode: When enabled, prevents all search operations. Useful for initial data loading or when you want to optimize for write performance.

Embedding

import {
  BrainyData,
  createTensorFlowEmbeddingFunction,
  createThreadedEmbeddingFunction
} from '@soulcraft/brainy'

// Use the standard TensorFlow Universal Sentence Encoder embedding function
const db = new BrainyData({
  embeddingFunction: createTensorFlowEmbeddingFunction()
})
await db.init()

// Or use the threaded embedding function for better performance
const threadedDb = new BrainyData({
  embeddingFunction: createThreadedEmbeddingFunction()
})
await threadedDb.init()

// Directly embed text to vectors
const vector = await db.embed("Some text to convert to a vector")

// Calculate similarity between two texts or vectors
const similarity = await db.calculateSimilarity(
  "Cats are furry pets",
  "Felines make good companions"
)
console.log(`Similarity score: ${similarity}`) // Higher value means more similar

// Calculate similarity with custom options
const vectorA = await db.embed("First text")
const vectorB = await db.embed("Second text")
const customSimilarity = await db.calculateSimilarity(
  vectorA, // Can use pre-computed vectors
  vectorB,
  {
    forceEmbed: false, // Skip embedding if inputs are already vectors
    distanceFunction: cosineDistance // Optional custom distance function
  }
)

The threaded embedding function runs in a separate thread (Web Worker in browsers, Worker Thread in Node.js) to improve performance, especially for embedding operations. It uses GPU acceleration when available (via WebGL in browsers) and falls back to CPU processing for compatibility. Universal Sentence Encoder is always used for embeddings. The implementation includes worker reuse and model caching for optimal performance.

Performance Tuning

Brainy includes comprehensive performance optimizations that work across all environments (browser, CLI, Node.js, container, server):

GPU and CPU Optimization

Brainy uses GPU and CPU optimization for compute-intensive operations:

1. GPU-Accelerated Embeddings: Generate text embeddings using TensorFlow.js with WebGL backend when available
2. Automatic Fallback: Falls back to CPU backend when GPU is not available
3. Optimized Distance Calculations: Perform vector similarity calculations with optimized algorithms
4. Cross-Environment Support: Works consistently across browsers and Node.js environments
5. Memory Management: Properly disposes of tensors to prevent memory leaks

Multithreading Support

Brainy includes comprehensive multithreading support to improve performance across all environments:

1. Parallel Batch Processing: Add multiple items concurrently with controlled parallelism
2. Multithreaded Vector Search: Perform distance calculations in parallel for faster search operations
3. Threaded Embedding Generation: Generate embeddings in separate threads to avoid blocking the main thread
4. Worker Reuse: Maintains a pool of workers to avoid the overhead of creating and terminating workers
5. Model Caching: Initializes the embedding model once per worker and reuses it for multiple operations
6. Batch Embedding: Processes multiple items in a single embedding operation for better performance
7. Automatic Environment Detection: Adapts to browser (Web Workers) and Node.js (Worker Threads) environments

import { BrainyData, euclideanDistance } from '@soulcraft/brainy'

// Configure with custom options
const db = new BrainyData({
  // Use Euclidean distance instead of default cosine distance
  distanceFunction: euclideanDistance,

  // HNSW index configuration for search performance
  hnsw: {
    M: 16,              // Max connections per noun
    efConstruction: 200, // Construction candidate list size
    efSearch: 50,       // Search candidate list size
  },

  // Performance optimization options
  performance: {
    useParallelization: true, // Enable multithreaded search operations
  },

  // Noun and Verb type validation
  typeValidation: {
    enforceNounTypes: true,  // Validate noun types against NounType enum
    enforceVerbTypes: true,  // Validate verb types against VerbType enum
  },

  // Storage configuration
  storage: {
    requestPersistentStorage: true,
    // Example configuration for cloud storage (replace with your own values):
    // s3Storage: {
    //   bucketName: 'your-s3-bucket-name',
    //   region: 'your-aws-region'
    //   // Credentials should be provided via environment variables
    //   // AWS_ACCESS_KEY_ID and AWS_SECRET_ACCESS_KEY
    // }
  }
})

Optimized HNSW for Large Datasets

Brainy includes an optimized HNSW index implementation for large datasets that may not fit entirely in memory, using a hybrid approach:

1. Product Quantization - Reduces vector dimensionality while preserving similarity relationships
2. Disk-Based Storage - Offloads vectors to disk when memory usage exceeds a threshold
3. Memory-Efficient Indexing - Optimizes memory usage for large-scale vector collections

import { BrainyData } from '@soulcraft/brainy'

// Configure with optimized HNSW index for large datasets
const db = new BrainyData({
  hnswOptimized: {
    // Standard HNSW parameters
    M: 16,              // Max connections per noun
    efConstruction: 200, // Construction candidate list size
    efSearch: 50,       // Search candidate list size

    // Memory threshold in bytes - when exceeded, will use disk-based approach
    memoryThreshold: 1024 * 1024 * 1024, // 1GB default threshold

    // Product quantization settings for dimensionality reduction
    productQuantization: {
      enabled: true,              // Enable product quantization
      numSubvectors: 16,          // Number of subvectors to split the vector into
      numCentroids: 256           // Number of centroids per subvector
    },

    // Whether to use disk-based storage for the index
    useDiskBasedIndex: true         // Enable disk-based storage
  },

  // Storage configuration (required for disk-based index)
  storage: {
    requestPersistentStorage: true
  }
})

// The optimized index automatically adapts based on dataset size:
// 1. For small datasets: Uses standard in-memory approach
// 2. For medium datasets: Applies product quantization to reduce memory usage
// 3. For large datasets: Combines product quantization with disk-based storage

// Check status to see memory usage and optimization details
const status = await db.status()
console.log(status.details.index)

Distance Functions

Brainy provides several distance functions for vector similarity calculations:

• cosineDistance (default): Measures the cosine of the angle between vectors (1 - cosine similarity)
• euclideanDistance: Measures the straight-line distance between vectors
• manhattanDistance: Measures the sum of absolute differences between vector components
• dotProductDistance: Measures the negative dot product between vectors

All distance functions are optimized for performance and automatically use the most efficient implementation based on the dataset size and available resources. For large datasets and high-dimensional vectors, Brainy uses batch processing and multithreading when available to improve performance.

Backup and Restore

Brainy provides backup and restore capabilities that allow you to:

• Back up your data
• Transfer data between Brainy instances
• Restore existing data into Brainy for vectorization and indexing
• Backup data for analysis or visualization in other tools

Backing Up Data

// Backup all data from the database
const backupData = await db.backup()

// The backup data includes:
// - All nouns (entities) with their vectors and metadata
// - All verbs (relationships) between nouns
// - Noun types and verb types
// - HNSW index data for fast similarity search
// - Version information

// Save the backup data to a file (Node.js environment)
import fs from 'fs'

fs.writeFileSync('brainy-backup.json', JSON.stringify(backupData, null, 2))

Restoring Data

Brainy's restore functionality can handle:

1. Complete backups with vectors and index data
2. Sparse data without vectors (vectors will be created during restore)
3. Data without HNSW index (index will be reconstructed if needed)

// Restore data with all options
const restoreResult = await db.restore(backupData, {
  clearExisting: true // Whether to clear existing data before restore
})

// Import sparse data (without vectors)
// Vectors will be automatically created using the embedding function
const sparseData = {
  nouns: [
    {
      id: '123',
      // No vector field - will be created during import
      metadata: {
        noun: 'Thing',
        text: 'This text will be used to generate a vector'
      }
    }
  ],
  verbs: [],
  version: '1.0.0'
}

const sparseImportResult = await db.importSparseData(sparseData)

CLI Backup/Restore

# Backup data to a file
brainy backup --output brainy-backup.json

# Restore data from a file
brainy restore --input brainy-backup.json --clear-existing

# Import sparse data (without vectors)
brainy import-sparse --input sparse-data.json

Embedding

Brainy uses the following embedding approach:

• TensorFlow Universal Sentence Encoder (high-quality text embeddings)
• GPU acceleration when available (via WebGL in browsers)
• Batch embedding for processing multiple items efficiently
• Worker reuse and model caching for optimal performance
• Custom embedding functions can be plugged in for specialized domains

Extensions

Brainy includes an augmentation system for extending functionality:

• Memory Augmentations: Different storage backends
• Sense Augmentations: Process raw data
• Cognition Augmentations: Reasoning and inference
• Dialog Augmentations: Text processing and interaction
• Perception Augmentations: Data interpretation and visualization
• Activation Augmentations: Trigger actions

Simplified Augmentation System

Brainy provides a simplified factory system for creating, importing, and executing augmentations with minimal boilerplate:

import {
  createMemoryAugmentation,
  createConduitAugmentation,
  createSenseAugmentation,
  addWebSocketSupport,
  executeStreamlined,
  processStaticData,
  processStreamingData,
  createPipeline
} from '@soulcraft/brainy'

// Create a memory augmentation with minimal code
const memoryAug = createMemoryAugmentation({
  name: 'simple-memory',
  description: 'A simple in-memory storage augmentation',
  autoRegister: true,
  autoInitialize: true,

  // Implement only the methods you need
  storeData: async (key, data) => {
    // Your implementation here
    return {
      success: true,
      data: true
    }
  },

  retrieveData: async (key) => {
    // Your implementation here
    return {
      success: true,
      data: { example: 'data', key }
    }
  }
})

// Add WebSocket support to any augmentation
const wsAugmentation = addWebSocketSupport(memoryAug, {
  connectWebSocket: async (url) => {
    // Your implementation here
    return {
      connectionId: 'ws-1',
      url,
      status: 'connected'
    }
  }
})

// Process static data through a pipeline
const result = await processStaticData(
  'Input data',
  [
    {
      augmentation: senseAug,
      method: 'processRawData',
      transformArgs: (data) => [data, 'text']
    },
    {
      augmentation: memoryAug,
      method: 'storeData',
      transformArgs: (data) => ['processed-data', data]
    }
  ]
)

// Create a reusable pipeline
const pipeline = createPipeline([
  {
    augmentation: senseAug,
    method: 'processRawData',
    transformArgs: (data) => [data, 'text']
  },
  {
    augmentation: memoryAug,
    method: 'storeData',
    transformArgs: (data) => ['processed-data', data]
  }
])

// Use the pipeline
const result = await pipeline('New input data')

// Dynamically load augmentations at runtime
const loadedAugmentations = await loadAugmentationModule(
  import('./my-augmentations.js'),
  {
    autoRegister: true,
    autoInitialize: true
  }
)

The simplified augmentation system provides:

1. Factory Functions - Create augmentations with minimal boilerplate
2. WebSocket Support - Add WebSocket capabilities to any augmentation
3. Streamlined Pipeline - Process data through augmentations more efficiently
4. Dynamic Loading - Load augmentations at runtime when needed
5. Static & Streaming Data - Handle both static and streaming data with the same API

WebSocket Augmentation Types

Brainy exports several WebSocket augmentation types that can be used by augmentation creators to add WebSocket capabilities to their augmentations:

import {
  // Base WebSocket support interface
  IWebSocketSupport,

  // Combined WebSocket augmentation types
  IWebSocketSenseAugmentation,
  IWebSocketConduitAugmentation,
  IWebSocketCognitionAugmentation,
  IWebSocketMemoryAugmentation,
  IWebSocketPerceptionAugmentation,
  IWebSocketDialogAugmentation,
  IWebSocketActivationAugmentation,

  // Function to add WebSocket support to any augmentation
  addWebSocketSupport
} from '@soulcraft/brainy'

// Example: Creating a typed WebSocket-enabled sense augmentation
const mySenseAug = createSenseAugmentation({
  name: 'my-sense',
  processRawData: async (data, dataType) => {
    // Implementation
    return {
      success: true,
      data: { nouns: [], verbs: [] }
    }
  }
}) as IWebSocketSenseAugmentation

// Add WebSocket support
addWebSocketSupport(mySenseAug, {
  connectWebSocket: async (url) => {
    // WebSocket implementation
    return {
      connectionId: 'ws-1',
      url,
      status: 'connected'
    }
  },
  sendWebSocketMessage: async (connectionId, data) => {
    // Send message implementation
  },
  onWebSocketMessage: async (connectionId, callback) => {
    // Register callback implementation
  },
  offWebSocketMessage: async (connectionId, callback) => {
    // Remove callback implementation
  },
  closeWebSocket: async (connectionId, code, reason) => {
    // Close connection implementation
  }
})

// Now mySenseAug has both sense augmentation methods and WebSocket methods
await mySenseAug.processRawData('data', 'text')
await mySenseAug.connectWebSocket('wss://example.com')

These WebSocket augmentation types combine the base augmentation interfaces with the IWebSocketSupport interface, providing type safety and autocompletion for augmentations with WebSocket capabilities.

Model Control Protocol (MCP)

Brainy includes a Model Control Protocol (MCP) implementation that allows external models to access Brainy data and use the augmentation pipeline as tools:

• BrainyMCPAdapter: Provides access to Brainy data through MCP
• MCPAugmentationToolset: Exposes the augmentation pipeline as tools
• BrainyMCPService: Integrates the adapter and toolset, providing WebSocket and REST server implementations

Environment compatibility:

• BrainyMCPAdapter and MCPAugmentationToolset can run in any environment (browser, Node.js, server)
• BrainyMCPService core functionality works in any environment

For detailed documentation and usage examples, see the MCP documentation.

Cross-Environment Compatibility

Brainy is designed to run seamlessly in any environment, from browsers to Node.js to serverless functions and containers. All Brainy data, functions, and augmentations are environment-agnostic, allowing you to use the same code everywhere.

Environment Detection

Brainy automatically detects the environment it's running in:

import { environment } from '@soulcraft/brainy'

// Check which environment we're running in
console.log(`Running in ${
  environment.isBrowser ? 'browser' :
    environment.isNode ? 'Node.js' :
      'serverless/unknown'
} environment`)

Adaptive Storage

Storage adapters are automatically selected based on the environment:

• Browser: Uses Origin Private File System (OPFS) when available, falls back to in-memory storage
• Node.js: Uses file system storage by default, with options for S3-compatible cloud storage
• Serverless: Uses in-memory storage with options for cloud persistence
• Container: Automatically detects and uses the appropriate storage based on available capabilities

Dynamic Imports

Brainy uses dynamic imports to load environment-specific dependencies only when needed, keeping the bundle size small and ensuring compatibility across environments.

Browser Support

Works in all modern browsers:

• Chrome 86+
• Edge 86+
• Opera 72+
• Chrome for Android 86+

For browsers without OPFS support, falls back to in-memory storage.

Related Projects

• Cartographer - A companion project that provides standardized interfaces for interacting with Brainy

Demo

The repository includes a comprehensive demo that showcases Brainy's main features:

• demo/index.html - A single demo page with animations demonstrating Brainy's features.
  • Try the live demo - Check out the interactive demo on GitHub Pages
  • Or run it locally with npm run demo (see demo instructions for details)
  • To deploy your own version to GitHub Pages, use the GitHub Actions workflow in .github/workflows/deploy-demo.yml, which automatically deploys when pushing to the main branch or can be manually triggered
  • To use a custom domain (like www.soulcraft.com):

    1. A CNAME file is already included in the demo directory 2. In your GitHub repository settings, go to Pages > Custom domain and enter your domain 3. Configure your domain's DNS settings to point to GitHub Pages:

       • Add a CNAME record for www pointing to <username>.github.io (e.g., soulcraft-research.github.io)
       • Or for an apex domain (soulcraft.com), add A records pointing to GitHub Pages IP addresses

The demo showcases:

• How Brainy runs in different environments (browser, Node.js, server, cloud)
• How the noun-verb data model works
• How HNSW search works

Syncing Brainy Instances

You can use the conduit augmentations to sync Brainy instances:

• WebSocket iConduit: For syncing between browsers and servers, or between servers. WebSockets cannot be used for direct browser-to-browser communication without a server in the middle.
• WebRTC iConduit: For direct peer-to-peer syncing between browsers. This is the recommended approach for browser-to-browser communication.

WebSocket Sync Example

import {
  BrainyData,
  pipeline,
  createConduitAugmentation
} from '@soulcraft/brainy'

// Create and initialize the database
const db = new BrainyData()
await db.init()

// Create a WebSocket conduit augmentation
const wsConduit = await createConduitAugmentation('websocket', 'my-websocket-sync')

// Register the augmentation with the pipeline
pipeline.register(wsConduit)

// Connect to another Brainy instance (server or browser)
// Replace the example URL below with your actual WebSocket server URL
const connectionResult = await pipeline.executeConduitPipeline(
  'establishConnection',
  ['wss://example-websocket-server.com/brainy-sync', { protocols: 'brainy-sync' }]
)

if (connectionResult[0] && (await connectionResult[0]).success) {
  const connection = (await connectionResult[0]).data

  // Read data from the remote instance
  const readResult = await pipeline.executeConduitPipeline(
    'readData',
    [{ connectionId: connection.connectionId, query: { type: 'getAllNouns' } }]
  )

  // Process and add the received data to the local instance
  if (readResult[0] && (await readResult[0]).success) {
    const remoteNouns = (await readResult[0]).data
    for (const noun of remoteNouns) {
      await db.add(noun.vector, noun.metadata)
    }
  }

  // Set up real-time sync by monitoring the stream
  await wsConduit.monitorStream(connection.connectionId, async (data) => {
    // Handle incoming data (e.g., new nouns, verbs, updates)
    if (data.type === 'newNoun') {
      await db.add(data.vector, data.metadata)
    } else if (data.type === 'newVerb') {
      await db.addVerb(data.sourceId, data.targetId, data.vector, data.options)
    }
  })
}

WebRTC Peer-to-Peer Sync Example

import {
  BrainyData,
  pipeline,
  createConduitAugmentation
} from '@soulcraft/brainy'

// Create and initialize the database
const db = new BrainyData()
await db.init()

// Create a WebRTC conduit augmentation
const webrtcConduit = await createConduitAugmentation('webrtc', 'my-webrtc-sync')

// Register the augmentation with the pipeline
pipeline.register(webrtcConduit)

// Connect to a peer using a signaling server
// Replace the example values below with your actual configuration
const connectionResult = await pipeline.executeConduitPipeline(
  'establishConnection',
  [
    'peer-id-to-connect-to', // Replace with actual peer ID
    {
      signalServerUrl: 'wss://example-signal-server.com', // Replace with your signal server
      localPeerId: 'my-local-peer-id', // Replace with your local peer ID
      iceServers: [{ urls: 'stun:stun.l.google.com:19302' }] // Public STUN server
    }
  ]
)

if (connectionResult[0] && (await connectionResult[0]).success) {
  const connection = (await connectionResult[0]).data

  // Set up real-time sync by monitoring the stream
  await webrtcConduit.monitorStream(connection.connectionId, async (data) => {
    // Handle incoming data (e.g., new nouns, verbs, updates)
    if (data.type === 'newNoun') {
      await db.add(data.vector, data.metadata)
    } else if (data.type === 'newVerb') {
      await db.addVerb(data.sourceId, data.targetId, data.vector, data.options)
    }
  })

  // When adding new data locally, also send to the peer
  const nounId = await db.add("New data to sync", { noun: "Thing" })

  // Send the new noun to the peer
  await pipeline.executeConduitPipeline(
    'writeData',
    [
      {
        connectionId: connection.connectionId,
        data: {
          type: 'newNoun',
          id: nounId,
          vector: (await db.get(nounId)).vector,
          metadata: (await db.get(nounId)).metadata
        }
      }
    ]
  )
}

Browser-Server Search Example

Brainy supports searching a server-hosted instance from a browser, storing results locally, and performing further searches against the local instance:

import { BrainyData } from '@soulcraft/brainy'

// Create and initialize the database with remote server configuration
// Replace the example URL below with your actual Brainy server URL
const db = new BrainyData({
  remoteServer: {
    url: 'wss://example-brainy-server.com/ws', // Replace with your server URL
    protocols: 'brainy-sync',
    autoConnect: true // Connect automatically during initialization
  }
})
await db.init()

// Or connect manually after initialization
if (!db.isConnectedToRemoteServer()) {
  // Replace the example URL below with your actual Brainy server URL
  await db.connectToRemoteServer('wss://example-brainy-server.com/ws', 'brainy-sync')
}

// Search the remote server (results are stored locally)
const remoteResults = await db.searchText('machine learning', 5, { searchMode: 'remote' })

// Search the local database (includes previously stored results)
const localResults = await db.searchText('machine learning', 5, { searchMode: 'local' })

// Perform a combined search (local first, then remote if needed)
const combinedResults = await db.searchText('neural networks', 5, { searchMode: 'combined' })

// Add data to both local and remote instances
const id = await db.addToBoth('Deep learning is a subset of machine learning', {
  noun: 'Concept',
  category: 'AI',
  tags: ['deep learning', 'neural networks']
})

// Clean up when done (this also cleans up worker pools)
await db.shutDown()

---

📈 Scaling Strategy

Brainy is designed to handle datasets of various sizes, from small collections to large-scale deployments. For terabyte-scale data that can't fit entirely in memory, we provide several approaches:

• Disk-Based HNSW: Modified implementations using intelligent caching and partial loading
• Distributed HNSW: Sharding and partitioning across multiple machines
• Hybrid Solutions: Combining quantization techniques with multi-tier architectures

For detailed information on how to scale Brainy for large datasets, vector dimension standardization, threading implementation, storage testing, and other technical topics, see our comprehensive Technical Guides.

Recent Changes and Performance Improvements

Enhanced Memory Management and Scalability

Brainy has been significantly improved to handle larger datasets more efficiently:

• Pagination Support: All data retrieval methods now support pagination to avoid loading entire datasets into memory at once. The deprecated getAllNouns() and getAllVerbs() methods have been replaced with getNouns() and getVerbs() methods that support pagination, filtering, and cursor-based navigation.

• Multi-level Caching: A sophisticated three-level caching strategy has been implemented:

  • Level 1: Hot cache (most accessed nodes) - RAM (automatically detecting and adjusting in each environment)
  • Level 2: Warm cache (recent nodes) - OPFS, Filesystem or S3 depending on environment
  • Level 3: Cold storage (all nodes) - OPFS, Filesystem or S3 depending on environment

• Adaptive Memory Usage: The system automatically detects available memory and adjusts cache sizes accordingly:

  • In Node.js: Uses 10% of free memory (minimum 1000 entries)
  • In browsers: Scales based on device memory (500 entries per GB, minimum 1000)

• Intelligent Cache Eviction: Implements a Least Recently Used (LRU) policy that evicts the oldest 20% of items when the cache reaches the configured threshold.

• Prefetching Strategy: Implements batch prefetching to improve performance while avoiding overwhelming system resources.

S3-Compatible Storage Improvements

• Enhanced Cloud Storage: Improved support for S3-compatible storage services including AWS S3, Cloudflare R2, and others.

• Optimized Data Access: Batch operations and error handling for efficient cloud storage access.

• Change Log Management: Efficient synchronization through change logs to track updates.

Data Compatibility

Yes, you can use existing data indexed from an old version. Brainy includes robust data migration capabilities:

• Vector Regeneration: If vectors are missing in imported data, they will be automatically created using the embedding function.

• HNSW Index Reconstruction: The system can reconstruct the HNSW index from backup data, ensuring compatibility with previous versions.

• Sparse Data Import: Support for importing sparse data (without vectors) through the importSparseData() method.

System Requirements

Default Mode

• Memory:

  • Minimum: 512MB RAM
  • Recommended: 2GB+ RAM for medium datasets, 8GB+ for large datasets

• CPU:

  • Minimum: 2 cores
  • Recommended: 4+ cores for better performance with parallel operations

• Storage:

  • Minimum: 1GB available storage
  • Recommended: Storage space at least 3x the size of your dataset

Read-Only Mode

Read-only mode prevents all write operations (add, update, delete) and is optimized for search operations.

• Memory:

  • Minimum: 256MB RAM
  • Recommended: 1GB+ RAM

• CPU:

  • Minimum: 1 core
  • Recommended: 2+ cores

• Storage:

  • Minimum: Storage space equal to the size of your dataset
  • Recommended: 2x the size of your dataset for caching

• New Feature: Lazy loading support in read-only mode for improved performance with large datasets.

Write-Only Mode

Write-only mode prevents all search operations and is optimized for initial data loading or when you want to optimize for write performance.

• Memory:

  • Minimum: 512MB RAM
  • Recommended: 2GB+ RAM

• CPU:

  • Minimum: 2 cores
  • Recommended: 4+ cores for faster data ingestion

• Storage:

  • Minimum: Storage space at least 2x the size of your dataset
  • Recommended: 4x the size of your dataset for optimal performance

Performance Tuning Parameters

Brainy offers comprehensive configuration options for performance tuning, with enhanced support for large datasets in S3 or other remote storage. All configuration is optional - the system automatically detects the optimal settings based on your environment, dataset size, and usage patterns.

Intelligent Defaults

Brainy uses intelligent defaults that automatically adapt to your environment:

• Environment Detection: Automatically detects whether you're running in Node.js, browser, or worker environment
• Memory-Aware Caching: Adjusts cache sizes based on available system memory
• Dataset Size Adaptation: Tunes parameters based on the size of your dataset
• Usage Pattern Optimization: Adjusts to read-heavy vs. write-heavy workloads
• Storage Type Awareness: Optimizes for local vs. remote storage (S3, R2, etc.)
• Operating Mode Specialization: Special optimizations for read-only and write-only modes

Cache Configuration (Optional)

You can override any of these automatically tuned parameters if needed:

• Hot Cache Size: Control the maximum number of items to keep in memory.

  • For large datasets (>100K items), consider values between 5,000-50,000 depending on available memory.
  • In read-only mode, larger values (10,000-100,000) can be used for better performance.

• Eviction Threshold: Set the threshold at which cache eviction begins (default: 0.8 or 80% of max size).

  • For write-heavy workloads, lower values (0.6-0.7) may improve performance.
  • For read-heavy workloads, higher values (0.8-0.9) are recommended.

• Warm Cache TTL: Set the time-to-live for items in the warm cache (default: 3600000 ms or 1 hour).

  • For frequently changing data, shorter TTLs are recommended.
  • For relatively static data, longer TTLs improve performance.

• Batch Size: Control the number of items to process in a single batch for operations like prefetching.

  • For S3 or remote storage with large datasets, larger values (50-200) significantly improve throughput.
  • In read-only mode with remote storage, even larger values (100-300) can be used.

Auto-Tuning (Enabled by Default)

• Auto-Tune: Enable or disable automatic tuning of cache parameters based on usage patterns (default: true).
• Auto-Tune Interval: Set how frequently the system adjusts cache parameters (default: 60000 ms or 1 minute).

Read-Only Mode Optimizations (Automatic)

Read-only mode includes special optimizations for search performance that are automatically applied:

• Larger Cache Sizes: Automatically uses more memory for caching (up to 40% of free memory for large datasets).
• Aggressive Prefetching: Loads more data in each batch to reduce the number of storage requests.
• Prefetch Strategy: Defaults to 'aggressive' prefetching strategy in read-only mode.

Example Configuration for Large S3 Datasets

const brainy = new BrainyData({
  readOnly: true,
  lazyLoadInReadOnlyMode: true,
  storage: {
    type: 's3',
    s3Storage: {
      bucketName: 'your-bucket',
      accessKeyId: 'your-access-key',
      secretAccessKey: 'your-secret-key',
      region: 'your-region'
    }
  },
  cache: {
    hotCacheMaxSize: 20000,
    hotCacheEvictionThreshold: 0.85,
    batchSize: 100,
    readOnlyMode: {
      hotCacheMaxSize: 50000,
      batchSize: 200,
      prefetchStrategy: 'aggressive'
    }
  }
});

These configuration options make Brainy more efficient, scalable, and adaptable to different environments and usage patterns, especially for large datasets in cloud storage.

Testing

Brainy uses Vitest for testing. For detailed information about testing in Brainy, including test configuration, scripts, reporting tools, and best practices, see our Testing Guide.

Here are some common test commands:

# Run all tests
npm test

# Run tests with comprehensive reporting
npm run test:report

# Run tests with coverage
npm run test:coverage

Contributing

For detailed contribution guidelines, please see CONTRIBUTING.md.

For developer documentation, including building, testing, and publishing instructions, please see DEVELOPERS.md.

We have a Code of Conduct that all contributors are expected to follow.

Commit Message Format

For best results with automatic changelog generation, follow the Conventional Commits specification for your commit messages:

AI Template for automated commit messages:

Use Conventional Commit format
Specify the changes in a structured format
Add information about the purpose of the commit

<type>(<scope>): <description>

[optional body]

[optional footer(s)]

Where <type> is one of:

• feat: A new feature (maps to Added section)
• fix: A bug fix (maps to Fixed section)
• chore: Regular maintenance tasks (maps to Changed section)
• docs: Documentation changes (maps to Documentation section)
• refactor: Code changes that neither fix bugs nor add features (maps to Changed section)
• perf: Performance improvements (maps to Changed section)

Manual Release Process

If you need more control over the release process, you can use the individual commands:

# Update version and generate changelog
npm run _release:patch  # or _release:minor, _release:major

# Create GitHub release
npm run _github-release

# Publish to NPM
npm publish

License

MIT