perf: pre-compute type embeddings at build time (zero runtime cost)

Major optimization - all type embeddings now built into package:

Build-time generation:
- Created scripts/buildTypeEmbeddings.ts to generate all type embeddings
- Generates embeddings for 31 NounTypes + 40 VerbTypes at build time
- Stores as base64-encoded binary data in embeddedTypeEmbeddings.ts
- Added check script to rebuild only when needed

Updated all consumers:
- NeuralEntityExtractor: loads pre-computed embeddings (instant)
- BrainyTypes: loads pre-computed embeddings (instant init)
- NaturalLanguageProcessor: loads pre-computed embeddings (instant init)

Build process:
- Added npm run build:types to generate embeddings
- Added npm run build:types:if-needed for conditional rebuild
- Integrated into main build pipeline
- Auto-rebuilds only when types or build script change

Benefits:
- Zero runtime cost - embeddings loaded instantly
- Survives all container restarts
- All 71 types always available (31 nouns + 40 verbs)
- ~100KB memory overhead for permanent performance gain
- Eliminates 5-10 second initialization delay

This completes the type embedding optimization started in v3.32.5
This commit is contained in:
David Snelling 2025-10-09 18:08:57 -07:00
parent 87eb60d527
commit 0d649b8a79
9 changed files with 643 additions and 111 deletions

View file

@ -0,0 +1,361 @@
#!/usr/bin/env node
/**
* Build embedded type embeddings with pre-computed vectors
* Generates embeddings for all 31 NounTypes + 40 VerbTypes
* NO runtime computation, NO external files needed!
*/
import { TransformerEmbedding } from '../src/utils/embedding.js'
import * as fs from 'fs/promises'
import * as path from 'path'
import { fileURLToPath } from 'url'
import { NounType, VerbType } from '../src/types/graphTypes.js'
const __dirname = path.dirname(fileURLToPath(import.meta.url))
/**
* Type descriptions for semantic matching
* Copied from BrainyTypes for consistency
*/
const NOUN_TYPE_DESCRIPTIONS: Record<string, string> = {
// Core Entity Types
[NounType.Person]: 'person human individual user employee customer citizen member author creator agent actor participant',
[NounType.Organization]: 'organization company business corporation institution agency department team group committee board',
[NounType.Location]: 'location place address city country region area zone coordinate position site venue building',
[NounType.Thing]: 'thing object item product device equipment tool instrument asset artifact material physical tangible',
[NounType.Concept]: 'concept idea theory principle philosophy belief value abstract intangible notion thought',
[NounType.Event]: 'event occurrence incident activity happening meeting conference celebration milestone timestamp date',
// Digital/Content Types
[NounType.Document]: 'document file report article paper text pdf word contract agreement record documentation',
[NounType.Media]: 'media image photo video audio music podcast multimedia graphic visualization animation',
[NounType.File]: 'file digital data binary code script program software archive package bundle',
[NounType.Message]: 'message email chat communication notification alert announcement broadcast transmission',
[NounType.Content]: 'content information data text material resource publication post blog webpage',
// Collection Types
[NounType.Collection]: 'collection group set list array category folder directory catalog inventory database',
[NounType.Dataset]: 'dataset data table spreadsheet database records statistics metrics measurements analysis',
// Business/Application Types
[NounType.Product]: 'product item merchandise offering service feature application software solution package',
[NounType.Service]: 'service offering subscription support maintenance utility function capability',
[NounType.User]: 'user account profile member subscriber customer client participant identity credentials',
[NounType.Task]: 'task action todo item job assignment duty responsibility activity step procedure',
[NounType.Project]: 'project initiative program campaign effort endeavor plan scheme venture undertaking',
// Descriptive Types
[NounType.Process]: 'process workflow procedure method algorithm sequence pipeline operation routine protocol',
[NounType.State]: 'state status condition phase stage mode situation circumstance configuration setting',
[NounType.Role]: 'role position title function responsibility duty job capacity designation authority',
[NounType.Topic]: 'subject topic theme category tag keyword area domain field discipline specialty',
[NounType.Language]: 'language dialect locale tongue vernacular communication speech linguistics vocabulary',
[NounType.Currency]: 'currency money dollar euro pound yen bitcoin payment financial monetary unit',
[NounType.Measurement]: 'measurement metric quantity value amount size dimension weight height volume distance',
// Scientific/Research Types
[NounType.Hypothesis]: 'hypothesis theory proposition thesis assumption premise conjecture speculation prediction',
[NounType.Experiment]: 'experiment test trial study research investigation analysis observation examination',
// Legal/Regulatory Types
[NounType.Contract]: 'contract agreement deal treaty pact covenant license terms conditions policy',
[NounType.Regulation]: 'regulation law rule policy standard compliance requirement guideline ordinance statute',
// Technical Infrastructure Types
[NounType.Interface]: 'interface API endpoint protocol specification contract schema definition connection',
[NounType.Resource]: 'resource infrastructure server database storage compute memory bandwidth capacity asset'
}
const VERB_TYPE_DESCRIPTIONS: Record<string, string> = {
// Core Relationship Types
[VerbType.RelatedTo]: 'related connected associated linked correlated relevant pertinent applicable',
[VerbType.Contains]: 'contains includes holds stores encompasses comprises consists incorporates',
[VerbType.PartOf]: 'part component element member piece portion section segment constituent',
[VerbType.LocatedAt]: 'located situated positioned placed found exists resides occupies',
[VerbType.References]: 'references cites mentions points links refers quotes sources',
// Temporal/Causal Types
[VerbType.Precedes]: 'precedes before earlier prior previous antecedent preliminary foregoing',
[VerbType.Succeeds]: 'succeeds follows after later subsequent next ensuing succeeding',
[VerbType.Causes]: 'causes triggers induces produces generates results influences affects',
[VerbType.DependsOn]: 'depends requires needs relies necessitates contingent prerequisite',
[VerbType.Requires]: 'requires needs demands necessitates mandates obliges compels entails',
// Creation/Transformation Types
[VerbType.Creates]: 'creates makes produces generates builds constructs forms establishes',
[VerbType.Transforms]: 'transforms converts changes modifies alters transitions morphs evolves',
[VerbType.Becomes]: 'becomes turns evolves transforms changes transitions develops grows',
[VerbType.Modifies]: 'modifies changes updates alters edits revises adjusts adapts',
[VerbType.Consumes]: 'consumes uses utilizes depletes expends absorbs takes processes',
// Ownership/Attribution Types
[VerbType.Owns]: 'owns possesses holds controls manages administers governs maintains',
[VerbType.AttributedTo]: 'attributed credited assigned ascribed authored written composed',
[VerbType.CreatedBy]: 'created made produced generated built developed authored written',
[VerbType.BelongsTo]: 'belongs property possession part member affiliate associated owned',
// Social/Organizational Types
[VerbType.MemberOf]: 'member participant affiliate associate belongs joined enrolled registered',
[VerbType.WorksWith]: 'works collaborates cooperates partners teams assists helps supports',
[VerbType.FriendOf]: 'friend companion buddy pal acquaintance associate connection relationship',
[VerbType.Follows]: 'follows subscribes tracks monitors watches observes trails pursues',
[VerbType.Likes]: 'likes enjoys appreciates favors prefers admires values endorses',
[VerbType.ReportsTo]: 'reports answers subordinate accountable responsible supervised managed',
[VerbType.Supervises]: 'supervises manages oversees directs leads controls guides administers',
[VerbType.Mentors]: 'mentors teaches guides coaches instructs trains advises counsels',
[VerbType.Communicates]: 'communicates talks speaks messages contacts interacts corresponds exchanges',
// Descriptive/Functional Types
[VerbType.Describes]: 'describes explains details documents specifies outlines depicts characterizes',
[VerbType.Defines]: 'defines specifies establishes determines sets declares identifies designates',
[VerbType.Categorizes]: 'categorizes classifies groups sorts organizes arranges labels tags',
[VerbType.Measures]: 'measures quantifies gauges assesses evaluates calculates determines counts',
[VerbType.Evaluates]: 'evaluates assesses analyzes reviews examines appraises judges rates',
[VerbType.Uses]: 'uses utilizes employs applies operates handles manipulates exploits',
[VerbType.Implements]: 'implements executes realizes performs accomplishes carries delivers completes',
[VerbType.Extends]: 'extends expands enhances augments amplifies broadens enlarges develops',
// Enhanced Relationships
[VerbType.Inherits]: 'inherits derives extends receives obtains acquires succeeds legacy',
[VerbType.Conflicts]: 'conflicts contradicts opposes clashes disputes disagrees incompatible inconsistent',
[VerbType.Synchronizes]: 'synchronizes coordinates aligns harmonizes matches corresponds parallels coincides',
[VerbType.Competes]: 'competes rivals contends contests challenges opposes vies struggles'
}
async function buildTypeEmbeddings() {
console.log('🧠 Building embedded type embeddings for Brainy...')
// Count types
const nounTypes = Object.keys(NOUN_TYPE_DESCRIPTIONS)
const verbTypes = Object.keys(VERB_TYPE_DESCRIPTIONS)
console.log(`📊 Processing ${nounTypes.length} noun types and ${verbTypes.length} verb types...`)
// Initialize TransformerEmbedding for embedding (one-time only!)
const embedder = new TransformerEmbedding({
verbose: true,
localFilesOnly: false // Allow downloading models during build
})
await embedder.init()
console.log('✅ TransformerEmbedding initialized')
// Generate noun type embeddings
const nounEmbeddings = new Map<string, number[]>()
console.log('📝 Generating noun type embeddings...')
for (const [type, description] of Object.entries(NOUN_TYPE_DESCRIPTIONS)) {
try {
const embedding = await embedder.embed(description)
if (embedding && Array.isArray(embedding)) {
nounEmbeddings.set(type, embedding)
console.log(`${type}`)
}
} catch (error) {
console.warn(` ⚠️ Failed to embed noun type: ${type}`)
}
}
// Generate verb type embeddings
const verbEmbeddings = new Map<string, number[]>()
console.log('📝 Generating verb type embeddings...')
for (const [type, description] of Object.entries(VERB_TYPE_DESCRIPTIONS)) {
try {
const embedding = await embedder.embed(description)
if (embedding && Array.isArray(embedding)) {
verbEmbeddings.set(type, embedding)
console.log(`${type}`)
}
} catch (error) {
console.warn(` ⚠️ Failed to embed verb type: ${type}`)
}
}
console.log(`✅ Generated ${nounEmbeddings.size} noun embeddings and ${verbEmbeddings.size} verb embeddings`)
// Get embedding dimension
const embeddingDim = nounEmbeddings.size > 0 ?
Array.from(nounEmbeddings.values())[0]?.length ?? 384 :
384
// Convert to compact binary format
const totalTypes = nounTypes.length + verbTypes.length
const totalFloats = totalTypes * embeddingDim
const buffer = new ArrayBuffer(totalFloats * 4)
const view = new DataView(buffer)
let offset = 0
// Pack noun embeddings
for (const type of nounTypes) {
const embedding = nounEmbeddings.get(type) || new Array(embeddingDim).fill(0)
for (let i = 0; i < embeddingDim; i++) {
view.setFloat32(offset, embedding[i], true) // little-endian
offset += 4
}
}
// Pack verb embeddings
for (const type of verbTypes) {
const embedding = verbEmbeddings.get(type) || new Array(embeddingDim).fill(0)
for (let i = 0; i < embeddingDim; i++) {
view.setFloat32(offset, embedding[i], true) // little-endian
offset += 4
}
}
// Convert to base64
const uint8 = new Uint8Array(buffer)
const base64 = Buffer.from(uint8).toString('base64')
// Generate TypeScript file
const tsContent = `/**
* 🧠 BRAINY EMBEDDED TYPE EMBEDDINGS
*
* AUTO-GENERATED - DO NOT EDIT
* Generated: ${new Date().toISOString()}
* Noun Types: ${nounTypes.length}
* Verb Types: ${verbTypes.length}
*
* This file contains pre-computed embeddings for all NounTypes and VerbTypes.
* No runtime computation needed, instant availability!
*/
import { NounType, VerbType } from '../types/graphTypes.js'
import { Vector } from '../coreTypes.js'
// Type metadata
export const TYPE_METADATA = {
nounTypes: ${nounTypes.length},
verbTypes: ${verbTypes.length},
totalTypes: ${totalTypes},
embeddingDimensions: ${embeddingDim},
generatedAt: "${new Date().toISOString()}",
sizeBytes: {
embeddings: ${buffer.byteLength},
base64: ${base64.length}
}
}
// All noun types in order
const NOUN_TYPE_ORDER: NounType[] = ${JSON.stringify(nounTypes)}
// All verb types in order
const VERB_TYPE_ORDER: VerbType[] = ${JSON.stringify(verbTypes)}
// Pre-computed embeddings (${(base64.length / 1024).toFixed(1)}KB base64)
const EMBEDDINGS_BASE64 = "${base64}"
// Decode embeddings at startup (happens once, <10ms)
function decodeEmbeddings(): Uint8Array {
if (typeof Buffer !== 'undefined') {
// Node.js environment
return Buffer.from(EMBEDDINGS_BASE64, 'base64')
} else if (typeof atob !== 'undefined') {
// Browser environment
const binaryString = atob(EMBEDDINGS_BASE64)
const bytes = new Uint8Array(binaryString.length)
for (let i = 0; i < binaryString.length; i++) {
bytes[i] = binaryString.charCodeAt(i)
}
return bytes
}
return new Uint8Array(0)
}
// Cached decoded embeddings
let decodedEmbeddings: Uint8Array | null = null
/**
* Get noun type embeddings as a Map for fast lookup
* This is called once and cached
*/
export function getNounTypeEmbeddings(): Map<NounType, Vector> {
if (!decodedEmbeddings) {
decodedEmbeddings = decodeEmbeddings()
}
const embeddings = new Map<NounType, Vector>()
const view = new DataView(decodedEmbeddings.buffer)
const embeddingSize = ${embeddingDim}
NOUN_TYPE_ORDER.forEach((type, index) => {
const offset = index * embeddingSize * 4
const embedding = new Float32Array(embeddingSize)
for (let i = 0; i < embeddingSize; i++) {
embedding[i] = view.getFloat32(offset + i * 4, true)
}
embeddings.set(type, Array.from(embedding))
})
return embeddings
}
/**
* Get verb type embeddings as a Map for fast lookup
* This is called once and cached
*/
export function getVerbTypeEmbeddings(): Map<VerbType, Vector> {
if (!decodedEmbeddings) {
decodedEmbeddings = decodeEmbeddings()
}
const embeddings = new Map<VerbType, Vector>()
const view = new DataView(decodedEmbeddings.buffer)
const embeddingSize = ${embeddingDim}
// Verb embeddings start after noun embeddings
const verbStartOffset = ${nounTypes.length} * embeddingSize * 4
VERB_TYPE_ORDER.forEach((type, index) => {
const offset = verbStartOffset + index * embeddingSize * 4
const embedding = new Float32Array(embeddingSize)
for (let i = 0; i < embeddingSize; i++) {
embedding[i] = view.getFloat32(offset + i * 4, true)
}
embeddings.set(type, Array.from(embedding))
})
return embeddings
}
// Import logging
import { prodLog } from '../utils/logger.js'
prodLog.info(\`🧠 Brainy Type Embeddings loaded: \${TYPE_METADATA.nounTypes} nouns, \${TYPE_METADATA.verbTypes} verbs, \${(TYPE_METADATA.sizeBytes.embeddings / 1024).toFixed(1)}KB\`)
`
// Write the TypeScript file
const outputPath = path.join(__dirname, '..', 'src', 'neural', 'embeddedTypeEmbeddings.ts')
await fs.writeFile(outputPath, tsContent)
// Report statistics
console.log(`
EMBEDDED TYPE EMBEDDINGS BUILT SUCCESSFULLY!
================================================
Noun Types: ${nounTypes.length}
Verb Types: ${verbTypes.length}
Total Types: ${totalTypes}
Embedding Dimensions: ${embeddingDim}
File sizes:
Embeddings binary: ${(buffer.byteLength / 1024).toFixed(1)} KB
Base64 encoded: ${(base64.length / 1024).toFixed(1)} KB
Output: ${outputPath}
Type embeddings are now embedded directly in Brainy!
No runtime computation needed, instant availability.
`)
}
// Run if called directly
if (import.meta.url === `file://${process.argv[1]}`) {
buildTypeEmbeddings().catch(console.error)
}
export { buildTypeEmbeddings }