fix: implement stub methods in Neural API clustering

Previously, clusterByDomain() and clusterByTime() methods contained
stub implementations that always returned empty arrays. This caused
empty results when attempting domain-based or temporal clustering.

Changes:
- Implement _getItemsByField() to query brain storage
- Implement _getItemsByTimeWindow() to filter by time windows
- Fix _groupByDomain() to check root, metadata, and data fields
- Implement _findCrossDomainMembers() for cross-domain analysis
- Implement _findCrossDomainClusters() to merge similar clusters
- Add comprehensive tests for domain and time clustering
- Update documentation structure to include VFS guides

The methods now properly query the brain's storage, filter results,
and return functional clustering data.
This commit is contained in:
David Snelling 2025-10-07 13:53:41 -07:00
parent df13c196be
commit 1d2da823ed
3 changed files with 555 additions and 12 deletions

View file

@ -104,6 +104,11 @@ docs/
│ ├── noun-verb-taxonomy.md # Data model
│ ├── triple-intelligence.md # Query system
│ └── storage.md # Storage layer
├── vfs/ # Virtual Filesystem
│ ├── README.md # VFS overview
│ ├── SEMANTIC_VFS.md # Semantic projections
│ ├── VFS_API_GUIDE.md # Complete API reference
│ └── QUICK_START.md # 5-minute setup
└── api/ # API documentation
├── README.md # API overview
├── brainy-data.md # Main class

View file

@ -2414,8 +2414,70 @@ export class ImprovedNeuralAPI {
}
private async _getItemsByField(field: string): Promise<any[]> {
// Implementation would query items by metadata field
return []
try {
// Query all items from brain (limit to reasonable number for clustering)
const result = await this.brain.find({
query: '',
limit: 10000 // Max items for clustering
})
if (!result || !Array.isArray(result)) {
return []
}
// Filter items that have the specified field (check both root level and metadata)
const itemsWithField = result.filter((item: any) => {
if (!item || !item.entity) return false
const entity = item.entity
// Check root level fields first (e.g., 'noun' for type)
if (field === 'type' || field === 'nounType') {
return entity.noun != null
}
// Check if field exists at root level
if (entity[field] != null) {
return true
}
// Check if field exists in metadata/data
if (entity.metadata?.[field] != null) {
return true
}
if (entity.data?.[field] != null) {
return true
}
return false
})
// Map to format expected by clustering methods
return itemsWithField.map((item: any) => {
const entity = item.entity
return {
id: entity.id,
vector: entity.embedding || entity.vector || [],
metadata: {
...(entity.metadata || {}),
...(entity.data || {}),
// Include root-level fields in metadata for easy access
noun: entity.noun,
type: entity.noun,
createdAt: entity.createdAt,
updatedAt: entity.updatedAt,
label: entity.label
},
nounType: entity.noun,
label: entity.label || entity.data || '',
data: entity.data
}
})
} catch (error) {
console.error('Error in _getItemsByField:', error)
return []
}
}
// ===== TRIPLE INTELLIGENCE INTEGRATION =====
@ -2875,11 +2937,39 @@ export class ImprovedNeuralAPI {
private _groupByDomain(items: any[], field: string): Map<string, any[]> {
const groups = new Map()
for (const item of items) {
const domain = item.metadata?.[field] || 'unknown'
if (!groups.has(domain)) {
groups.set(domain, [])
// Check multiple locations for the field value
let domain: any = 'unknown'
// Special handling for type/nounType field
if (field === 'type' || field === 'nounType') {
domain = item.nounType || item.metadata?.noun || item.metadata?.type || 'unknown'
} else {
// Check root level first
domain = item[field]
// Then check metadata
if (domain == null) {
domain = item.metadata?.[field]
}
// Then check data
if (domain == null) {
domain = item.data?.[field]
}
// Fallback to unknown
if (domain == null) {
domain = 'unknown'
}
}
groups.get(domain).push(item)
// Convert domain to string for Map key
const domainKey = String(domain)
if (!groups.has(domainKey)) {
groups.set(domainKey, [])
}
groups.get(domainKey).push(item)
}
return groups
}
@ -2902,18 +2992,203 @@ export class ImprovedNeuralAPI {
}
private async _findCrossDomainMembers(cluster: SemanticCluster, threshold: number): Promise<string[]> {
// Find members that might belong to multiple domains
return []
try {
// Find cluster members that have high similarity to items in other domains
const crossDomainMembers: string[] = []
for (const memberId of cluster.members) {
try {
// Get neighbors for this member
const neighbors = await this.neighbors(memberId, {
limit: 10,
minSimilarity: threshold
})
if (Array.isArray(neighbors) && neighbors.length > 0) {
// Check if any neighbors are NOT in this cluster
const hasExternalNeighbors = neighbors.some(neighbor =>
!cluster.members.includes(typeof neighbor === 'object' ? neighbor.id : neighbor)
)
if (hasExternalNeighbors) {
crossDomainMembers.push(memberId)
}
}
} catch (error) {
// Skip members that can't be processed
continue
}
}
return crossDomainMembers
} catch (error) {
console.error('Error in _findCrossDomainMembers:', error)
return []
}
}
private async _findCrossDomainClusters(clusters: DomainCluster[], threshold: number): Promise<DomainCluster[]> {
// Find clusters that span multiple domains
return []
try {
const crossDomainClusters: DomainCluster[] = []
// Group clusters by domain
const domainMap = new Map<string, DomainCluster[]>()
for (const cluster of clusters) {
const domain = cluster.domain || 'unknown'
if (!domainMap.has(domain)) {
domainMap.set(domain, [])
}
domainMap.get(domain)!.push(cluster)
}
// Find clusters with high inter-domain similarity
const domains = Array.from(domainMap.keys())
for (let i = 0; i < domains.length; i++) {
for (let j = i + 1; j < domains.length; j++) {
const domain1 = domains[i]
const domain2 = domains[j]
const clusters1 = domainMap.get(domain1)!
const clusters2 = domainMap.get(domain2)!
// Compare clusters between domains
for (const c1 of clusters1) {
for (const c2 of clusters2) {
try {
// Calculate similarity between cluster centroids
if (!c1.centroid || !c2.centroid || c1.centroid.length === 0 || c2.centroid.length === 0) {
continue
}
const similarity = 1 - cosineDistance(
Array.from(c1.centroid) as number[],
Array.from(c2.centroid) as number[]
)
if (similarity >= threshold) {
// Create a cross-domain cluster
const mergedMembers = [...new Set([...c1.members, ...c2.members])]
const mergedCentroid = this._averageVectors([
Array.from(c1.centroid) as number[],
Array.from(c2.centroid) as number[]
])
crossDomainClusters.push({
...c1,
id: `cross-${domain1}-${domain2}-${crossDomainClusters.length}`,
label: `Cross-domain: ${c1.label} + ${c2.label}`,
members: mergedMembers,
centroid: mergedCentroid,
domain: `${domain1}+${domain2}`,
domainConfidence: similarity,
crossDomainMembers: mergedMembers
})
}
} catch (error) {
// Skip cluster pairs that can't be compared
continue
}
}
}
}
}
return crossDomainClusters
} catch (error) {
console.error('Error in _findCrossDomainClusters:', error)
return []
}
}
private _averageVectors(vectors: number[][]): number[] {
if (vectors.length === 0) return []
if (vectors.length === 1) return [...vectors[0]]
const dim = vectors[0].length
const result = new Array(dim).fill(0)
for (const vector of vectors) {
for (let i = 0; i < dim; i++) {
result[i] += vector[i]
}
}
for (let i = 0; i < dim; i++) {
result[i] /= vectors.length
}
return result
}
private async _getItemsByTimeWindow(timeField: string, window: TimeWindow): Promise<any[]> {
// Implementation would query items within time window
return []
try {
// Query all items from brain
const result = await this.brain.find({
query: '',
limit: 10000 // Max items for clustering
})
if (!result || !Array.isArray(result)) {
return []
}
// Filter items within the time window
const itemsInWindow = result.filter((item: any) => {
if (!item || !item.entity) return false
const entity = item.entity
// Get timestamp value from various possible locations
let timestamp: any = null
// Check root level first
if (timeField === 'createdAt' || timeField === 'updatedAt') {
timestamp = entity[timeField]
}
// Check metadata/data
if (timestamp == null) {
timestamp = entity.metadata?.[timeField] || entity.data?.[timeField]
}
if (timestamp == null) {
return false
}
// Convert to Date if needed
const itemDate = timestamp instanceof Date ? timestamp : new Date(timestamp)
if (isNaN(itemDate.getTime())) {
return false // Invalid date
}
// Check if item falls within window
return itemDate >= window.start && itemDate <= window.end
})
// Map to format expected by clustering methods
return itemsInWindow.map((item: any) => {
const entity = item.entity
return {
id: entity.id,
vector: entity.embedding || entity.vector || [],
metadata: {
...(entity.metadata || {}),
...(entity.data || {}),
noun: entity.noun,
type: entity.noun,
createdAt: entity.createdAt,
updatedAt: entity.updatedAt,
label: entity.label
},
nounType: entity.noun,
label: entity.label || entity.data || '',
data: entity.data
}
})
} catch (error) {
console.error('Error in _getItemsByTimeWindow:', error)
return []
}
}
private async _calculateTemporalMetrics(cluster: SemanticCluster, items: any[], timeField: string): Promise<any> {

View file

@ -0,0 +1,263 @@
/**
* Domain and Time Clustering Tests
*
* Tests for clusterByDomain() and clusterByTime() methods
* that were previously stub implementations.
*/
import { describe, it, expect, beforeEach } from 'vitest'
import { Brainy } from '../../../src/brainy'
import { NounType } from '../../../src/types/graphTypes'
import { createAddParams } from '../../helpers/test-factory'
describe('Domain and Time Clustering', () => {
let brain: Brainy
beforeEach(async () => {
brain = new Brainy({ enableCache: false })
await brain.init()
})
describe('clusterByDomain() - Field-based clustering', () => {
it('should cluster entities by type field', async () => {
// Add entities of different types
await brain.add(createAddParams({
data: 'John Smith is a person',
type: NounType.Person
}))
await brain.add(createAddParams({
data: 'Jane Doe is also a person',
type: NounType.Person
}))
await brain.add(createAddParams({
data: 'Technical document about AI',
type: NounType.Document
}))
await brain.add(createAddParams({
data: 'Research paper on machine learning',
type: NounType.Document
}))
await brain.add(createAddParams({
data: 'Microsoft Corporation',
type: NounType.Organization
}))
// Cluster by type field
const clusters = await brain.neural().clusterByDomain('type', {
minClusterSize: 1,
maxClusters: 10
})
// Should have clusters for each type
expect(Array.isArray(clusters)).toBe(true)
expect(clusters.length).toBeGreaterThan(0)
// Verify domain values exist
const domains = new Set(clusters.map(c => c.domain))
expect(domains.has(NounType.Person) || domains.has('person')).toBe(true)
expect(domains.has(NounType.Document) || domains.has('document')).toBe(true)
})
it('should cluster entities by metadata field', async () => {
// Add entities with category metadata
await brain.add(createAddParams({
data: 'JavaScript programming guide',
type: NounType.Document,
metadata: { category: 'programming' }
}))
await brain.add(createAddParams({
data: 'Python tutorial',
type: NounType.Document,
metadata: { category: 'programming' }
}))
await brain.add(createAddParams({
data: 'Chocolate cake recipe',
type: NounType.Document,
metadata: { category: 'cooking' }
}))
await brain.add(createAddParams({
data: 'Pasta preparation',
type: NounType.Document,
metadata: { category: 'cooking' }
}))
// Cluster by category field
const clusters = await brain.neural().clusterByDomain('category', {
minClusterSize: 1,
maxClusters: 5
})
expect(Array.isArray(clusters)).toBe(true)
expect(clusters.length).toBeGreaterThan(0)
// Verify categories are in domains
const domains = new Set(clusters.map(c => c.domain))
expect(domains.has('programming')).toBe(true)
expect(domains.has('cooking')).toBe(true)
})
it('should handle entities without the specified field', async () => {
// Add entities with and without category
await brain.add(createAddParams({
data: 'Has category',
metadata: { category: 'tech' }
}))
await brain.add(createAddParams({
data: 'No category'
}))
const clusters = await brain.neural().clusterByDomain('category', {
minClusterSize: 1
})
expect(Array.isArray(clusters)).toBe(true)
// Should have 'tech' and 'unknown' domains
const domains = new Set(clusters.map(c => c.domain))
expect(domains.has('tech')).toBe(true)
expect(domains.has('unknown')).toBe(true)
})
})
describe('clusterByTime() - Temporal clustering', () => {
it('should cluster entities by time windows', async () => {
const now = new Date()
const oneDayAgo = new Date(now.getTime() - 24 * 60 * 60 * 1000)
const oneWeekAgo = new Date(now.getTime() - 7 * 24 * 60 * 60 * 1000)
const oneMonthAgo = new Date(now.getTime() - 30 * 24 * 60 * 60 * 1000)
// Add entities with different timestamps
await brain.add(createAddParams({
data: 'Recent item 1',
metadata: { publishedAt: now.toISOString() }
}))
await brain.add(createAddParams({
data: 'Recent item 2',
metadata: { publishedAt: oneDayAgo.toISOString() }
}))
await brain.add(createAddParams({
data: 'Old item 1',
metadata: { publishedAt: oneWeekAgo.toISOString() }
}))
await brain.add(createAddParams({
data: 'Very old item',
metadata: { publishedAt: oneMonthAgo.toISOString() }
}))
// Define time windows
const timeWindows = [
{
start: new Date(now.getTime() - 2 * 24 * 60 * 60 * 1000), // Last 2 days
end: now,
label: 'Recent'
},
{
start: new Date(now.getTime() - 14 * 24 * 60 * 60 * 1000), // 2-14 days ago
end: new Date(now.getTime() - 2 * 24 * 60 * 60 * 1000),
label: 'This Week'
},
{
start: new Date(now.getTime() - 60 * 24 * 60 * 60 * 1000), // 14-60 days ago
end: new Date(now.getTime() - 14 * 24 * 60 * 60 * 1000),
label: 'Older'
}
]
// Cluster by time
const clusters = await brain.neural().clusterByTime('publishedAt', timeWindows, {
timeField: 'publishedAt',
windows: timeWindows
})
expect(Array.isArray(clusters)).toBe(true)
expect(clusters.length).toBeGreaterThan(0)
// Verify time windows are represented
const windowLabels = new Set(clusters.map(c => c.timeWindow?.label))
expect(windowLabels.size).toBeGreaterThan(0)
})
it('should cluster entities by createdAt timestamps', async () => {
// These will use the auto-generated createdAt timestamps
const id1 = await brain.add(createAddParams({
data: 'First item'
}))
// Wait a bit to ensure different timestamps
await new Promise(resolve => setTimeout(resolve, 10))
const id2 = await brain.add(createAddParams({
data: 'Second item'
}))
const now = new Date()
const timeWindows = [
{
start: new Date(now.getTime() - 60 * 60 * 1000), // Last hour
end: new Date(now.getTime() + 60 * 60 * 1000), // Next hour (to include all)
label: 'Now'
}
]
const clusters = await brain.neural().clusterByTime('createdAt', timeWindows, {
timeField: 'createdAt',
windows: timeWindows
})
expect(Array.isArray(clusters)).toBe(true)
// Both items should be in the 'Now' time window
const nowCluster = clusters.find(c => c.timeWindow?.label === 'Now')
expect(nowCluster).toBeDefined()
if (nowCluster) {
expect(nowCluster.members.length).toBeGreaterThanOrEqual(2)
}
})
it('should handle empty time windows gracefully', async () => {
const futureStart = new Date(Date.now() + 365 * 24 * 60 * 60 * 1000) // 1 year from now
const futureEnd = new Date(Date.now() + 2 * 365 * 24 * 60 * 60 * 1000) // 2 years from now
const timeWindows = [
{
start: futureStart,
end: futureEnd,
label: 'Future'
}
]
const clusters = await brain.neural().clusterByTime('createdAt', timeWindows, {
timeField: 'createdAt',
windows: timeWindows
})
// Should return empty array or array with empty clusters
expect(Array.isArray(clusters)).toBe(true)
})
})
describe('Cross-domain functionality', () => {
it('should find cross-domain clusters when enabled', async () => {
// Add entities from different domains with similar content
await brain.add(createAddParams({
data: 'Machine learning and artificial intelligence',
type: NounType.Document,
metadata: { category: 'tech' }
}))
await brain.add(createAddParams({
data: 'AI and neural networks',
type: NounType.Concept,
metadata: { category: 'science' }
}))
const clusters = await brain.neural().clusterByDomain('category', {
minClusterSize: 1,
preserveDomainBoundaries: false, // Enable cross-domain clustering
crossDomainThreshold: 0.5
})
expect(Array.isArray(clusters)).toBe(true)
expect(clusters.length).toBeGreaterThan(0)
})
})
})