feat(cluster): implement cluster detection and bridge notes discovery

Add automatic note clustering using density-based algorithm (DBSCAN variant)
and bridge notes detection for connecting different thematic clusters.

Features:
- NoteCluster, ClusterMember, BridgeNote, BridgeSuggestion models
- Clustering service with pgvector cosine similarity
- Bridge notes detection (notes connecting >=2 clusters)
- AI-powered suggestions for missing cluster connections
- /insights page with React Flow visualization
- Cron endpoint for automatic recalculation

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
This commit is contained in:
Antigravity
2026-05-23 20:26:25 +00:00
parent 2aed148dc2
commit 077e665dfc
13 changed files with 2882 additions and 12 deletions

View File

@@ -0,0 +1,85 @@
import { NextRequest, NextResponse } from 'next/server'
import { auth } from '@/auth'
import prisma from '@/lib/prisma'
import { bridgeNotesService } from '@/lib/ai/services/bridge-notes.service'
/**
* GET /api/bridge-notes
* Get all bridge notes for the current user.
*/
export async function GET(request: NextRequest) {
try {
const session = await auth()
if (!session?.user?.id) {
return NextResponse.json({ error: 'Unauthorized' }, { status: 401 })
}
const userId = session.user.id
const { searchParams } = new URL(request.url)
const includeDetails = searchParams.get('details') === 'true'
const bridgeNotes = await bridgeNotesService.getBridgeNotes(userId)
// Optionally include note details
let enrichedNotes = bridgeNotes
if (includeDetails && bridgeNotes.length > 0) {
const noteIds = bridgeNotes.map(b => b.noteId)
const notes = await prisma.note.findMany({
where: { id: { in: noteIds } },
select: { id: true, title: true, content: true, notebookId: true }
})
const noteMap = new Map(notes.map(n => [n.id, n]))
enrichedNotes = bridgeNotes.map(b => ({
...b,
note: noteMap.get(b.noteId)
}))
}
return NextResponse.json({
bridgeNotes: enrichedNotes,
count: enrichedNotes.length
})
} catch (error) {
console.error('Error fetching bridge notes:', error)
return NextResponse.json(
{ error: 'Failed to fetch bridge notes' },
{ status: 500 }
)
}
}
/**
* DELETE /api/bridge-notes
* Dismiss a bridge suggestion.
*/
export async function DELETE(request: NextRequest) {
try {
const session = await auth()
if (!session?.user?.id) {
return NextResponse.json({ error: 'Unauthorized' }, { status: 401 })
}
const userId = session.user.id
const body = await request.json()
const { clusterAId, clusterBId } = body
if (typeof clusterAId !== 'number' || typeof clusterBId !== 'number') {
return NextResponse.json(
{ error: 'Invalid cluster IDs' },
{ status: 400 }
)
}
await bridgeNotesService.dismissSuggestion(userId, clusterAId, clusterBId)
return NextResponse.json({ success: true })
} catch (error) {
console.error('Error dismissing suggestion:', error)
return NextResponse.json(
{ error: 'Failed to dismiss suggestion' },
{ status: 500 }
)
}
}

View File

@@ -0,0 +1,66 @@
import { NextRequest, NextResponse } from 'next/server'
import { auth } from '@/auth'
import { bridgeNotesService } from '@/lib/ai/services/bridge-notes.service'
/**
* GET /api/bridge-notes/suggestions
* Get AI-powered bridge note suggestions for unconnected clusters.
*/
export async function GET(request: NextRequest) {
try {
const session = await auth()
if (!session?.user?.id) {
return NextResponse.json({ error: 'Unauthorized' }, { status: 401 })
}
const userId = session.user.id
const { searchParams } = new URL(request.url)
const includeDismissed = searchParams.get('dismissed') === 'true'
const suggestions = await bridgeNotesService.getBridgeSuggestions(userId, includeDismissed)
return NextResponse.json({
suggestions,
count: suggestions.length
})
} catch (error) {
console.error('Error fetching bridge suggestions:', error)
return NextResponse.json(
{ error: 'Failed to fetch suggestions' },
{ status: 500 }
)
}
}
/**
* POST /api/bridge-notes/suggestions/generate
* Generate new bridge suggestions (force regeneration).
*/
export async function POST(request: NextRequest) {
try {
const session = await auth()
if (!session?.user?.id) {
return NextResponse.json({ error: 'Unauthorized' }, { status: 401 })
}
const userId = session.user.id
// Generate new suggestions
const suggestions = await bridgeNotesService.generateBridgeSuggestions(userId)
// Save to database
await bridgeNotesService.saveBridgeSuggestions(userId, suggestions)
return NextResponse.json({
suggestions,
count: suggestions.length,
message: `Generated ${suggestions.length} bridge suggestions`
})
} catch (error) {
console.error('Error generating bridge suggestions:', error)
return NextResponse.json(
{ error: 'Failed to generate suggestions' },
{ status: 500 }
)
}
}

View File

@@ -0,0 +1,124 @@
import { NextRequest, NextResponse } from 'next/server'
import { auth } from '@/auth'
import prisma from '@/lib/prisma'
import { clusteringService } from '@/lib/ai/services/clustering.service'
import { bridgeNotesService } from '@/lib/ai/services/bridge-notes.service'
/**
* GET /api/clusters
* Get all clusters for the current user.
*/
export async function GET(request: NextRequest) {
try {
const session = await auth()
if (!session?.user?.id) {
return NextResponse.json({ error: 'Unauthorized' }, { status: 401 })
}
const userId = session.user.id
// Check for cached results
const cached = await clusteringService.getCachedClusters(userId)
if (cached) {
return NextResponse.json({
clusters: cached,
cached: true,
totalNotes: cached.reduce((sum, c) => sum + c.noteIds.length, 0)
})
}
// No cached results, check if user has enough notes
const notesCount = await prisma.note.count({
where: { userId, trashedAt: null }
})
if (notesCount < 10) {
return NextResponse.json({
clusters: [],
message: 'Need at least 10 notes to generate clusters',
totalNotes: notesCount
})
}
// Trigger background recalculation
return NextResponse.json({
clusters: [],
message: 'Calculating clusters... Please check back later',
totalNotes: notesCount
})
} catch (error) {
console.error('Error fetching clusters:', error)
return NextResponse.json(
{ error: 'Failed to fetch clusters' },
{ status: 500 }
)
}
}
/**
* POST /api/clusters/recalculate
* Trigger a full recalculation of clusters and bridge notes.
*/
export async function POST(request: NextRequest) {
try {
const session = await auth()
if (!session?.user?.id) {
return NextResponse.json({ error: 'Unauthorized' }, { status: 401 })
}
const userId = session.user.id
const body = await request.json()
const force = body.force === true
// Check if recalculation is needed
const shouldRecalc = force || await clusteringService.shouldRecalculate(userId)
if (!shouldRecalc) {
const cached = await clusteringService.getCachedClusters(userId)
if (cached) {
return NextResponse.json({
clusters: cached,
cached: true,
message: 'Using cached results (data has not changed significantly)'
})
}
}
// Perform clustering
const results = await clusteringService.clusterNotes(userId)
if (results.clusters.length === 0) {
return NextResponse.json({
clusters: [],
message: 'Could not generate clusters. Need more diverse notes.',
noiseCount: results.noiseCount
})
}
// Generate cluster names
for (const cluster of results.clusters) {
cluster.name = await clusteringService.generateClusterName(cluster.clusterId, userId)
}
// Save results
await clusteringService.saveClusteringResults(userId, results)
// Detect and save bridge notes
const bridgeNotes = await bridgeNotesService.detectBridgeNotes(userId)
await bridgeNotesService.saveBridgeNotes(userId, bridgeNotes)
return NextResponse.json({
clusters: results.clusters,
bridgeNotes: bridgeNotes.slice(0, 10), // Return top 10
totalNotes: results.clusters.reduce((sum, c) => sum + c.noteIds.length, 0) + results.noiseCount,
noiseCount: results.noiseCount,
message: `Generated ${results.clusters.length} clusters`
})
} catch (error) {
console.error('Error recalculating clusters:', error)
return NextResponse.json(
{ error: 'Failed to recalculate clusters' },
{ status: 500 }
)
}
}

View File

@@ -0,0 +1,224 @@
import { NextRequest, NextResponse } from 'next/server'
import { auth } from '@/auth'
import prisma from '@/lib/prisma'
import { clusteringService } from '@/lib/ai/services/clustering.service'
import { bridgeNotesService } from '@/lib/ai/services/bridge-notes.service'
/**
* Cron endpoint for automatic cluster recalculation.
*
* This endpoint is designed to be called by a cron job or scheduler.
* It processes users who need cluster updates based on:
* 1. Data change percentage (> 5%)
* 2. Time since last calculation (> 24 hours)
*
* GET /api/cron/clusters?secret=CRON_SECRET
*
* The secret parameter must match the environment variable CRON_SECRET.
*/
export async function GET(request: NextRequest) {
try {
// Verify cron secret
const { searchParams } = new URL(request.url)
const secret = searchParams.get('secret')
if (secret !== process.env.CRON_SECRET) {
return NextResponse.json({ error: 'Unauthorized' }, { status: 401 })
}
// Find users who need recalculation
// Users who either:
// 1. Have no clusters calculated yet
// 2. Have data changes > 5% since last calculation
// 3. Haven't had calculations in > 24 hours
const oneDayAgo = new Date(Date.now() - 24 * 60 * 60 * 1000)
const usersToProcess = await prisma.user.findMany({
where: {
// Only active users
OR: [
// Users who never had clusters calculated
{
noteClusters: {
none: {}
}
},
// Users with stale cluster data
{
noteClusters: {
some: {
lastCalculated: {
lt: oneDayAgo
}
}
}
}
]
},
select: {
id: true,
email: true
},
take: 50 // Process up to 50 users per run
})
const results = {
processed: 0,
skipped: 0,
errors: 0,
details: [] as Array<{ userId: string; clusters: number; bridges: number; error?: string }>
}
for (const user of usersToProcess) {
try {
// Check if recalculation is needed based on data change
const shouldRecalc = await clusteringService.shouldRecalculate(user.id)
if (!shouldRecalc) {
results.skipped++
continue
}
// Perform clustering
const clusterResults = await clusteringService.clusterNotes(user.id)
if (clusterResults.clusters.length === 0) {
results.details.push({
userId: user.id,
clusters: 0,
bridges: 0,
error: 'Could not generate clusters'
})
continue
}
// Generate cluster names
for (const cluster of clusterResults.clusters) {
cluster.name = await clusteringService.generateClusterName(cluster.clusterId, user.id)
}
// Save results
await clusteringService.saveClusteringResults(user.id, clusterResults)
// Detect and save bridge notes
const bridgeNotes = await bridgeNotesService.detectBridgeNotes(user.id)
await bridgeNotesService.saveBridgeNotes(user.id, bridgeNotes)
results.processed++
results.details.push({
userId: user.id,
clusters: clusterResults.clusters.length,
bridges: bridgeNotes.length
})
} catch (error) {
console.error(`Error processing user ${user.id}:`, error)
results.errors++
results.details.push({
userId: user.id,
clusters: 0,
bridges: 0,
error: error instanceof Error ? error.message : 'Unknown error'
})
}
}
return NextResponse.json({
success: true,
...results,
message: `Processed ${results.processed} users, skipped ${results.skipped}, ${results.errors} errors`
})
} catch (error) {
console.error('Error in cluster cron job:', error)
return NextResponse.json(
{ error: 'Cron job failed', details: error instanceof Error ? error.message : 'Unknown error' },
{ status: 500 }
)
}
}
/**
* POST /api/cron/clusters
* Manual trigger for cluster recalculation (admin only).
* Requires admin authentication.
*/
export async function POST(request: NextRequest) {
try {
const session = await auth()
if (!session?.user?.id) {
return NextResponse.json({ error: 'Unauthorized' }, { status: 401 })
}
// Check if user is admin
const user = await prisma.user.findUnique({
where: { id: session.user.id },
select: { role: true }
})
if (user?.role !== 'ADMIN') {
return NextResponse.json({ error: 'Forbidden' }, { status: 403 })
}
const body = await request.json()
const { userId, forceAll } = body
if (userId) {
// Process specific user
const clusterResults = await clusteringService.clusterNotes(userId)
const bridgeNotes = await bridgeNotesService.detectBridgeNotes(userId)
await clusteringService.saveClusteringResults(userId, clusterResults)
await bridgeNotesService.saveBridgeNotes(userId, bridgeNotes)
return NextResponse.json({
success: true,
userId,
clusters: clusterResults.clusters.length,
bridges: bridgeNotes.length
})
}
if (forceAll) {
// Process all users (use with caution)
const users = await prisma.user.findMany({
select: { id: true },
take: 100 // Limit to 100 users per request
})
const results = {
processed: 0,
errors: 0
}
for (const user of users) {
try {
const clusterResults = await clusteringService.clusterNotes(user.id)
const bridgeNotes = await bridgeNotesService.detectBridgeNotes(user.id)
await clusteringService.saveClusteringResults(user.id, clusterResults)
await bridgeNotesService.saveBridgeNotes(user.id, bridgeNotes)
results.processed++
} catch {
results.errors++
}
}
return NextResponse.json({
success: true,
...results
})
}
return NextResponse.json(
{ error: 'Must provide userId or forceAll=true' },
{ status: 400 }
)
} catch (error) {
console.error('Error in manual cluster trigger:', error)
return NextResponse.json(
{ error: 'Failed to trigger recalculation' },
{ status: 500 }
)
}
}