feat: Add core memory clustering visualization and hooks configuration

- Implemented core memory clustering visualization in core-memory-clusters.js
- Added functions for loading, rendering, and managing clusters and their members
- Created example hooks configuration in hooks-config-example.json for session management
- Developed test script for hooks integration in test-hooks.js
- Included error handling and notifications for user interactions
This commit is contained in:
catlog22
2025-12-18 23:06:58 +08:00
parent 68f9de0c69
commit 9f6e6852da
24 changed files with 4543 additions and 590 deletions

View File

@@ -164,15 +164,58 @@ export class CoreMemoryStore {
}
/**
* Migrate database by removing old tables
* Migrate database by removing old tables, views, and triggers
*/
private migrateDatabase(): void {
const oldTables = ['knowledge_graph', 'knowledge_graph_edges', 'evolution_history'];
for (const table of oldTables) {
try {
// Disable foreign key constraints during migration
this.db.pragma('foreign_keys = OFF');
// Drop any triggers that might reference old tables
const triggers = this.db.prepare(
`SELECT name FROM sqlite_master WHERE type='trigger'`
).all() as { name: string }[];
for (const trigger of triggers) {
try {
this.db.exec(`DROP TRIGGER IF EXISTS "${trigger.name}"`);
} catch (e) {
// Ignore trigger drop errors
}
}
// Drop any views that might reference old tables
const views = this.db.prepare(
`SELECT name FROM sqlite_master WHERE type='view'`
).all() as { name: string }[];
for (const view of views) {
try {
this.db.exec(`DROP VIEW IF EXISTS "${view.name}"`);
} catch (e) {
// Ignore view drop errors
}
}
// Now drop the old tables
for (const table of oldTables) {
try {
this.db.exec(`DROP TABLE IF EXISTS "${table}"`);
} catch (e) {
// Ignore if table doesn't exist
}
}
// Re-enable foreign key constraints
this.db.pragma('foreign_keys = ON');
} catch (e) {
// If migration fails, continue - tables may not exist
try {
this.db.exec(`DROP TABLE IF EXISTS ${table}`);
} catch (e) {
// Ignore if table doesn't exist
this.db.pragma('foreign_keys = ON');
} catch (_) {
// Ignore
}
}
}
@@ -202,7 +245,10 @@ export class CoreMemoryStore {
const hours = String(now.getHours()).padStart(2, '0');
const minutes = String(now.getMinutes()).padStart(2, '0');
const seconds = String(now.getSeconds()).padStart(2, '0');
return `CLST-${year}${month}${day}-${hours}${minutes}${seconds}`;
const ms = String(now.getMilliseconds()).padStart(3, '0');
// Add random 2-digit suffix to ensure uniqueness
const random = String(Math.floor(Math.random() * 100)).padStart(2, '0');
return `CLST-${year}${month}${day}-${hours}${minutes}${seconds}${ms}${random}`;
}
/**

View File

@@ -90,6 +90,7 @@ const MODULE_FILES = [
'views/memory.js',
'views/core-memory.js',
'views/core-memory-graph.js',
'views/core-memory-clusters.js',
'views/prompt-history.js',
'views/skills-manager.js',
'views/rules-manager.js',

View File

@@ -77,7 +77,7 @@ export async function handleCodexLensRoutes(ctx: RouteContext): Promise<boolean>
// API: CodexLens Index List - Get all indexed projects with details
if (pathname === '/api/codexlens/indexes') {
try {
// First get config to find index directory
// Get config for index directory path
const configResult = await executeCodexLens(['config', '--json']);
let indexDir = '';
@@ -85,109 +85,127 @@ export async function handleCodexLensRoutes(ctx: RouteContext): Promise<boolean>
try {
const config = extractJSON(configResult.output);
if (config.success && config.result) {
indexDir = config.result.index_root || '';
// CLI returns index_dir (not index_root)
indexDir = config.result.index_dir || config.result.index_root || '';
}
} catch (e) {
console.error('[CodexLens] Failed to parse config for index list:', e.message);
}
}
// Get detailed status including projects
const statusResult = await executeCodexLens(['status', '--json']);
// Get project list using 'projects list' command
const projectsResult = await executeCodexLens(['projects', 'list', '--json']);
let indexes: any[] = [];
let totalSize = 0;
let vectorIndexCount = 0;
let normalIndexCount = 0;
if (projectsResult.success) {
try {
const projectsData = extractJSON(projectsResult.output);
if (projectsData.success && Array.isArray(projectsData.result)) {
const { statSync, existsSync } = await import('fs');
const { basename, join } = await import('path');
for (const project of projectsData.result) {
// Skip test/temp projects
if (project.source_root && (
project.source_root.includes('\\Temp\\') ||
project.source_root.includes('/tmp/') ||
project.total_files === 0
)) {
continue;
}
let projectSize = 0;
let hasVectorIndex = false;
let hasNormalIndex = true; // All projects have FTS index
let lastModified = null;
// Try to get actual index size from index_root
if (project.index_root && existsSync(project.index_root)) {
try {
const { readdirSync } = await import('fs');
const files = readdirSync(project.index_root);
for (const file of files) {
try {
const filePath = join(project.index_root, file);
const stat = statSync(filePath);
projectSize += stat.size;
if (!lastModified || stat.mtime > lastModified) {
lastModified = stat.mtime;
}
// Check for vector/embedding files
if (file.includes('vector') || file.includes('embedding') ||
file.endsWith('.faiss') || file.endsWith('.npy') ||
file.includes('semantic_chunks')) {
hasVectorIndex = true;
}
} catch (e) {
// Skip files we can't stat
}
}
} catch (e) {
// Can't read index directory
}
}
if (hasVectorIndex) vectorIndexCount++;
if (hasNormalIndex) normalIndexCount++;
totalSize += projectSize;
// Use source_root as the display name
const displayName = project.source_root ? basename(project.source_root) : `project_${project.id}`;
indexes.push({
id: displayName,
path: project.source_root || '',
indexPath: project.index_root || '',
size: projectSize,
sizeFormatted: formatSize(projectSize),
fileCount: project.total_files || 0,
dirCount: project.total_dirs || 0,
hasVectorIndex,
hasNormalIndex,
status: project.status || 'active',
lastModified: lastModified ? lastModified.toISOString() : null
});
}
// Sort by file count (most files first), then by name
indexes.sort((a, b) => {
if (b.fileCount !== a.fileCount) return b.fileCount - a.fileCount;
return a.id.localeCompare(b.id);
});
}
} catch (e) {
console.error('[CodexLens] Failed to parse projects list:', e.message);
}
}
// Also get summary stats from status command
const statusResult = await executeCodexLens(['status', '--json']);
let statusSummary: any = {};
if (statusResult.success) {
try {
const status = extractJSON(statusResult.output);
if (status.success && status.result) {
const projectsCount = status.result.projects_count || 0;
// Try to get project list from index directory
if (indexDir) {
const { readdirSync, statSync, existsSync } = await import('fs');
const { join } = await import('path');
const { homedir } = await import('os');
// Expand ~ in path
const expandedDir = indexDir.startsWith('~')
? join(homedir(), indexDir.slice(1))
: indexDir;
if (existsSync(expandedDir)) {
try {
const entries = readdirSync(expandedDir, { withFileTypes: true });
for (const entry of entries) {
if (entry.isDirectory()) {
const projectDir = join(expandedDir, entry.name);
let projectSize = 0;
let hasVectorIndex = false;
let hasNormalIndex = false;
let fileCount = 0;
let lastModified = null;
try {
// Check for index files
const projectFiles = readdirSync(projectDir);
for (const file of projectFiles) {
const filePath = join(projectDir, file);
try {
const stat = statSync(filePath);
projectSize += stat.size;
fileCount++;
if (!lastModified || stat.mtime > lastModified) {
lastModified = stat.mtime;
}
// Check index type
if (file.includes('vector') || file.includes('embedding') || file.endsWith('.faiss') || file.endsWith('.npy')) {
hasVectorIndex = true;
}
if (file.includes('fts') || file.endsWith('.db') || file.endsWith('.sqlite')) {
hasNormalIndex = true;
}
} catch (e) {
// Skip files we can't stat
}
}
} catch (e) {
// Can't read project directory
}
if (hasVectorIndex) vectorIndexCount++;
if (hasNormalIndex) normalIndexCount++;
totalSize += projectSize;
indexes.push({
id: entry.name,
path: projectDir,
size: projectSize,
sizeFormatted: formatSize(projectSize),
fileCount,
hasVectorIndex,
hasNormalIndex,
lastModified: lastModified ? lastModified.toISOString() : null
});
}
}
// Sort by last modified (most recent first)
indexes.sort((a, b) => {
if (!a.lastModified) return 1;
if (!b.lastModified) return -1;
return new Date(b.lastModified).getTime() - new Date(a.lastModified).getTime();
});
} catch (e) {
console.error('[CodexLens] Failed to read index directory:', e.message);
}
}
statusSummary = {
totalProjects: status.result.projects_count || indexes.length,
totalFiles: status.result.total_files || 0,
totalDirs: status.result.total_dirs || 0,
indexSizeBytes: status.result.index_size_bytes || totalSize,
indexSizeMb: status.result.index_size_mb || 0,
embeddings: status.result.embeddings || {}
};
// Use status total size if available
if (status.result.index_size_bytes) {
totalSize = status.result.index_size_bytes;
}
}
} catch (e) {
console.error('[CodexLens] Failed to parse status for index list:', e.message);
console.error('[CodexLens] Failed to parse status:', e.message);
}
}
@@ -201,7 +219,8 @@ export async function handleCodexLensRoutes(ctx: RouteContext): Promise<boolean>
totalSize,
totalSizeFormatted: formatSize(totalSize),
vectorIndexCount,
normalIndexCount
normalIndexCount,
...statusSummary
}
}));
} catch (err) {
@@ -280,7 +299,8 @@ export async function handleCodexLensRoutes(ctx: RouteContext): Promise<boolean>
try {
const config = extractJSON(configResult.output);
if (config.success && config.result) {
responseData.index_dir = config.result.index_root || responseData.index_dir;
// CLI returns index_dir (not index_root)
responseData.index_dir = config.result.index_dir || config.result.index_root || responseData.index_dir;
}
} catch (e) {
console.error('[CodexLens] Failed to parse config:', e.message);

View File

@@ -1,7 +1,7 @@
import * as http from 'http';
import { URL } from 'url';
import { getCoreMemoryStore } from '../core-memory-store.js';
import type { CoreMemory } from '../core-memory-store.js';
import type { CoreMemory, SessionCluster, ClusterMember, ClusterRelation } from '../core-memory-store.js';
/**
* Route context interface
@@ -197,5 +197,329 @@ export async function handleCoreMemoryRoutes(ctx: RouteContext): Promise<boolean
return true;
}
// ============================================================
// Session Clustering API Endpoints
// ============================================================
// API: Get all clusters
if (pathname === '/api/core-memory/clusters' && req.method === 'GET') {
const projectPath = url.searchParams.get('path') || initialPath;
const status = url.searchParams.get('status') || undefined;
try {
const store = getCoreMemoryStore(projectPath);
const clusters = store.listClusters(status);
// Add member count to each cluster
const clustersWithCount = clusters.map(c => ({
...c,
memberCount: store.getClusterMembers(c.id).length
}));
res.writeHead(200, { 'Content-Type': 'application/json' });
res.end(JSON.stringify({ success: true, clusters: clustersWithCount }));
} catch (error: unknown) {
res.writeHead(500, { 'Content-Type': 'application/json' });
res.end(JSON.stringify({ error: (error as Error).message }));
}
return true;
}
// API: Get cluster detail with members
if (pathname.match(/^\/api\/core-memory\/clusters\/[^\/]+$/) && req.method === 'GET') {
const clusterId = pathname.split('/').pop()!;
const projectPath = url.searchParams.get('path') || initialPath;
try {
const store = getCoreMemoryStore(projectPath);
const cluster = store.getCluster(clusterId);
if (!cluster) {
res.writeHead(404, { 'Content-Type': 'application/json' });
res.end(JSON.stringify({ error: 'Cluster not found' }));
return true;
}
const members = store.getClusterMembers(clusterId);
const relations = store.getClusterRelations(clusterId);
// Get metadata for each member
const membersWithMetadata = members.map(m => ({
...m,
metadata: store.getSessionMetadata(m.session_id)
}));
res.writeHead(200, { 'Content-Type': 'application/json' });
res.end(JSON.stringify({
success: true,
cluster,
members: membersWithMetadata,
relations
}));
} catch (error: unknown) {
res.writeHead(500, { 'Content-Type': 'application/json' });
res.end(JSON.stringify({ error: (error as Error).message }));
}
return true;
}
// API: Auto-cluster sessions
if (pathname === '/api/core-memory/clusters/auto' && req.method === 'POST') {
handlePostRequest(req, res, async (body) => {
const { scope = 'recent', minClusterSize = 2, path: projectPath } = body;
const basePath = projectPath || initialPath;
try {
const { SessionClusteringService } = await import('../session-clustering-service.js');
const service = new SessionClusteringService(basePath);
const validScope: 'all' | 'recent' | 'unclustered' =
scope === 'all' || scope === 'recent' || scope === 'unclustered' ? scope : 'recent';
const result = await service.autocluster({
scope: validScope,
minClusterSize
});
// Broadcast update event
broadcastToClients({
type: 'CLUSTERS_UPDATED',
payload: {
...result,
timestamp: new Date().toISOString()
}
});
return {
success: true,
...result
};
} catch (error: unknown) {
return { error: (error as Error).message, status: 500 };
}
});
return true;
}
// API: Create new cluster
if (pathname === '/api/core-memory/clusters' && req.method === 'POST') {
handlePostRequest(req, res, async (body) => {
const { name, description, intent, metadata, path: projectPath } = body;
if (!name) {
return { error: 'name is required', status: 400 };
}
const basePath = projectPath || initialPath;
try {
const store = getCoreMemoryStore(basePath);
const cluster = store.createCluster({
name,
description,
intent,
metadata: metadata ? JSON.stringify(metadata) : undefined
});
// Broadcast update event
broadcastToClients({
type: 'CLUSTER_UPDATED',
payload: {
cluster,
timestamp: new Date().toISOString()
}
});
return {
success: true,
cluster
};
} catch (error: unknown) {
return { error: (error as Error).message, status: 500 };
}
});
return true;
}
// API: Update cluster (supports both PUT and PATCH)
if (pathname.match(/^\/api\/core-memory\/clusters\/[^\/]+$/) && (req.method === 'PUT' || req.method === 'PATCH')) {
const clusterId = pathname.split('/').pop()!;
handlePostRequest(req, res, async (body) => {
const { name, description, intent, status, metadata, path: projectPath } = body;
const basePath = projectPath || initialPath;
try {
const store = getCoreMemoryStore(basePath);
const cluster = store.updateCluster(clusterId, {
name,
description,
intent,
status,
metadata: metadata ? JSON.stringify(metadata) : undefined
});
if (!cluster) {
return { error: 'Cluster not found', status: 404 };
}
// Broadcast update event
broadcastToClients({
type: 'CLUSTER_UPDATED',
payload: {
cluster,
timestamp: new Date().toISOString()
}
});
return {
success: true,
cluster
};
} catch (error: unknown) {
return { error: (error as Error).message, status: 500 };
}
});
return true;
}
// API: Delete cluster
if (pathname.match(/^\/api\/core-memory\/clusters\/[^\/]+$/) && req.method === 'DELETE') {
const clusterId = pathname.split('/').pop()!;
const projectPath = url.searchParams.get('path') || initialPath;
try {
const store = getCoreMemoryStore(projectPath);
const deleted = store.deleteCluster(clusterId);
if (!deleted) {
res.writeHead(404, { 'Content-Type': 'application/json' });
res.end(JSON.stringify({ error: 'Cluster not found' }));
return true;
}
// Broadcast update event
broadcastToClients({
type: 'CLUSTER_UPDATED',
payload: {
clusterId,
deleted: true,
timestamp: new Date().toISOString()
}
});
res.writeHead(204, { 'Content-Type': 'application/json' });
res.end();
} catch (error: unknown) {
res.writeHead(500, { 'Content-Type': 'application/json' });
res.end(JSON.stringify({ error: (error as Error).message }));
}
return true;
}
// API: Add member to cluster
if (pathname.match(/^\/api\/core-memory\/clusters\/[^\/]+\/members$/) && req.method === 'POST') {
const clusterId = pathname.split('/')[4];
handlePostRequest(req, res, async (body) => {
const { session_id, session_type, sequence_order, relevance_score, path: projectPath } = body;
if (!session_id || !session_type) {
return { error: 'session_id and session_type are required', status: 400 };
}
const basePath = projectPath || initialPath;
try {
const store = getCoreMemoryStore(basePath);
const member = store.addClusterMember({
cluster_id: clusterId,
session_id,
session_type,
sequence_order: sequence_order ?? 0,
relevance_score: relevance_score ?? 1.0
});
// Broadcast update event
broadcastToClients({
type: 'CLUSTER_UPDATED',
payload: {
clusterId,
member,
timestamp: new Date().toISOString()
}
});
return {
success: true,
member
};
} catch (error: unknown) {
return { error: (error as Error).message, status: 500 };
}
});
return true;
}
// API: Remove member from cluster
if (pathname.match(/^\/api\/core-memory\/clusters\/[^\/]+\/members\/[^\/]+$/) && req.method === 'DELETE') {
const parts = pathname.split('/');
const clusterId = parts[4];
const sessionId = parts[6];
const projectPath = url.searchParams.get('path') || initialPath;
try {
const store = getCoreMemoryStore(projectPath);
const removed = store.removeClusterMember(clusterId, sessionId);
if (!removed) {
res.writeHead(404, { 'Content-Type': 'application/json' });
res.end(JSON.stringify({ error: 'Member not found' }));
return true;
}
// Broadcast update event
broadcastToClients({
type: 'CLUSTER_UPDATED',
payload: {
clusterId,
removedSessionId: sessionId,
timestamp: new Date().toISOString()
}
});
res.writeHead(204, { 'Content-Type': 'application/json' });
res.end();
} catch (error: unknown) {
res.writeHead(500, { 'Content-Type': 'application/json' });
res.end(JSON.stringify({ error: (error as Error).message }));
}
return true;
}
// API: Search sessions by keyword
if (pathname === '/api/core-memory/sessions/search' && req.method === 'GET') {
const keyword = url.searchParams.get('q') || '';
const projectPath = url.searchParams.get('path') || initialPath;
if (!keyword) {
res.writeHead(400, { 'Content-Type': 'application/json' });
res.end(JSON.stringify({ error: 'Query parameter q is required' }));
return true;
}
try {
const store = getCoreMemoryStore(projectPath);
const results = store.searchSessionsByKeyword(keyword);
res.writeHead(200, { 'Content-Type': 'application/json' });
res.end(JSON.stringify({ success: true, results }));
} catch (error: unknown) {
res.writeHead(500, { 'Content-Type': 'application/json' });
res.end(JSON.stringify({ error: (error as Error).message }));
}
return true;
}
return false;
}

View File

@@ -138,8 +138,16 @@ function findAllIndexDbs(dir: string): string[] {
/**
* Map codex-lens symbol kinds to graph node types
* Returns null for non-code symbols (markdown headings, etc.)
*/
function mapSymbolKind(kind: string): string {
function mapSymbolKind(kind: string): string | null {
const kindLower = kind.toLowerCase();
// Exclude markdown headings
if (/^h[1-6]$/.test(kindLower)) {
return null;
}
const kindMap: Record<string, string> = {
'function': 'FUNCTION',
'class': 'CLASS',
@@ -148,8 +156,13 @@ function mapSymbolKind(kind: string): string {
'module': 'MODULE',
'interface': 'CLASS', // TypeScript interfaces as CLASS
'type': 'CLASS', // Type aliases as CLASS
'constant': 'VARIABLE',
'property': 'VARIABLE',
'parameter': 'VARIABLE',
'import': 'MODULE',
'export': 'MODULE',
};
return kindMap[kind.toLowerCase()] || 'VARIABLE';
return kindMap[kindLower] || 'VARIABLE';
}
/**
@@ -224,13 +237,19 @@ async function querySymbols(projectPath: string, fileFilter?: string, moduleFilt
db.close();
allNodes.push(...rows.map((row: any) => ({
id: `${row.file}:${row.name}:${row.start_line}`,
name: row.name,
type: mapSymbolKind(row.kind),
file: row.file,
line: row.start_line,
})));
// Filter out non-code symbols (markdown headings, etc.)
rows.forEach((row: any) => {
const type = mapSymbolKind(row.kind);
if (type !== null) {
allNodes.push({
id: `${row.file}:${row.name}:${row.start_line}`,
name: row.name,
type,
file: row.file,
line: row.start_line,
});
}
});
} catch (err) {
const message = err instanceof Error ? err.message : String(err);
console.error(`[Graph] Failed to query symbols from ${dbPath}: ${message}`);

View File

@@ -202,6 +202,46 @@ export async function handleHooksRoutes(ctx: RouteContext): Promise<boolean> {
resolvedSessionId = extractSessionIdFromPath(filePath);
}
// Handle context hooks (session-start, context)
if (type === 'session-start' || type === 'context') {
try {
const projectPath = url.searchParams.get('path') || initialPath;
const { SessionClusteringService } = await import('../session-clustering-service.js');
const clusteringService = new SessionClusteringService(projectPath);
const format = url.searchParams.get('format') || 'markdown';
// Pass type and prompt to getProgressiveIndex
// session-start: returns recent sessions by time
// context: returns intent-matched sessions based on prompt
const index = await clusteringService.getProgressiveIndex({
type: type as 'session-start' | 'context',
sessionId: resolvedSessionId,
prompt: extraData.prompt // Pass user prompt for intent matching
});
// Return context directly
return {
success: true,
type: 'context',
format,
content: index,
sessionId: resolvedSessionId
};
} catch (error) {
console.error('[Hooks] Failed to generate context:', error);
// Return empty content on failure (fail silently)
return {
success: true,
type: 'context',
format: 'markdown',
content: '',
sessionId: resolvedSessionId,
error: (error as Error).message
};
}
}
// Broadcast to all connected WebSocket clients
const notification = {
type: type || 'session_updated',

View File

@@ -132,6 +132,7 @@ const MODULE_FILES = [
'views/memory.js',
'views/core-memory.js',
'views/core-memory-graph.js',
'views/core-memory-clusters.js',
'views/prompt-history.js',
'views/skills-manager.js',
'views/rules-manager.js',

View File

@@ -0,0 +1,842 @@
/**
* Session Clustering Service
* Intelligently groups related sessions into clusters using multi-dimensional similarity analysis
*/
import { CoreMemoryStore, SessionCluster, ClusterMember, SessionMetadataCache } from './core-memory-store.js';
import { CliHistoryStore } from '../tools/cli-history-store.js';
import { StoragePaths } from '../config/storage-paths.js';
import { readdirSync, readFileSync, statSync, existsSync } from 'fs';
import { join } from 'path';
// Clustering dimension weights
const WEIGHTS = {
fileOverlap: 0.3,
temporalProximity: 0.2,
semanticSimilarity: 0.3,
intentAlignment: 0.2,
};
// Clustering threshold
const CLUSTER_THRESHOLD = 0.6;
export interface ClusteringOptions {
scope?: 'all' | 'recent' | 'unclustered';
timeRange?: { start: string; end: string };
minClusterSize?: number;
}
export interface ClusteringResult {
clustersCreated: number;
sessionsProcessed: number;
sessionsClustered: number;
}
export class SessionClusteringService {
private coreMemoryStore: CoreMemoryStore;
private cliHistoryStore: CliHistoryStore;
private projectPath: string;
constructor(projectPath: string) {
this.projectPath = projectPath;
this.coreMemoryStore = new CoreMemoryStore(projectPath);
this.cliHistoryStore = new CliHistoryStore(projectPath);
}
/**
* Collect all session sources
*/
async collectSessions(options?: ClusteringOptions): Promise<SessionMetadataCache[]> {
const sessions: SessionMetadataCache[] = [];
// 1. Core Memories
const memories = this.coreMemoryStore.getMemories({ archived: false, limit: 1000 });
for (const memory of memories) {
const cached = this.coreMemoryStore.getSessionMetadata(memory.id);
if (cached) {
sessions.push(cached);
} else {
const metadata = this.extractMetadata(memory, 'core_memory');
sessions.push(metadata);
}
}
// 2. CLI History
const history = this.cliHistoryStore.getHistory({ limit: 1000 });
for (const exec of history.executions) {
const cached = this.coreMemoryStore.getSessionMetadata(exec.id);
if (cached) {
sessions.push(cached);
} else {
const conversation = this.cliHistoryStore.getConversation(exec.id);
if (conversation) {
const metadata = this.extractMetadata(conversation, 'cli_history');
sessions.push(metadata);
}
}
}
// 3. Workflow Sessions (WFS-*)
const workflowSessions = await this.parseWorkflowSessions();
sessions.push(...workflowSessions);
// Apply scope filter
if (options?.scope === 'recent') {
// Last 30 days
const cutoff = new Date();
cutoff.setDate(cutoff.getDate() - 30);
const cutoffStr = cutoff.toISOString();
return sessions.filter(s => (s.created_at || '') >= cutoffStr);
} else if (options?.scope === 'unclustered') {
// Only sessions not in any cluster
return sessions.filter(s => {
const clusters = this.coreMemoryStore.getSessionClusters(s.session_id);
return clusters.length === 0;
});
}
return sessions;
}
/**
* Extract metadata from a session
*/
extractMetadata(session: any, type: 'core_memory' | 'workflow' | 'cli_history' | 'native'): SessionMetadataCache {
let content = '';
let title = '';
let created_at = '';
if (type === 'core_memory') {
content = session.content || '';
created_at = session.created_at;
// Extract title from first line
const lines = content.split('\n');
title = lines[0].replace(/^#+\s*/, '').trim().substring(0, 100);
} else if (type === 'cli_history') {
// Extract from conversation turns
const turns = session.turns || [];
if (turns.length > 0) {
content = turns.map((t: any) => t.prompt).join('\n');
title = turns[0].prompt.substring(0, 100);
created_at = session.created_at || turns[0].timestamp;
}
} else if (type === 'workflow') {
content = session.content || '';
title = session.title || 'Workflow Session';
created_at = session.created_at || '';
}
const summary = content.substring(0, 200).trim();
const keywords = this.extractKeywords(content);
const file_patterns = this.extractFilePatterns(content);
const token_estimate = Math.ceil(content.length / 4);
return {
session_id: session.id,
session_type: type,
title,
summary,
keywords,
token_estimate,
file_patterns,
created_at,
last_accessed: new Date().toISOString(),
access_count: 0
};
}
/**
* Extract keywords from content
*/
private extractKeywords(content: string): string[] {
const keywords = new Set<string>();
// 1. File paths (src/xxx, .ts, .js, etc)
const filePathRegex = /(?:^|\s|["'`])((?:\.\/|\.\.\/|\/)?[\w-]+(?:\/[\w-]+)*\.[\w]+)(?:\s|["'`]|$)/g;
let match;
while ((match = filePathRegex.exec(content)) !== null) {
keywords.add(match[1]);
}
// 2. Function/Class names (camelCase, PascalCase)
const camelCaseRegex = /\b([A-Z][a-z]+(?:[A-Z][a-z]+)+|[a-z]+[A-Z][a-z]+(?:[A-Z][a-z]+)*)\b/g;
while ((match = camelCaseRegex.exec(content)) !== null) {
keywords.add(match[1]);
}
// 3. Technical terms (common frameworks/libraries)
const techTerms = [
'react', 'vue', 'angular', 'typescript', 'javascript', 'node', 'express',
'auth', 'authentication', 'jwt', 'oauth', 'session', 'token',
'api', 'rest', 'graphql', 'database', 'sql', 'mongodb', 'redis',
'test', 'testing', 'jest', 'mocha', 'vitest',
'refactor', 'refactoring', 'optimization', 'performance',
'bug', 'fix', 'error', 'issue', 'debug'
];
const lowerContent = content.toLowerCase();
for (const term of techTerms) {
if (lowerContent.includes(term)) {
keywords.add(term);
}
}
// Return top 20 keywords
return Array.from(keywords).slice(0, 20);
}
/**
* Extract file patterns from content
*/
private extractFilePatterns(content: string): string[] {
const patterns = new Set<string>();
// Extract directory patterns (src/xxx/, lib/xxx/)
const dirRegex = /\b((?:src|lib|test|dist|build|public|components|utils|services|config|core|tools)(?:\/[\w-]+)*)\//g;
let match;
while ((match = dirRegex.exec(content)) !== null) {
patterns.add(match[1] + '/**');
}
// Extract file extension patterns
const extRegex = /\.(\w+)(?:\s|$|["'`])/g;
const extensions = new Set<string>();
while ((match = extRegex.exec(content)) !== null) {
extensions.add(match[1]);
}
// Add extension patterns
if (extensions.size > 0) {
patterns.add(`**/*.{${Array.from(extensions).join(',')}}`);
}
return Array.from(patterns).slice(0, 10);
}
/**
* Calculate relevance score between two sessions
*/
calculateRelevance(session1: SessionMetadataCache, session2: SessionMetadataCache): number {
const fileScore = this.calculateFileOverlap(session1, session2);
const temporalScore = this.calculateTemporalProximity(session1, session2);
const semanticScore = this.calculateSemanticSimilarity(session1, session2);
const intentScore = this.calculateIntentAlignment(session1, session2);
return (
fileScore * WEIGHTS.fileOverlap +
temporalScore * WEIGHTS.temporalProximity +
semanticScore * WEIGHTS.semanticSimilarity +
intentScore * WEIGHTS.intentAlignment
);
}
/**
* Calculate file path overlap score (Jaccard similarity)
*/
private calculateFileOverlap(s1: SessionMetadataCache, s2: SessionMetadataCache): number {
const files1 = new Set(s1.file_patterns || []);
const files2 = new Set(s2.file_patterns || []);
if (files1.size === 0 || files2.size === 0) return 0;
const intersection = new Set([...files1].filter(f => files2.has(f)));
const union = new Set([...files1, ...files2]);
return intersection.size / union.size;
}
/**
* Calculate temporal proximity score
* 24h: 1.0, 7d: 0.7, 30d: 0.4, >30d: 0.1
*/
private calculateTemporalProximity(s1: SessionMetadataCache, s2: SessionMetadataCache): number {
if (!s1.created_at || !s2.created_at) return 0.1;
const t1 = new Date(s1.created_at).getTime();
const t2 = new Date(s2.created_at).getTime();
const diffMs = Math.abs(t1 - t2);
const diffHours = diffMs / (1000 * 60 * 60);
if (diffHours <= 24) return 1.0;
if (diffHours <= 24 * 7) return 0.7;
if (diffHours <= 24 * 30) return 0.4;
return 0.1;
}
/**
* Calculate semantic similarity using keyword overlap (Jaccard similarity)
*/
private calculateSemanticSimilarity(s1: SessionMetadataCache, s2: SessionMetadataCache): number {
const kw1 = new Set(s1.keywords || []);
const kw2 = new Set(s2.keywords || []);
if (kw1.size === 0 || kw2.size === 0) return 0;
const intersection = new Set([...kw1].filter(k => kw2.has(k)));
const union = new Set([...kw1, ...kw2]);
return intersection.size / union.size;
}
/**
* Calculate intent alignment score
* Based on title/summary keyword matching
*/
private calculateIntentAlignment(s1: SessionMetadataCache, s2: SessionMetadataCache): number {
const text1 = ((s1.title || '') + ' ' + (s1.summary || '')).toLowerCase();
const text2 = ((s2.title || '') + ' ' + (s2.summary || '')).toLowerCase();
if (!text1 || !text2) return 0;
// Simple word-based TF-IDF approximation
const words1 = text1.split(/\s+/).filter(w => w.length > 3);
const words2 = text2.split(/\s+/).filter(w => w.length > 3);
const set1 = new Set(words1);
const set2 = new Set(words2);
const intersection = new Set([...set1].filter(w => set2.has(w)));
const union = new Set([...set1, ...set2]);
return intersection.size / union.size;
}
/**
* Run auto-clustering algorithm
*/
async autocluster(options?: ClusteringOptions): Promise<ClusteringResult> {
// 1. Collect sessions
const sessions = await this.collectSessions(options);
console.log(`[Clustering] Collected ${sessions.length} sessions`);
// 2. Update metadata cache
for (const session of sessions) {
this.coreMemoryStore.upsertSessionMetadata(session);
}
// 3. Calculate relevance matrix
const n = sessions.length;
const relevanceMatrix: number[][] = Array(n).fill(0).map(() => Array(n).fill(0));
for (let i = 0; i < n; i++) {
for (let j = i + 1; j < n; j++) {
const score = this.calculateRelevance(sessions[i], sessions[j]);
relevanceMatrix[i][j] = score;
relevanceMatrix[j][i] = score;
}
}
// 4. Agglomerative clustering
const clusters = this.agglomerativeClustering(sessions, relevanceMatrix, CLUSTER_THRESHOLD);
console.log(`[Clustering] Generated ${clusters.length} clusters`);
// 5. Create session_clusters
let clustersCreated = 0;
let sessionsClustered = 0;
for (const cluster of clusters) {
if (cluster.length < (options?.minClusterSize || 2)) {
continue; // Skip small clusters
}
const clusterName = this.generateClusterName(cluster);
const clusterIntent = this.generateClusterIntent(cluster);
const clusterRecord = this.coreMemoryStore.createCluster({
name: clusterName,
description: `Auto-generated cluster with ${cluster.length} sessions`,
intent: clusterIntent,
status: 'active'
});
// Add members
cluster.forEach((session, index) => {
this.coreMemoryStore.addClusterMember({
cluster_id: clusterRecord.id,
session_id: session.session_id,
session_type: session.session_type as 'core_memory' | 'workflow' | 'cli_history' | 'native',
sequence_order: index + 1,
relevance_score: 1.0 // TODO: Calculate based on centrality
});
});
clustersCreated++;
sessionsClustered += cluster.length;
}
return {
clustersCreated,
sessionsProcessed: sessions.length,
sessionsClustered
};
}
/**
* Agglomerative clustering algorithm
* Returns array of clusters (each cluster is array of sessions)
*/
private agglomerativeClustering(
sessions: SessionMetadataCache[],
relevanceMatrix: number[][],
threshold: number
): SessionMetadataCache[][] {
const n = sessions.length;
// Initialize: each session is its own cluster
const clusters: Set<number>[] = sessions.map((_, i) => new Set([i]));
while (true) {
let maxScore = -1;
let mergeI = -1;
let mergeJ = -1;
// Find pair of clusters with highest average linkage
for (let i = 0; i < clusters.length; i++) {
for (let j = i + 1; j < clusters.length; j++) {
const score = this.averageLinkage(clusters[i], clusters[j], relevanceMatrix);
if (score > maxScore) {
maxScore = score;
mergeI = i;
mergeJ = j;
}
}
}
// Stop if no pair exceeds threshold
if (maxScore < threshold) break;
// Merge clusters
const merged = new Set([...clusters[mergeI], ...clusters[mergeJ]]);
clusters.splice(mergeJ, 1); // Remove j first (higher index)
clusters.splice(mergeI, 1);
clusters.push(merged);
}
// Convert cluster indices to sessions
return clusters.map(cluster =>
Array.from(cluster).map(i => sessions[i])
);
}
/**
* Calculate average linkage between two clusters
*/
private averageLinkage(
cluster1: Set<number>,
cluster2: Set<number>,
relevanceMatrix: number[][]
): number {
let sum = 0;
let count = 0;
for (const i of cluster1) {
for (const j of cluster2) {
sum += relevanceMatrix[i][j];
count++;
}
}
return count > 0 ? sum / count : 0;
}
/**
* Generate cluster name from members
*/
private generateClusterName(members: SessionMetadataCache[]): string {
// Count keyword frequency
const keywordFreq = new Map<string, number>();
for (const member of members) {
for (const keyword of member.keywords || []) {
keywordFreq.set(keyword, (keywordFreq.get(keyword) || 0) + 1);
}
}
// Get top 2 keywords
const sorted = Array.from(keywordFreq.entries())
.sort((a, b) => b[1] - a[1])
.map(([kw]) => kw);
if (sorted.length >= 2) {
return `${sorted[0]}-${sorted[1]}`;
} else if (sorted.length === 1) {
return sorted[0];
} else {
return 'unnamed-cluster';
}
}
/**
* Generate cluster intent from members
*/
private generateClusterIntent(members: SessionMetadataCache[]): string {
// Extract common action words from titles
const actionWords = ['implement', 'refactor', 'fix', 'add', 'create', 'update', 'optimize'];
const titles = members.map(m => (m.title || '').toLowerCase());
for (const action of actionWords) {
const count = titles.filter(t => t.includes(action)).length;
if (count >= members.length / 2) {
const topic = this.generateClusterName(members);
return `${action.charAt(0).toUpperCase() + action.slice(1)} ${topic}`;
}
}
return `Work on ${this.generateClusterName(members)}`;
}
/**
* Get progressive disclosure index for hook
* @param options - Configuration options
* @param options.type - 'session-start' returns recent sessions, 'context' returns intent-matched sessions
* @param options.sessionId - Current session ID (optional)
* @param options.prompt - User prompt for intent matching (required for 'context' type)
*/
async getProgressiveIndex(options: {
type: 'session-start' | 'context';
sessionId?: string;
prompt?: string;
}): Promise<string> {
const { type, sessionId, prompt } = options;
// For session-start: return recent sessions by time
if (type === 'session-start') {
return this.getRecentSessionsIndex();
}
// For context: return intent-matched sessions based on prompt
if (type === 'context' && prompt) {
return this.getIntentMatchedIndex(prompt, sessionId);
}
// Fallback to recent sessions
return this.getRecentSessionsIndex();
}
/**
* Get recent sessions index (for session-start)
*/
private async getRecentSessionsIndex(): Promise<string> {
const sessions = await this.collectSessions({ scope: 'recent' });
// Sort by created_at descending (most recent first)
const sortedSessions = sessions
.filter(s => s.created_at)
.sort((a, b) => (b.created_at || '').localeCompare(a.created_at || ''))
.slice(0, 10); // Top 10 recent sessions
if (sortedSessions.length === 0) {
return `<ccw-session-context>
## 📋 Recent Sessions
No recent sessions found. Start a new workflow to begin tracking.
**MCP Tools**:
\`\`\`
# Search sessions
Use tool: mcp__ccw-tools__core_memory
Parameters: { "action": "search", "query": "<keyword>" }
# Create new session
Parameters: { "action": "save", "content": "<context>" }
\`\`\`
</ccw-session-context>`;
}
// Generate table
let table = `| # | Session | Type | Title | Date |\n`;
table += `|---|---------|------|-------|------|\n`;
sortedSessions.forEach((s, idx) => {
const type = s.session_type === 'core_memory' ? 'Core' :
s.session_type === 'workflow' ? 'Workflow' : 'CLI';
const title = (s.title || '').substring(0, 40);
const date = s.created_at ? new Date(s.created_at).toLocaleDateString() : '';
table += `| ${idx + 1} | ${s.session_id} | ${type} | ${title} | ${date} |\n`;
});
return `<ccw-session-context>
## 📋 Recent Sessions (Last 30 days)
${table}
**Resume via MCP**:
\`\`\`
Use tool: mcp__ccw-tools__core_memory
Parameters: { "action": "load", "id": "${sortedSessions[0].session_id}" }
\`\`\`
---
**Tip**: Sessions are sorted by most recent. Use \`search\` action to find specific topics.
</ccw-session-context>`;
}
/**
* Get intent-matched sessions index (for context with prompt)
*/
private async getIntentMatchedIndex(prompt: string, sessionId?: string): Promise<string> {
const sessions = await this.collectSessions({ scope: 'all' });
if (sessions.length === 0) {
return `<ccw-session-context>
## 📋 Related Sessions
No sessions available for intent matching.
</ccw-session-context>`;
}
// Create a virtual session from the prompt for similarity calculation
const promptSession: SessionMetadataCache = {
session_id: 'prompt-virtual',
session_type: 'native',
title: prompt.substring(0, 100),
summary: prompt.substring(0, 200),
keywords: this.extractKeywords(prompt),
token_estimate: Math.ceil(prompt.length / 4),
file_patterns: this.extractFilePatterns(prompt),
created_at: new Date().toISOString(),
last_accessed: new Date().toISOString(),
access_count: 0
};
// Calculate relevance scores for all sessions
const scoredSessions = sessions
.filter(s => s.session_id !== sessionId) // Exclude current session
.map(s => ({
session: s,
score: this.calculateRelevance(promptSession, s)
}))
.filter(item => item.score >= 0.3) // Minimum relevance threshold
.sort((a, b) => b.score - a.score)
.slice(0, 8); // Top 8 relevant sessions
if (scoredSessions.length === 0) {
return `<ccw-session-context>
## 📋 Related Sessions
No sessions match current intent. Consider:
- Starting fresh with a new approach
- Using \`search\` to find sessions by keyword
**MCP Tools**:
\`\`\`
Use tool: mcp__ccw-tools__core_memory
Parameters: { "action": "search", "query": "<keyword>" }
\`\`\`
</ccw-session-context>`;
}
// Group by relevance tier
const highRelevance = scoredSessions.filter(s => s.score >= 0.6);
const mediumRelevance = scoredSessions.filter(s => s.score >= 0.4 && s.score < 0.6);
const lowRelevance = scoredSessions.filter(s => s.score < 0.4);
// Generate output
let output = `<ccw-session-context>
## 📋 Intent-Matched Sessions
**Detected Intent**: ${promptSession.keywords.slice(0, 5).join(', ') || 'General'}
`;
if (highRelevance.length > 0) {
output += `### 🔥 Highly Relevant (${highRelevance.length})\n`;
output += `| Session | Type | Match | Summary |\n`;
output += `|---------|------|-------|--------|\n`;
for (const item of highRelevance) {
const type = item.session.session_type === 'core_memory' ? 'Core' :
item.session.session_type === 'workflow' ? 'Workflow' : 'CLI';
const matchPct = Math.round(item.score * 100);
const summary = (item.session.title || item.session.summary || '').substring(0, 35);
output += `| ${item.session.session_id} | ${type} | ${matchPct}% | ${summary} |\n`;
}
output += `\n`;
}
if (mediumRelevance.length > 0) {
output += `### 📌 Related (${mediumRelevance.length})\n`;
output += `| Session | Type | Match | Summary |\n`;
output += `|---------|------|-------|--------|\n`;
for (const item of mediumRelevance) {
const type = item.session.session_type === 'core_memory' ? 'Core' :
item.session.session_type === 'workflow' ? 'Workflow' : 'CLI';
const matchPct = Math.round(item.score * 100);
const summary = (item.session.title || item.session.summary || '').substring(0, 35);
output += `| ${item.session.session_id} | ${type} | ${matchPct}% | ${summary} |\n`;
}
output += `\n`;
}
if (lowRelevance.length > 0) {
output += `### 💡 May Be Useful (${lowRelevance.length})\n`;
const sessionList = lowRelevance.map(s => s.session.session_id).join(', ');
output += `${sessionList}\n\n`;
}
// Add resume command for top match
const topMatch = scoredSessions[0];
output += `**Resume Top Match**:
\`\`\`
Use tool: mcp__ccw-tools__core_memory
Parameters: { "action": "load", "id": "${topMatch.session.session_id}" }
\`\`\`
---
**Tip**: Sessions ranked by semantic similarity to your prompt.
</ccw-session-context>`;
return output;
}
/**
* Legacy method for backward compatibility
* @deprecated Use getProgressiveIndex({ type, sessionId, prompt }) instead
*/
async getProgressiveIndexLegacy(sessionId?: string): Promise<string> {
let activeCluster: SessionCluster | null = null;
let members: SessionMetadataCache[] = [];
if (sessionId) {
const clusters = this.coreMemoryStore.getSessionClusters(sessionId);
if (clusters.length > 0) {
activeCluster = clusters[0];
const clusterMembers = this.coreMemoryStore.getClusterMembers(activeCluster.id);
members = clusterMembers
.map(m => this.coreMemoryStore.getSessionMetadata(m.session_id))
.filter((m): m is SessionMetadataCache => m !== null)
.sort((a, b) => (a.created_at || '').localeCompare(b.created_at || ''));
}
}
if (!activeCluster || members.length === 0) {
return `<ccw-session-context>
## 📋 Related Sessions Index
No active cluster found. Start a new workflow or continue from recent sessions.
**MCP Tools**:
\`\`\`
# Search sessions
Use tool: mcp__ccw-tools__core_memory
Parameters: { "action": "search", "query": "<keyword>" }
# Trigger clustering
Parameters: { "action": "cluster", "scope": "auto" }
\`\`\`
</ccw-session-context>`;
}
// Generate table
let table = `| # | Session | Type | Summary | Tokens |\n`;
table += `|---|---------|------|---------|--------|\n`;
members.forEach((m, idx) => {
const type = m.session_type === 'core_memory' ? 'Core' :
m.session_type === 'workflow' ? 'Workflow' : 'CLI';
const summary = (m.summary || '').substring(0, 40);
const token = `~${m.token_estimate || 0}`;
table += `| ${idx + 1} | ${m.session_id} | ${type} | ${summary} | ${token} |\n`;
});
// Generate timeline - show multiple recent sessions
let timeline = '';
if (members.length > 0) {
const timelineEntries: string[] = [];
const displayCount = Math.min(members.length, 3); // Show last 3 sessions
for (let i = members.length - displayCount; i < members.length; i++) {
const member = members[i];
const date = member.created_at ? new Date(member.created_at).toLocaleDateString() : '';
const title = member.title?.substring(0, 30) || 'Untitled';
const isCurrent = i === members.length - 1;
const marker = isCurrent ? ' ← Current' : '';
timelineEntries.push(`${date} ─●─ ${member.session_id} (${title})${marker}`);
}
timeline = `\`\`\`\n${timelineEntries.join('\n │\n')}\n\`\`\``;
}
return `<ccw-session-context>
## 📋 Related Sessions Index
### 🔗 Active Cluster: ${activeCluster.name} (${members.length} sessions)
**Intent**: ${activeCluster.intent || 'No intent specified'}
${table}
**Resume via MCP**:
\`\`\`
Use tool: mcp__ccw-tools__core_memory
Parameters: { "action": "load", "id": "${members[members.length - 1].session_id}" }
Or load entire cluster:
{ "action": "load-cluster", "clusterId": "${activeCluster.id}" }
\`\`\`
### 📊 Timeline
${timeline}
---
**Tip**: Use \`mcp__ccw-tools__core_memory({ action: "search", query: "<keyword>" })\` to find more sessions
</ccw-session-context>`;
}
/**
* Parse workflow session files
*/
private async parseWorkflowSessions(): Promise<SessionMetadataCache[]> {
const sessions: SessionMetadataCache[] = [];
const workflowDir = join(this.projectPath, '.workflow', 'sessions');
if (!existsSync(workflowDir)) {
return sessions;
}
try {
const sessionDirs = readdirSync(workflowDir).filter(d => d.startsWith('WFS-'));
for (const sessionDir of sessionDirs) {
const sessionFile = join(workflowDir, sessionDir, 'session.json');
if (!existsSync(sessionFile)) continue;
try {
const content = readFileSync(sessionFile, 'utf8');
const sessionData = JSON.parse(content);
const metadata: SessionMetadataCache = {
session_id: sessionDir,
session_type: 'workflow',
title: sessionData.title || sessionDir,
summary: (sessionData.description || '').substring(0, 200),
keywords: this.extractKeywords(JSON.stringify(sessionData)),
token_estimate: Math.ceil(JSON.stringify(sessionData).length / 4),
file_patterns: this.extractFilePatterns(JSON.stringify(sessionData)),
created_at: sessionData.created_at || statSync(sessionFile).mtime.toISOString(),
last_accessed: new Date().toISOString(),
access_count: 0
};
sessions.push(metadata);
} catch (err) {
console.warn(`[Clustering] Failed to parse ${sessionFile}:`, err);
}
}
} catch (err) {
console.warn('[Clustering] Failed to read workflow sessions:', err);
}
return sessions;
}
/**
* Update metadata cache for all sessions
*/
async refreshMetadataCache(): Promise<number> {
const sessions = await this.collectSessions({ scope: 'all' });
for (const session of sessions) {
this.coreMemoryStore.upsertSessionMetadata(session);
}
return sessions.length;
}
}