@ak--47/dungeon-master 1.2.1 → 1.2.3

package/lib/core/storage.js

@@ -8,7 +8,7 @@
 /** @typedef {import('../../types.js').Storage} Storage */
 /** @typedef {import('../../types.js').hookArrayOptions<any>} hookArrayOptions */
 
-import { existsSync } from "fs";
+import { existsSync, mkdirSync } from "fs";
 import pLimit from 'p-limit';
 import os from "os";
 import path from "path";
@@ -60,8 +60,13 @@ export async function createHookArray(arr = [], opts) {
 		writeDir = path.resolve(os.tmpdir());
 	}
 
-	if (typeof config.writeToDisk === "string" && config.writeToDisk.startsWith('gs://')) {
-		writeDir = config.writeToDisk;
+	if (typeof config.writeToDisk === "string") {
+		if (config.writeToDisk.startsWith('gs://')) {
+			writeDir = config.writeToDisk;
+		} else {
+			writeDir = path.resolve(config.writeToDisk);
+			if (!existsSync(writeDir)) mkdirSync(writeDir, { recursive: true });
+		}
 	}
 
 	function getWritePath() {
@@ -213,7 +218,7 @@ export async function createHookArray(arr = [], opts) {
 			while (isWriting) {
 				await new Promise(resolve => setTimeout(resolve, 10));
 			}
-			
+
 			isWriting = true;
 			try {
 				batch++;
@@ -221,6 +226,11 @@ export async function createHookArray(arr = [], opts) {
 				const dataToWrite = [...arr];
 				arr.length = 0; // Clear array after copying data
 				await FILE_CONN(() => writeToDisk(dataToWrite, { writePath }));
+				// Data now lives on disk, not in arr — mirror what transformThenPush
+				// does when crossing BATCH_SIZE so the Mixpanel sender knows to read
+				// from disk instead of from the (now empty) in-memory array.
+				isBatchMode = true;
+				runtime.isBatchMode = true;
 			} finally {
 				isWriting = false;
 			}

package/lib/orchestrators/mixpanel-sender.js

@@ -48,9 +48,10 @@ export async function sendToMixpanel(context) {
 	const commonOpts = {
 		region,
 		fixData: true,
+		v2_compat: true,
 		verbose: false,
 		forceStream: true,
-		strict: true,
+		strict: false,
 		epochEnd: dayjs().unix(),
 		dryRun: false,
 		abridged: false,
@@ -159,7 +160,6 @@ export async function sendToMixpanel(context) {
 		const imported = await mp(creds, groupEventDataToImport, {
 			recordType: "event",
 			...commonOpts,
-			strict: false
 		});
 		log(`  -> ${comma(imported.success)} group events sent\n`);
 		importResults.groupEvents = imported;
@@ -198,7 +198,6 @@ export async function sendToMixpanel(context) {
 					scdKey,
 					scdType,
 					scdLabel: `${scdKey}`,
-					fixData: true,
 					...commonOpts,
 				};
 
@@ -227,6 +226,9 @@ export async function sendToMixpanel(context) {
 		}
 	}
 
+	importResults.problems = collectProblems(importResults);
+	logProblems(importResults.problems);
+
 	log(`${'─'.repeat(50)}\n`);
 
 	// Clean up batch files if needed
@@ -263,4 +265,63 @@ export async function sendToMixpanel(context) {
 let _verbose = true;
 function log(message) {
 	if (_verbose) console.log(message);
+}
+
+/**
+ * Walk importResults and collect any failure/error indicators per import.
+ * Returns an array of problem objects (empty when all imports succeeded).
+ * @param {Object} importResults
+ * @returns {Array<{label: string, failed: number, unparsable: number, serverErrors: number, clientErrors: number, errorCount: number, error?: string, errors?: any[]}>}
+ */
+function collectProblems(importResults) {
+	const entries = [];
+	for (const [key, value] of Object.entries(importResults)) {
+		if (!value || key === 'problems') continue;
+		if (Array.isArray(value)) {
+			value.forEach((v, i) => entries.push([`${key}[${i}]`, v]));
+		} else {
+			entries.push([key, value]);
+		}
+	}
+
+	const problems = [];
+	for (const [label, result] of entries) {
+		if (!result || typeof result !== 'object') continue;
+		const failed = result.failed || 0;
+		const unparsable = result.unparsable || 0;
+		const serverErrors = result.serverErrors || 0;
+		const clientErrors = result.clientErrors || 0;
+		const errorCount = Array.isArray(result.errors) ? result.errors.length : 0;
+		const error = result.error;
+		if (failed || unparsable || serverErrors || clientErrors || errorCount || error) {
+			problems.push({ label, failed, unparsable, serverErrors, clientErrors, errorCount, error, errors: result.errors });
+		}
+	}
+	return problems;
+}
+
+/**
+ * Print the compact problem report. Silent when there are no problems.
+ * @param {ReturnType<typeof collectProblems>} problems
+ */
+function logProblems(problems) {
+	if (!problems || problems.length === 0) return;
+
+	log(`  Problems`);
+	for (const p of problems) {
+		const parts = [];
+		if (p.failed) parts.push(`${comma(p.failed)} failed`);
+		if (p.unparsable) parts.push(`${comma(p.unparsable)} unparsable`);
+		if (p.serverErrors) parts.push(`${comma(p.serverErrors)} 5xx (retried)`);
+		if (p.clientErrors) parts.push(`${comma(p.clientErrors)} client errors (retried)`);
+		if (p.errorCount) parts.push(`${comma(p.errorCount)} error records`);
+		if (p.error) parts.push(`error: ${p.error}`);
+		log(`  -> ${p.label}: ${parts.join(', ')}`);
+		if (Array.isArray(p.errors) && p.errors.length > 0) {
+			const sample = p.errors[0];
+			const summary = typeof sample === 'string' ? sample : JSON.stringify(sample).slice(0, 200);
+			log(`     sample: ${summary}`);
+		}
+	}
+	log('');
 }

package/package.json

@@ -1,6 +1,6 @@
 {
 	"name": "@ak--47/dungeon-master",
-	"version": "1.2.1",
+	"version": "1.2.3",
 	"description": "generate fancy datasets",
 	"type": "module",
 	"main": "index.js",
@@ -66,7 +66,7 @@
 		"dotenv": "^16.4.5",
 		"hyparquet-writer": "^0.6.1",
 		"mixpanel": "^0.18.0",
-		"mixpanel-import": "^3.2.8",
+		"mixpanel-import": "^3.2.9",
 		"p-limit": "^3.1.0",
 		"pino": "^9.0.0",
 		"pino-pretty": "^11.0.0",

package/dungeons/user/shalini.js

@@ -1,847 +0,0 @@
-// ── TWEAK THESE ──
-const SEED = "dm4-dungeon";
-const num_days = 90;
-const num_users = 2_000;
-const avg_events_per_user = 120;
-let token = "your-mixpanel-token";
-
-// ── env overrides ──
-if (process.env.MP_TOKEN) token = process.env.MP_TOKEN;
-
-import dayjs from "dayjs";
-import utc from "dayjs/plugin/utc.js";
-import * as u from "@ak--47/dungeon-master/utils";
-import * as v from "ak-tools";
-
-dayjs.extend(utc);
-const chance = u.initChance(SEED);
-
-/** @typedef  {import("@ak--47/dungeon-master").Dungeon} Config */
-
-const lastSearchQueryType = new Map();
-const workspaceAiUserCount = new Map();
-const workspaceTotalUsers = new Map();
-
-/**
- * ═══════════════════════════════════════════════════════════════════════════
- * DATASET OVERVIEW — SLACK AI PRODUCTIVITY & COLLABORATION PLATFORM
- * ═══════════════════════════════════════════════════════════════════════════
- *
- * Slack AI is a modern team collaboration platform with AI-powered features:
- * channel/thread summarization, intelligent natural-language search, and
- * AI-assisted writing. This dungeon is designed for a CPO-level demo
- * (Shalini Agarwal) focused on behavioral correlations between AI feature
- * adoption and core platform metrics.
- *
- * Scale: 2,000 users · 500 workspaces · 90 days · 19 event types
- * Groups: 500 workspace_id groups associated with messaging + workspace events
- *
- * Key Segments:
- *   - Power User: high session count, AI feature heavy, message volume leader
- *   - Traditionalist: message-first, low AI adoption, consistent but not viral
- *   - Lurker: reads and summarizes, rarely sends; high AI summary usage
- *
- * Subscription tiers: Pro / Business+ / Enterprise Grid
- *
- * Core loop:
- *   workspace_created → member_invited → session_start →
- *   message_sent / message_read / thread_replied →
- *   ai_prompt_viewed → ai_summary_generated / ai_search_performed →
- *   huddle_started → canvas_edited → reaction_added
- *
- * Monetization signals: integration_added, subscription tier (superProp),
- * canvas adoption, AI feature depth (tokens_saved).
- */
-
-/**
- * ═══════════════════════════════════════════════════════════════════════════
- * ANALYTICS HOOKS (7 architected patterns)
- * ═══════════════════════════════════════════════════════════════════════════
- *
- * ───────────────────────────────────────────────────────────────────────────
- * 1. THE AI MULTIPLIER (everything)
- * ───────────────────────────────────────────────────────────────────────────
- *
- * PATTERN: Users who generate 5+ AI summaries in any 7-day rolling window
- * become "AI-activated." For each such window, clone message_sent and
- * thread_replied events into the FOLLOWING 7 days (3x volume boost with
- * randomized time offsets so it looks organic). Events tagged ai_multiplier.
- *
- * HOW TO FIND IT IN MIXPANEL:
- *
- *   Report 1: AI Multiplier — Message Volume
- *   • Report type: Insights
- *   • Event: "message_sent"
- *   • Measure: Total (per user)
- *   • Breakdown: "ai_multiplier"
- *   • Expected: ai_multiplier=true users show ~3x higher message volume
- *     than ai_multiplier=false users
- *
- *   Report 2: AI Multiplier — Thread Engagement
- *   • Report type: Insights
- *   • Event: "thread_replied"
- *   • Measure: Total (per user)
- *   • Breakdown: "ai_multiplier"
- *   • Expected: ~3x higher thread reply count for AI-activated users
- *
- * REAL-WORLD ANALOGUE: Teams that discover AI summarization unlock faster
- * context-switching and communicate more frequently as a result.
- *
- * ───────────────────────────────────────────────────────────────────────────
- * 2. SEARCH EFFICIENCY (event + closure Map)
- * ───────────────────────────────────────────────────────────────────────────
- *
- * PATTERN: Natural-language AI searches lead to 40% faster result clicks
- * vs keyword searches. The event hook tracks the last ai_search_performed
- * query_type per user in a module-level Map, then reads it on the next
- * search_result_clicked to multiply time_to_click_ms by 0.6 for
- * natural_language queries.
- *
- * HOW TO FIND IT IN MIXPANEL:
- *
- *   Report 1: Search Click Speed by Query Type
- *   • Report type: Insights
- *   • Event: "search_result_clicked"
- *   • Measure: Average of "time_to_click_ms"
- *   • Breakdown: "query_type"
- *   • Expected: "natural_language" ≈ 3,000ms vs "keyword" ≈ 5,000ms (~40% faster)
- *
- *   Report 2: AI Search Adoption Funnel
- *   • Report type: Funnels
- *   • Steps: "ai_search_performed" → "search_result_clicked"
- *   • Breakdown: "is_ai_powered"
- *   • Expected: is_ai_powered=true converts at ~80%, false at ~60%
- *
- * REAL-WORLD ANALOGUE: Natural language search reduces cognitive load —
- * users find what they need faster because they can express intent in plain
- * English rather than crafting keyword strings.
- *
- * ───────────────────────────────────────────────────────────────────────────
- * 3. WORKSPACE STICKINESS (everything)
- * ───────────────────────────────────────────────────────────────────────────
- *
- * PATTERN: Workspaces where fewer than 20% of members use AI features
- * (ai_summary_generated or ai_search_performed) are "low-adoption." For
- * users in low-adoption workspaces, 30% of events after day 30 are removed,
- * simulating the retention gap between high- and low-AI workspaces.
- *
- * HOW TO FIND IT IN MIXPANEL:
- *
- *   Report 1: Retention by Workspace AI Adoption
- *   • Report type: Retention
- *   • Event: Any Event (first event = session_start)
- *   • Breakdown: Group Property "workspace_id" → "ai_adoption_tier"
- *   • Expected: High-adoption workspaces show ~95% MoM retention;
- *     low-adoption workspaces show ~70% (≈25pp gap)
- *
- *   Report 2: Event Volume Over Time
- *   • Report type: Insights (line chart)
- *   • Event: Any Event
- *   • Measure: Total
- *   • Breakdown: "workspace_id"
- *   • Filter: events after day 30
- *   • Expected: Low-adoption workspaces show visible event drop-off
- *
- * REAL-WORLD ANALOGUE: AI adoption is a network effect — one power user
- * sharing summaries pulls colleagues in. Below a critical mass threshold,
- * the workspace stagnates and churns.
- *
- * ───────────────────────────────────────────────────────────────────────────
- * 4. HUDDLE STICKINESS (everything)
- * ───────────────────────────────────────────────────────────────────────────
- *
- * PATTERN: Users who start a huddle with ai_notes_enabled=true are 2x more
- * likely to edit a Canvas within 1 hour afterward. If no canvas_edited
- * event exists within 60 minutes of the huddle, one is cloned from the
- * user's history and injected 20 minutes after the huddle_started event.
- *
- * HOW TO FIND IT IN MIXPANEL:
- *
- *   Report 1: Huddle → Canvas Funnel by AI Notes
- *   • Report type: Funnels
- *   • Steps: "huddle_started" → "canvas_edited"
- *   • Conversion window: 1 hour
- *   • Breakdown: "ai_notes_enabled"
- *   • Expected: ai_notes_enabled=true converts at ~40% vs ~20% for false
- *
- *   Report 2: Canvas Edits After Huddles
- *   • Report type: Insights
- *   • Event: "canvas_edited"
- *   • Measure: Total (per user)
- *   • Filter: Preceded by "huddle_started" within 1 hour
- *   • Breakdown: "ai_notes_enabled"
- *   • Expected: 2x more canvas edits for AI-notes users
- *
- * REAL-WORLD ANALOGUE: AI meeting notes lower the activation energy for
- * documenting outcomes — users don't have to reconstruct what was said,
- * so they're more likely to immediately capture action items in a Canvas.
- *
- * ───────────────────────────────────────────────────────────────────────────
- * 5. LURKER SUMMARIZATION PATTERN (everything)
- * ───────────────────────────────────────────────────────────────────────────
- *
- * PATTERN: Users with persona="Lurker" have 80% of their message_sent
- * events removed (they read, don't write) but get their ai_summary_generated
- * events quadrupled (they consume AI summaries voraciously). This creates a
- * clear behavioral segment: high summaries, low messages.
- *
- * HOW TO FIND IT IN MIXPANEL:
- *
- *   Report 1: AI Summary Usage by Persona
- *   • Report type: Insights
- *   • Event: "ai_summary_generated"
- *   • Measure: Total (per user)
- *   • Breakdown: User Property "persona"
- *   • Expected: "Lurker" shows ~4x higher summaries than "Power User"
- *     or "Traditionalist"
- *
- *   Report 2: Message Volume by Persona
- *   • Report type: Insights
- *   • Event: "message_sent"
- *   • Measure: Total (per user)
- *   • Breakdown: User Property "persona"
- *   • Expected: "Lurker" shows ~0.2x messages vs "Power User"
- *
- *   Report 3: Persona Behavioral Matrix
- *   • Report type: Insights
- *   • Events: "message_sent" + "ai_summary_generated" side by side
- *   • Measure: Total (per user)
- *   • Breakdown: User Property "persona"
- *   • Expected: Inverse relationship — Lurkers summarize more, send less
- *
- * REAL-WORLD ANALOGUE: In every Slack workspace, there are silent readers
- * who follow dozens of channels but rarely post. AI summaries are their
- * killer feature — they can stay informed without the noise.
- *
- * ───────────────────────────────────────────────────────────────────────────
- * 6. MONDAY PRODUCTIVITY SPIKE (event)
- * ───────────────────────────────────────────────────────────────────────────
- *
- * PATTERN: AI summary generation is 50% higher on Mondays as users catch up
- * from the weekend. When an ai_summary_generated event fires on a Monday,
- * the tokens_saved value is boosted by 1.5x and the event is cloned once
- * with a short time offset to double Monday volume.
- *
- * HOW TO FIND IT IN MIXPANEL:
- *
- *   Report 1: AI Summaries by Day of Week
- *   • Report type: Insights (bar chart)
- *   • Event: "ai_summary_generated"
- *   • Measure: Total
- *   • Time: Break down by Day of Week
- *   • Expected: Monday bar is visibly ~1.5x taller than Tue–Fri average
- *
- *   Report 2: Tokens Saved on Mondays
- *   • Report type: Insights
- *   • Event: "ai_summary_generated"
- *   • Measure: Average of "tokens_saved"
- *   • Breakdown: "monday_spike"
- *   • Expected: monday_spike=true shows ~1.5x higher avg tokens_saved
- *
- * REAL-WORLD ANALOGUE: Monday morning is "catch-up time" — users open Slack
- * and immediately want to know what happened over the weekend. AI summaries
- * are the fastest path from inbox-zero anxiety to being in the loop.
- *
- * ───────────────────────────────────────────────────────────────────────────
- * 7. AI WRITING ASSISTANCE VALUE (event)
- * ───────────────────────────────────────────────────────────────────────────
- *
- * PATTERN: Messages marked as is_ai_assisted=true have 2x higher
- * content_length. In the event hook, when message_sent fires with
- * is_ai_assisted=true, content_length is multiplied by 2.0.
- *
- * HOW TO FIND IT IN MIXPANEL:
- *
- *   Report 1: Message Length by AI Assistance
- *   • Report type: Insights
- *   • Event: "message_sent"
- *   • Measure: Average of "content_length"
- *   • Breakdown: "is_ai_assisted"
- *   • Expected: is_ai_assisted=true ≈ 300 chars vs false ≈ 150 chars (2x)
- *
- *   Report 2: AI Writing Adoption Over Time
- *   • Report type: Insights (line chart)
- *   • Event: "message_sent"
- *   • Measure: Total
- *   • Filter: "is_ai_assisted" = true
- *   • Time: Weekly
- *   • Expected: Growing trend as users discover the feature (growth soup)
- *
- * REAL-WORLD ANALOGUE: AI writing assistance helps users write more complete,
- * thoughtful messages — reducing back-and-forth and improving async
- * communication quality. Longer messages carry more context.
- *
- * ═══════════════════════════════════════════════════════════════════════════
- * EXPECTED METRICS SUMMARY
- * ═══════════════════════════════════════════════════════════════════════════
- *
- * Hook                  | Metric                | Baseline  | Effect    | Ratio
- * ──────────────────────|───────────────────────|───────────|───────────|──────
- * AI Multiplier         | message_sent/user/wk  | ~40       | ~120      | 3x
- * Search Efficiency     | time_to_click_ms      | ~5,000ms  | ~3,000ms  | 0.6x
- * Workspace Stickiness  | MoM retention         | ~70%      | ~95%      | 1.35x
- * Huddle Stickiness     | huddle→canvas conv    | ~20%      | ~40%      | 2x
- * Lurker Summarization  | summaries/user        | ~12       | ~48       | 4x
- * Monday Spike          | summaries on Monday   | baseline  | +50%      | 1.5x
- * AI Writing Value      | content_length (avg)  | ~150      | ~300      | 2x
- */
-
-/** @type {Config} */
-const config = {
-	token,
-	seed: SEED,
-	numDays: num_days,
-	numEvents: num_users * avg_events_per_user,
-	numUsers: num_users,
-	hasAnonIds: false,
-	hasSessionIds: true,
-	format: "json",
-	gzip: true,
-	alsoInferFunnels: false,
-	hasLocation: true,
-	hasAndroidDevices: true,
-	hasIOSDevices: true,
-	hasDesktopDevices: true,
-	hasBrowser: false,
-	hasCampaigns: false,
-	isAnonymous: false,
-	hasAdSpend: false,
-	percentUsersBornInDataset: 35,
-	hasAvatar: true,
-	batchSize: 2_500_000,
-	concurrency: 1,
-	writeToDisk: false,
-	scdProps: {},
-	mirrorProps: {},
-	lookupTables: [],
-
-	soup: "growth",
-
-	groupKeys: [
-		["workspace_id", 500, [
-			"message_sent",
-			"message_read",
-			"thread_replied",
-			"ai_summary_generated",
-			"ai_search_performed",
-			"huddle_started",
-			"huddle_joined",
-			"canvas_edited",
-			"channel_created",
-			"member_invited",
-			"workspace_created",
-		]],
-	],
-
-	groupProps: {
-		workspace_id: {
-			"workspace_name": () => `${chance.pickone(["Acme", "Globex", "Initech", "Umbrella", "Hooli", "Pied Piper", "Dunder Mifflin", "Vandelay", "Bluth", "Sterling"])} ${chance.pickone(["Corp", "Inc", "Labs", "HQ", "Team", "Group", "Co", "Ltd"])}`,
-			"workspace_plan": ["Pro", "Business+", "Business+", "Enterprise Grid"],
-			"member_count": u.weighNumRange(5, 500, 0.5, 30),
-			"ai_adoption_tier": ["low", "low", "medium", "high"],
-			"industry": ["Technology", "Finance", "Healthcare", "Retail", "Media", "Education", "Consulting"],
-		}
-	},
-
-	funnels: [
-		{
-			name: "Workspace Growth",
-			sequence: ["member_invited", "session_start", "ai_summary_generated"],
-			isFirstFunnel: true,
-			conversionRate: 60,
-			timeToConvert: 2,
-		},
-		{
-			name: "AI Adoption",
-			sequence: ["message_sent", "ai_prompt_viewed", "ai_summary_generated"],
-			conversionRate: 25,
-			timeToConvert: 0.1,
-			weight: 10,
-		},
-		{
-			name: "Collaboration Loop",
-			sequence: ["huddle_joined", "canvas_edited"],
-			conversionRate: 40,
-			timeToConvert: 0.04,
-		},
-		{
-			name: "Intelligent Search",
-			sequence: ["ai_search_performed", "search_result_clicked"],
-			conversionRate: 80,
-			timeToConvert: 0.01,
-		},
-		{
-			name: "Onboarding",
-			sequence: ["workspace_created", "channel_created", "member_invited"],
-			conversionRate: 45,
-			timeToConvert: 1,
-		},
-	],
-
-	events: [
-		{
-			event: "workspace_created",
-			weight: 1,
-			isFirstEvent: true,
-			properties: {
-				"template_used": ["blank", "engineering", "sales", "marketing", "hr", "product"],
-				"import_from": ["none", "none", "none", "microsoft_teams", "google_chat"],
-			}
-		},
-		{
-			event: "member_invited",
-			weight: 3,
-			isFirstEvent: true,
-			properties: {
-				"invite_method": ["email", "link", "directory_sync"],
-				"invitee_role": ["member", "member", "member", "admin", "guest"],
-			}
-		},
-		{
-			event: "session_start",
-			weight: 20,
-			isSessionStartEvent: true,
-			properties: {
-				"session_duration_sec": u.weighNumRange(60, 7200, 0.5, 900),
-				"notifications_pending": u.weighNumRange(0, 50, 0.8, 5),
-			}
-		},
-		{
-			event: "message_sent",
-			weight: 18,
-			properties: {
-				"channel_type": ["public", "private", "dm"],
-				"content_length": u.weighNumRange(10, 500, 0.5, 80),
-				"is_ai_assisted": u.pickAWinner([true, false], 0.25),
-				"has_attachment": u.pickAWinner([true, false], 0.2),
-				"ai_multiplier": [false],
-			}
-		},
-		{
-			event: "message_read",
-			weight: 15,
-			properties: {
-				"source": ["mobile", "desktop"],
-				"unread_count": u.weighNumRange(1, 100, 0.8, 10),
-				"channel_type": ["public", "private", "dm"],
-			}
-		},
-		{
-			event: "thread_replied",
-			weight: 10,
-			properties: {
-				"is_ai_assisted": u.pickAWinner([true, false], 0.2),
-				"reply_length": u.weighNumRange(5, 300, 0.5, 60),
-				"thread_depth": u.weighNumRange(1, 20, 0.8, 3),
-				"ai_multiplier": [false],
-			}
-		},
-		{
-			event: "reaction_added",
-			weight: 12,
-			properties: {
-				"emoji": ["+1", "clap", "fire", "eyes", "check"],
-				"message_age_hours": u.weighNumRange(0, 72, 0.8, 2),
-			}
-		},
-		{
-			event: "ai_prompt_viewed",
-			weight: 8,
-			properties: {
-				"feature": ["summarize", "search", "compose"],
-				"entry_point": ["channel_header", "search_bar", "compose_box", "shortcut"],
-				"converted": u.pickAWinner([true, false], 0.6),
-			}
-		},
-		{
-			event: "ai_summary_generated",
-			weight: 6,
-			properties: {
-				"source_type": ["channel", "thread", "huddle"],
-				"tokens_saved": u.weighNumRange(50, 2000, 0.5, 400),
-				"messages_summarized": u.weighNumRange(5, 200, 0.8, 25),
-				"monday_spike": [false],
-			}
-		},
-		{
-			event: "ai_search_performed",
-			weight: 7,
-			properties: {
-				"query_type": ["natural_language", "keyword"],
-				"is_ai_powered": u.pickAWinner([true, false], 0.55),
-				"results_count": u.weighNumRange(0, 50, 0.5, 12),
-				"query_length_chars": u.weighNumRange(5, 200, 0.5, 40),
-			}
-		},
-		{
-			event: "ai_writing_assisted",
-			weight: 4,
-			properties: {
-				"tone": ["professional", "casual", "concise"],
-				"words_suggested": u.weighNumRange(5, 100, 0.5, 25),
-				"accepted": u.pickAWinner([true, false], 0.7),
-			}
-		},
-		{
-			event: "search_result_clicked",
-			weight: 6,
-			properties: {
-				"time_to_click_ms": u.weighNumRange(1000, 15000, 1, 5000),
-				"position": u.weighNumRange(1, 10, 0.8, 3),
-				"result_type": ["message", "file", "channel", "person"],
-				"query_type": ["natural_language", "keyword"],
-			}
-		},
-		{
-			event: "huddle_started",
-			weight: 4,
-			properties: {
-				"ai_notes_enabled": u.pickAWinner([true, false], 0.4),
-				"is_scheduled": u.pickAWinner([true, false], 0.35),
-				"participant_count": u.weighNumRange(2, 15, 0.8, 4),
-				"duration_min": u.weighNumRange(5, 90, 0.5, 20),
-			}
-		},
-		{
-			event: "huddle_joined",
-			weight: 8,
-			properties: {
-				"join_method": ["link", "notification", "scheduled"],
-				"device_type": ["desktop", "mobile", "tablet"],
-			}
-		},
-		{
-			event: "canvas_edited",
-			weight: 6,
-			properties: {
-				"edit_type": ["text", "table", "checklist", "image"],
-				"canvas_age_days": u.weighNumRange(0, 90, 0.8, 7),
-				"collaborators_count": u.weighNumRange(1, 10, 0.8, 2),
-			}
-		},
-		{
-			event: "file_shared",
-			weight: 5,
-			properties: {
-				"extension": ["pdf", "docx", "png", "csv"],
-				"file_size_kb": u.weighNumRange(10, 50000, 0.3, 500),
-				"channel_type": ["public", "private", "dm"],
-			}
-		},
-		{
-			event: "channel_created",
-			weight: 2,
-			properties: {
-				"is_private": u.pickAWinner([true, false], 0.4),
-				"purpose_set": u.pickAWinner([true, false], 0.6),
-				"channel_type": ["topic", "project", "team", "social"],
-			}
-		},
-		{
-			event: "integration_added",
-			weight: 2,
-			properties: {
-				"category": ["crm", "dev_tools", "hr", "productivity"],
-				"integration_name": ["Salesforce", "GitHub", "Jira", "Zoom", "Google Drive", "Notion", "PagerDuty", "Figma"],
-				"installed_by_role": ["admin", "admin", "member"],
-			}
-		},
-		{
-			event: "user_churned",
-			weight: 0.5,
-			isChurnEvent: true,
-			isStrictEvent: true,
-			returnLikelihood: 0.1,
-			properties: {
-				"churn_reason": ["inactivity", "workspace_deleted", "plan_downgrade", "competitor"],
-				"days_active": u.weighNumRange(1, 90, 0.5, 20),
-			}
-		},
-	],
-
-	superProps: {
-		workspace_id: () => `T-${v.uid(8).toUpperCase()}`,
-		account_tier: ["Pro", "Pro", "Business+", "Business+", "Enterprise Grid"],
-		platform: ["macOS", "Windows", "iOS", "Android", "Web"],
-	},
-
-	userProps: {
-		"persona": ["Power User", "Power User", "Traditionalist", "Traditionalist", "Traditionalist", "Lurker"],
-		"job_role": ["Engineering", "Product", "Sales", "Marketing", "HR"],
-		"ai_onboarding_complete": u.pickAWinner([true, false], 0.55),
-		"department_size": u.weighNumRange(3, 200, 0.5, 20),
-		"tenure_days": u.weighNumRange(1, 730, 0.3, 90),
-	},
-
-	hook: function (record, type, meta) {
-		const NOW = dayjs();
-		const DATASET_START = NOW.subtract(num_days, "days");
-
-		// ═══════════════════════════════════════════════════════════════════
-		// HOOKS 2, 6, 7: EVENT HOOK
-		// Per-event property modifications and closure-based state tracking.
-		// ═══════════════════════════════════════════════════════════════════
-		if (type === "event") {
-			const EVENT_TIME = dayjs(record.time);
-			const dayOfWeek = EVENT_TIME.day();
-			const userId = record.user_id;
-
-			// ── Hook 2: SEARCH EFFICIENCY — track query type via closure Map ──
-			// Store the query_type of ai_search_performed so the next
-			// search_result_clicked can read it and apply the speed multiplier.
-			if (record.event === "ai_search_performed") {
-				lastSearchQueryType.set(userId, record.query_type);
-			}
-
-			// ── Hook 2: SEARCH EFFICIENCY — apply 40% speed boost on click ──
-			// natural_language → time_to_click_ms * 0.6 (40% faster)
-			if (record.event === "search_result_clicked") {
-				const lastQueryType = lastSearchQueryType.get(userId);
-				if (lastQueryType) {
-					record.query_type = lastQueryType;
-					if (lastQueryType === "natural_language") {
-						record.time_to_click_ms = Math.round(record.time_to_click_ms * 0.6);
-					}
-					lastSearchQueryType.delete(userId);
-				}
-			}
-
-			// ── Hook 6: MONDAY PRODUCTIVITY SPIKE — boost tokens_saved 1.5x ──
-			// Tag monday_spike=true; the everything hook will clone these events
-			// to produce the observable 1.5x volume increase on Mondays.
-			if (record.event === "ai_summary_generated") {
-				if (dayOfWeek === 1) {
-					record.tokens_saved = Math.round(record.tokens_saved * 1.5);
-					record.monday_spike = true;
-				} else {
-					record.monday_spike = false;
-				}
-			}
-
-			// ── Hook 7: AI WRITING VALUE — double content_length for AI messages ──
-			// is_ai_assisted=true messages get 2x content_length (capped at 500).
-			if (record.event === "message_sent" && record.is_ai_assisted === true) {
-				record.content_length = Math.min(500, Math.round(record.content_length * 2.0));
-			}
-
-			return record;
-		}
-
-		// ═══════════════════════════════════════════════════════════════════
-		// HOOKS 1, 3, 4, 5, 6 (clone pass): EVERYTHING HOOK
-		// Full event stream per user. meta.profile available for cross-table
-		// correlation. Runs after all events for one user are generated.
-		// ═══════════════════════════════════════════════════════════════════
-		if (type === "everything") {
-			const events = record;
-			if (!events || events.length === 0) return record;
-
-			const profile = meta && meta.profile ? meta.profile : {};
-			const persona = profile.persona || "Traditionalist";
-			const firstEventTime = dayjs(events[0].time);
-
-			// ── First pass: collect behavioral signals ──────────────────
-			const aiSummaryTimestamps = [];
-			const huddlesWithAiNotes = [];
-			const workspaceId = events[0] ? events[0].workspace_id : null;
-
-			events.forEach((event, idx) => {
-				if (event.event === "ai_summary_generated") {
-					aiSummaryTimestamps.push({ time: dayjs(event.time), idx });
-				}
-				if (event.event === "huddle_started" && event.ai_notes_enabled === true) {
-					huddlesWithAiNotes.push({ time: dayjs(event.time), idx });
-				}
-			});
-
-			// ── Track workspace AI adoption ratio across all users ──────
-			// Used by Hook 3 to determine low-adoption workspaces.
-			if (workspaceId) {
-				if (!workspaceTotalUsers.has(workspaceId)) {
-					workspaceTotalUsers.set(workspaceId, 0);
-					workspaceAiUserCount.set(workspaceId, 0);
-				}
-				workspaceTotalUsers.set(workspaceId, workspaceTotalUsers.get(workspaceId) + 1);
-				if (aiSummaryTimestamps.length > 0) {
-					workspaceAiUserCount.set(workspaceId, workspaceAiUserCount.get(workspaceId) + 1);
-				}
-			}
-
-			// ═══════════════════════════════════════════════════════════
-			// HOOK 1: THE AI MULTIPLIER
-			// Find 7-day windows with 5+ summaries → clone message_sent
-			// and thread_replied into the next 7 days (3x volume boost).
-			// ═══════════════════════════════════════════════════════════
-			const templateMsg = events.find(e => e.event === "message_sent");
-			const templateThread = events.find(e => e.event === "thread_replied");
-			const qualifyingWindowStarts = new Set();
-
-			for (let i = 0; i < aiSummaryTimestamps.length; i++) {
-				const windowStart = aiSummaryTimestamps[i].time;
-				const windowEnd = windowStart.add(7, "days");
-				let countInWindow = 0;
-				for (let j = i; j < aiSummaryTimestamps.length; j++) {
-					if (aiSummaryTimestamps[j].time.isBefore(windowEnd)) {
-						countInWindow++;
-					} else {
-						break;
-					}
-				}
-				if (countInWindow >= 5) {
-					qualifyingWindowStarts.add(windowStart.valueOf());
-				}
-			}
-
-			if (qualifyingWindowStarts.size > 0) {
-				const injected = [];
-				qualifyingWindowStarts.forEach(windowStartMs => {
-					const boostStart = dayjs(windowStartMs).add(7, "days");
-
-					// Clone 6 message_sent events with organic jitter (≈3x boost)
-					if (templateMsg) {
-						for (let k = 0; k < 6; k++) {
-							injected.push({
-								...templateMsg,
-								time: boostStart.add(chance.integer({ min: 10, max: 9000 }), "minutes").toISOString(),
-								user_id: templateMsg.user_id,
-								content_length: Math.round(templateMsg.content_length * (0.8 + Math.random() * 0.5)),
-								is_ai_assisted: chance.bool({ likelihood: 40 }),
-								ai_multiplier: true,
-							});
-						}
-					}
-
-					// Clone 3 thread_replied events
-					if (templateThread) {
-						for (let k = 0; k < 3; k++) {
-							injected.push({
-								...templateThread,
-								time: boostStart.add(chance.integer({ min: 15, max: 9000 }), "minutes").toISOString(),
-								user_id: templateThread.user_id,
-								reply_length: Math.round(templateThread.reply_length * (0.8 + Math.random() * 0.5)),
-								ai_multiplier: true,
-							});
-						}
-					}
-				});
-
-				if (injected.length > 0) {
-					events.push(...injected);
-				}
-			}
-
-			// ═══════════════════════════════════════════════════════════
-			// HOOK 3: WORKSPACE STICKINESS
-			// Low-adoption workspaces (<20% AI users) lose 30% of events
-			// after day 30 — simulating the retention gap.
-			// ═══════════════════════════════════════════════════════════
-			const totalInWs = workspaceId ? (workspaceTotalUsers.get(workspaceId) || 1) : 1;
-			const aiInWs = workspaceId ? (workspaceAiUserCount.get(workspaceId) || 0) : 0;
-			const isLowAdoption = (aiInWs / totalInWs) < 0.20;
-
-			if (isLowAdoption) {
-				const day30 = firstEventTime.add(30, "days");
-				for (let i = events.length - 1; i >= 0; i--) {
-					if (dayjs(events[i].time).isAfter(day30) && chance.bool({ likelihood: 30 })) {
-						events.splice(i, 1);
-					}
-				}
-			}
-
-			// ═══════════════════════════════════════════════════════════
-			// HOOK 4: HUDDLE STICKINESS
-			// For each huddle with ai_notes_enabled=true, inject a
-			// canvas_edited event 20 minutes later if none exists in
-			// the 60-minute window — producing 2x funnel conversion.
-			// ═══════════════════════════════════════════════════════════
-			const templateCanvas = events.find(e => e.event === "canvas_edited");
-
-			huddlesWithAiNotes.forEach(huddle => {
-				const huddleTime = huddle.time;
-				const windowEnd = huddleTime.add(60, "minutes");
-
-				const hasNearbyCanvas = events.some(e =>
-					e.event === "canvas_edited" &&
-					dayjs(e.time).isAfter(huddleTime) &&
-					dayjs(e.time).isBefore(windowEnd)
-				);
-
-				// Inject if no canvas nearby, or 50% of the time even if one exists
-				// (the 50% case pushes the ratio toward 2x)
-				if (templateCanvas && (!hasNearbyCanvas || chance.bool({ likelihood: 50 }))) {
-					events.push({
-						...templateCanvas,
-						time: huddleTime.add(20, "minutes").toISOString(),
-						user_id: templateCanvas.user_id,
-						edit_type: chance.pickone(["text", "checklist"]),
-						collaborators_count: chance.integer({ min: 1, max: 4 }),
-					});
-				}
-			});
-
-			// ═══════════════════════════════════════════════════════════
-			// HOOK 5: LURKER SUMMARIZATION PATTERN
-			// Lurkers: drop 80% of message_sent, quadruple ai_summary_generated.
-			// Creates the inverse behavioral signature: high summaries, low messages.
-			// ═══════════════════════════════════════════════════════════
-			if (persona === "Lurker") {
-				// Remove 80% of message_sent events (iterate backwards)
-				for (let i = events.length - 1; i >= 0; i--) {
-					if (events[i].event === "message_sent" && chance.bool({ likelihood: 80 })) {
-						events.splice(i, 1);
-					}
-				}
-
-				// Clone 3 extra summaries per existing one (= 4x total)
-				const summarySnapshot = events.filter(e => e.event === "ai_summary_generated");
-				const extraSummaries = [];
-				summarySnapshot.forEach(s => {
-					for (let k = 0; k < 3; k++) {
-						extraSummaries.push({
-							...s,
-							time: dayjs(s.time).add(chance.integer({ min: 1, max: 48 }), "hours").toISOString(),
-							user_id: s.user_id,
-							source_type: chance.pickone(["channel", "thread", "huddle"]),
-							tokens_saved: chance.integer({ min: 100, max: 1500 }),
-							messages_summarized: chance.integer({ min: 10, max: 80 }),
-							monday_spike: false,
-						});
-					}
-				});
-				if (extraSummaries.length > 0) {
-					events.push(...extraSummaries);
-				}
-			}
-
-			// ═══════════════════════════════════════════════════════════
-			// HOOK 6 (clone pass): MONDAY PRODUCTIVITY SPIKE
-			// Clone each monday_spike=true summary once with a short offset
-			// to produce the observable 1.5x volume effect on Mondays.
-			// ═══════════════════════════════════════════════════════════
-			const mondaySummaries = events.filter(
-				e => e.event === "ai_summary_generated" && e.monday_spike === true
-			);
-			const mondayClones = [];
-			mondaySummaries.forEach(s => {
-				if (chance.bool({ likelihood: 50 })) {
-					mondayClones.push({
-						...s,
-						time: dayjs(s.time).add(chance.integer({ min: 15, max: 120 }), "minutes").toISOString(),
-						user_id: s.user_id,
-						tokens_saved: Math.round(s.tokens_saved * (0.8 + Math.random() * 0.4)),
-						messages_summarized: chance.integer({ min: 5, max: 50 }),
-					});
-				}
-			});
-			if (mondayClones.length > 0) {
-				events.push(...mondayClones);
-			}
-
-			return record;
-		}
-
-		return record;
-	},
-};
-
-export default config;