@ak--47/dungeon-master 1.2.3 → 1.3.0

package/dungeons/vertical/ai-platform.js

@@ -0,0 +1,799 @@
+// ── TWEAK THESE ──
+const SEED = "promptforge";
+const num_days = 120;
+const num_users = 8_000;
+const avg_events_per_user_per_day = 0.83;
+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 "dotenv/config";
+import * as u from "../../lib/utils/utils.js";
+import * as v from "ak-tools";
+
+dayjs.extend(utc);
+const chance = u.initChance(SEED);
+const NOW = dayjs();
+const DATASET_START = NOW.subtract(num_days, "days");
+
+/** @typedef  {import("../../types").Dungeon} Config */
+
+/**
+ * ===============================================================
+ * DATASET OVERVIEW
+ * ===============================================================
+ *
+ * PromptForge -- an LLM API platform (like Anthropic/OpenAI).
+ * Customers (developers and companies) send API requests for chat
+ * completions, embeddings, evaluations, and tool use. Billing is
+ * per input/output token. Key features: prompt caching, tool use,
+ * multi-turn conversations, batch API, model selection, and
+ * evaluation pipelines.
+ *
+ * - 8,000 users over 120 days, ~800K events
+ * - Three API tiers: Free, Build, Enterprise
+ * - Core loop: org created -> api key created -> api call -> iterate
+ * - Revenue: token-based billing with tier-based pricing
+ *
+ * Key entities:
+ * - model: LLM model version (sonnet-4, haiku-4, opus-4-6, opus-4-7)
+ * - api_tier: Free / Build / Enterprise (determines context window, rate limits)
+ * - tokens_used: total tokens consumed per API call (input + output)
+ * - cost_usd: dollar cost of a single API call
+ * - cache_enabled: prompt caching flag that reduces cost 70%
+ * - multi_turn: whether the call is part of a conversation
+ *
+ * ===============================================================
+ * ANALYTICS HOOKS (8 hooks)
+ * ===============================================================
+ *
+ * ---------------------------------------------------------------
+ * 1. PROMPT CACHING ADOPTION (CONVERSION — event + everything)
+ * ---------------------------------------------------------------
+ *
+ * PATTERN: Customers who enable prompt caching see 70% lower
+ * cost_per_call. Once any api call has cache_enabled=true, all
+ * subsequent calls for that user get cost_usd reduced by 70%.
+ *
+ * HOW TO FIND IT IN MIXPANEL:
+ *
+ *   Report 1: Cost Per Call by Cache Status
+ *   - Report type: Insights
+ *   - Event: "api call"
+ *   - Measure: Average of "cost_usd"
+ *   - Breakdown: "cache_enabled"
+ *   - Expected: cache_enabled=true ~ $0.003, false ~ $0.01 (70% cheaper)
+ *
+ *   Report 2: Cache Adoption Over Time
+ *   - Report type: Insights
+ *   - Event: "api call"
+ *   - Measure: Total
+ *   - Filter: cache_enabled = true
+ *   - Line chart by week
+ *   - Expected: steady growth in cached calls over the dataset
+ *
+ * REAL-WORLD ANALOGUE: Prompt caching avoids re-processing long
+ * system prompts on every call, dramatically reducing cost and latency.
+ *
+ * ---------------------------------------------------------------
+ * 2. MODEL MIGRATION WAVE (TIMED RELEASE — event)
+ * ---------------------------------------------------------------
+ *
+ * PATTERN: At day 60, new model "opus-4-7" releases. After day 60,
+ * 35% of api calls from Build/Enterprise users switch model to
+ * "opus-4-7". These calls use 1.5x tokens (smarter model, longer
+ * responses).
+ *
+ * HOW TO FIND IT IN MIXPANEL:
+ *
+ *   Report 1: Model Distribution Over Time
+ *   - Report type: Insights
+ *   - Event: "api call"
+ *   - Measure: Total
+ *   - Breakdown: "model"
+ *   - Line chart by week
+ *   - Expected: opus-4-7 appears at day 60, ramps to ~35% of paid calls
+ *
+ *   Report 2: Tokens Per Model
+ *   - Report type: Insights
+ *   - Event: "api call"
+ *   - Measure: Average of "tokens_used"
+ *   - Breakdown: "model"
+ *   - Expected: opus-4-7 ~ 1.5x tokens vs other models
+ *
+ * REAL-WORLD ANALOGUE: New flagship model launches cause migration
+ * waves among power users who want improved capabilities.
+ *
+ * ---------------------------------------------------------------
+ * 3. AGENTIC LOOP POWER USERS (BEHAVIORS TOGETHER — everything)
+ * ---------------------------------------------------------------
+ *
+ * PATTERN: Users who use both "tool use call" AND have multi_turn=true
+ * on any api call are agentic loop users. They get 8x tokens_used on
+ * all api calls and 3x extra api call events injected.
+ *
+ * HOW TO FIND IT IN MIXPANEL:
+ *
+ *   Report 1: Token Usage — Agentic vs Standard
+ *   - Report type: Insights
+ *   - Event: "api call"
+ *   - Measure: Average of "tokens_used"
+ *   - Breakdown: "is_agentic_user"
+ *   - Expected: is_agentic_user=true ~ 8x tokens (agentic ~ 40K, standard ~ 5K)
+ *
+ *   Report 2: API Call Volume — Agentic vs Standard
+ *   - Report type: Insights
+ *   - Event: "api call"
+ *   - Measure: Total per user (average)
+ *   - Breakdown: "is_agentic_user"
+ *   - Expected: agentic users ~ 3x more api calls
+ *
+ * REAL-WORLD ANALOGUE: Agentic workloads (coding agents, research
+ * assistants) consume dramatically more tokens via extended tool-use
+ * loops and multi-turn conversations.
+ *
+ * ---------------------------------------------------------------
+ * 4. RATE LIMIT CHURN (CHURN — everything)
+ * ---------------------------------------------------------------
+ *
+ * PATTERN: Users hitting "rate limit error" >= 5 times in first
+ * 7 days lose 60% of events after week 1. Rate-limited users
+ * churn from frustration.
+ *
+ * HOW TO FIND IT IN MIXPANEL:
+ *
+ *   Report 1: Retention by Early Rate Limiting
+ *   - Report type: Retention
+ *   - Event A: any event
+ *   - Event B: any event
+ *   - Breakdown: "hit_rate_limit_early" (user property)
+ *   - Expected: hit_rate_limit_early=true ~ 40% D30 retention
+ *     vs ~80% for others
+ *
+ *   Report 2: Event Volume Post Rate-Limit
+ *   - Report type: Insights
+ *   - Event: any event
+ *   - Measure: Total per user (average)
+ *   - Breakdown: "hit_rate_limit_early"
+ *   - Expected: rate-limited users ~ 40% of normal volume
+ *
+ * REAL-WORLD ANALOGUE: Developers who get rate-limited early in
+ * their evaluation often switch to a competitor platform.
+ *
+ * ---------------------------------------------------------------
+ * 5. TIER-BASED CONTEXT WINDOW (SUBSCRIPTION TIER — everything)
+ * ---------------------------------------------------------------
+ *
+ * PATTERN: Free users have context_window=200000, Build=1000000,
+ * Enterprise=2000000. Enterprise users send 4x larger input_tokens.
+ * Context window and input tokens are scaled by tier.
+ *
+ * HOW TO FIND IT IN MIXPANEL:
+ *
+ *   Report 1: Input Tokens by Tier
+ *   - Report type: Insights
+ *   - Event: "api call"
+ *   - Measure: Average of "input_tokens"
+ *   - Breakdown: "api_tier" (superProp)
+ *   - Expected: Enterprise ~ 4x Free (Enterprise ~ 8K, Free ~ 2K)
+ *
+ *   Report 2: Context Window by Tier
+ *   - Report type: Insights
+ *   - Event: "api call"
+ *   - Measure: Average of "context_window"
+ *   - Breakdown: "api_tier"
+ *   - Expected: Free=200K, Build=1M, Enterprise=2M
+ *
+ * REAL-WORLD ANALOGUE: Enterprise customers pay for larger context
+ * windows and use them for long-document analysis and code review.
+ *
+ * ---------------------------------------------------------------
+ * 6. OUTAGE DAY (TIME-BASED — event)
+ * ---------------------------------------------------------------
+ *
+ * PATTERN: Days 40-41, is_error is set to true on 40% of api call
+ * events. error_type is set to service errors. Simulates a major
+ * platform outage.
+ *
+ * HOW TO FIND IT IN MIXPANEL:
+ *
+ *   Report 1: Error Rate Over Time
+ *   - Report type: Insights
+ *   - Event: "api call"
+ *   - Measure: Total
+ *   - Filter: is_error = true
+ *   - Line chart by day
+ *   - Expected: massive spike on days 40-41 (8x baseline error rate)
+ *
+ *   Report 2: Error Types During Outage
+ *   - Report type: Insights
+ *   - Event: "api call"
+ *   - Filter: is_error = true
+ *   - Breakdown: "error_type"
+ *   - Date range: days 40-41
+ *   - Expected: service_overloaded and internal_server_error dominate
+ *
+ * REAL-WORLD ANALOGUE: API platforms experience periodic outages
+ * that spike error rates across all customers.
+ *
+ * ---------------------------------------------------------------
+ * 7. BATCH API DISCOUNT (PURCHASE VALUE — everything)
+ * ---------------------------------------------------------------
+ *
+ * PATTERN: Users who submit batch jobs get 50% lower cost_per_token
+ * on api calls but use 2x tokens_used. Batch processing is cheaper
+ * per token but encourages higher volume.
+ *
+ * HOW TO FIND IT IN MIXPANEL:
+ *
+ *   Report 1: Cost Per Token — Batch vs Interactive
+ *   - Report type: Insights
+ *   - Event: "api call"
+ *   - Measure: Average of "cost_per_token"
+ *   - Breakdown: "is_batch_user"
+ *   - Expected: is_batch_user=true ~ 50% lower cost per token
+ *
+ *   Report 2: Token Volume — Batch Users
+ *   - Report type: Insights
+ *   - Event: "api call"
+ *   - Measure: Average of "tokens_used"
+ *   - Breakdown: "is_batch_user"
+ *   - Expected: batch users ~ 2x token volume
+ *
+ * REAL-WORLD ANALOGUE: Batch API pricing incentivizes high-volume
+ * workloads with discounted per-token rates.
+ *
+ * ---------------------------------------------------------------
+ * 8. EVAL-DRIVEN RETENTION (RETENTION — everything)
+ * ---------------------------------------------------------------
+ *
+ * PATTERN: Users who run "eval job" in the first 7 days have 75%
+ * D30 retention vs 25% for non-eval users. Early eval adoption
+ * indicates serious platform investment.
+ *
+ * HOW TO FIND IT IN MIXPANEL:
+ *
+ *   Report 1: Retention by Early Eval Usage
+ *   - Report type: Retention
+ *   - Event A: any event
+ *   - Event B: any event
+ *   - Breakdown: "has_early_eval" (user property)
+ *   - Expected: has_early_eval=true ~ 75% D30 vs 25% for false
+ *
+ *   Report 2: Event Volume Over Time
+ *   - Report type: Insights
+ *   - Event: any event
+ *   - Measure: Total per user (average)
+ *   - Breakdown: "has_early_eval"
+ *   - Line chart by week
+ *   - Expected: early eval users sustain volume; non-eval users decay
+ *
+ * REAL-WORLD ANALOGUE: Teams that set up evaluation pipelines early
+ * are deeply invested in prompt quality and stick with the platform.
+ *
+ * ===============================================================
+ * EXPECTED METRICS SUMMARY
+ * ===============================================================
+ *
+ * Hook                        | Metric              | Baseline   | Effect       | Ratio
+ * ----------------------------|----------------------|------------|--------------|------
+ * Prompt Caching Adoption     | cost_usd             | $0.01      | $0.003       | 0.3x
+ * Model Migration Wave        | opus-4-7 share       | 0%         | ~35% (paid)  | new
+ * Agentic Loop Power Users    | tokens_used          | 5K         | 40K          | 8x
+ * Rate Limit Churn            | D30 retention        | 80%        | 40%          | 0.5x
+ * Tier-Based Context Window   | input_tokens         | 2K (Free)  | 8K (Ent)     | 4x
+ * Outage Day                  | error rate           | 5%         | 40%          | 8x
+ * Batch API Discount          | cost_per_token       | $0.00001   | $0.000005    | 0.5x
+ * Eval-Driven Retention       | D30 retention        | 25%        | 75%          | 3x
+ */
+
+/** @type {Config} */
+const config = {
+	token,
+	seed: SEED,
+	numDays: num_days,
+	avgEventsPerUserPerDay: avg_events_per_user_per_day,
+	numUsers: num_users,
+	hasAnonIds: false,
+	hasSessionIds: false,
+	format: "json",
+	gzip: true,
+	alsoInferFunnels: false,
+	hasLocation: true,
+	hasAndroidDevices: false,
+	hasIOSDevices: false,
+	hasDesktopDevices: true,
+	hasBrowser: true,
+	hasCampaigns: false,
+	isAnonymous: false,
+	hasAdSpend: false,
+	hasAvatar: true,
+	concurrency: 1,
+	writeToDisk: false,
+
+	soup: "growth",
+
+	scdProps: {
+		monthly_api_usage: {
+			values: u.weighNumRange(0, 1000000, 0.3, 50),
+			frequency: "week",
+			timing: "fuzzy",
+			max: 20,
+		},
+		api_tier_history: {
+			values: ["Free", "Build", "Enterprise"],
+			frequency: "month",
+			timing: "fixed",
+			max: 6,
+		},
+	},
+
+	// -- Events (18) ------------------------------------------
+	events: [
+		{
+			event: "organization created",
+			weight: 1,
+			isFirstEvent: true,
+			properties: {
+				org_size: ["solo", "startup", "growth", "enterprise"],
+				referral_source: ["docs", "blog", "github", "word_of_mouth", "search", "conference"],
+			},
+		},
+		{
+			event: "api key created",
+			weight: 2,
+			properties: {
+				key_type: ["development", "production", "staging"],
+				key_scope: ["full_access", "read_only", "completions_only"],
+			},
+		},
+		{
+			event: "api key rotated",
+			weight: 1,
+			properties: {
+				rotation_reason: ["scheduled", "compromised", "policy", "manual"],
+			},
+		},
+		{
+			event: "api call",
+			weight: 10,
+			properties: {
+				model: ["sonnet-4", "sonnet-4", "sonnet-4", "haiku-4", "haiku-4", "opus-4-6"],
+				input_tokens: u.weighNumRange(50, 8000, 0.4, 2000),
+				output_tokens: u.weighNumRange(10, 4000, 0.4, 500),
+				tokens_used: u.weighNumRange(100, 12000, 0.4, 2500),
+				cost_usd: [0.001, 0.002, 0.003, 0.003, 0.005, 0.005, 0.005, 0.008, 0.008, 0.01, 0.01, 0.01, 0.01, 0.015, 0.015, 0.02, 0.025, 0.03, 0.04, 0.05],
+				cost_per_token: [0.000002, 0.000003, 0.000005, 0.000005, 0.000008, 0.000008, 0.00001, 0.00001, 0.00001, 0.000012, 0.000015, 0.00002, 0.000025, 0.00003],
+				latency_ms: u.weighNumRange(100, 15000, 0.4, 1500),
+				cache_enabled: [false],
+				is_error: [false],
+				error_type: ["none"],
+				multi_turn: [false, false, false, true],
+				context_window: [200000],
+				is_agentic_user: [false],
+				is_batch_user: [false],
+				stream: [true, true, true, false],
+				stop_reason: ["end_turn", "end_turn", "end_turn", "max_tokens", "tool_use"],
+			},
+		},
+		{
+			event: "tool use call",
+			weight: 4,
+			properties: {
+				tool_name: ["web_search", "code_interpreter", "file_reader", "calculator", "database_query", "api_connector"],
+				execution_time_ms: u.weighNumRange(50, 10000, 0.4, 800),
+				success: [true, true, true, true, false],
+				tool_input_tokens: u.weighNumRange(50, 2000, 0.4, 300),
+				tool_output_tokens: u.weighNumRange(20, 5000, 0.4, 500),
+			},
+		},
+		{
+			event: "batch job submitted",
+			weight: 2,
+			properties: {
+				batch_size: u.weighNumRange(10, 10000, 0.3, 500),
+				model: ["sonnet-4", "haiku-4", "opus-4-6"],
+				estimated_tokens: u.weighNumRange(10000, 5000000, 0.3, 500000),
+				priority: ["standard", "standard", "standard", "express"],
+			},
+		},
+		{
+			event: "batch job completed",
+			weight: 2,
+			properties: {
+				batch_size: u.weighNumRange(10, 10000, 0.3, 500),
+				processing_time_sec: u.weighNumRange(60, 7200, 0.4, 900),
+				total_tokens: u.weighNumRange(10000, 5000000, 0.3, 500000),
+				success_rate: u.weighNumRange(90, 100, 0.8, 98),
+			},
+		},
+		{
+			event: "eval job",
+			weight: 3,
+			properties: {
+				eval_type: ["accuracy", "relevance", "safety", "latency", "cost", "custom"],
+				num_test_cases: u.weighNumRange(10, 1000, 0.3, 100),
+				model: ["sonnet-4", "haiku-4", "opus-4-6"],
+				dataset_name: ["prod_prompts", "safety_suite", "regression_set", "benchmark_v2", "custom_eval"],
+			},
+		},
+		{
+			event: "eval result",
+			weight: 3,
+			properties: {
+				eval_type: ["accuracy", "relevance", "safety", "latency", "cost", "custom"],
+				score: u.weighNumRange(0, 100, 0.6, 75),
+				pass_rate: u.weighNumRange(50, 100, 0.7, 85),
+				model: ["sonnet-4", "haiku-4", "opus-4-6"],
+				regression_detected: [false, false, false, false, true],
+			},
+		},
+		{
+			event: "rate limit error",
+			weight: 3,
+			properties: {
+				error_code: [429],
+				retry_after_ms: u.weighNumRange(1000, 60000, 0.3, 5000),
+				requests_per_minute: u.weighNumRange(50, 2000, 0.4, 500),
+				tier_limit: ["Free", "Build", "Enterprise"],
+			},
+		},
+		{
+			event: "billing payment",
+			weight: 2,
+			properties: {
+				amount_usd: u.weighNumRange(5, 50000, 0.2, 500),
+				payment_method: ["credit_card", "credit_card", "credit_card", "invoice", "wire_transfer"],
+				billing_period: ["monthly", "monthly", "annual"],
+				tokens_consumed: u.weighNumRange(100000, 50000000, 0.3, 5000000),
+			},
+		},
+		{
+			event: "model selected",
+			weight: 3,
+			properties: {
+				model: ["sonnet-4", "sonnet-4", "haiku-4", "opus-4-6"],
+				is_default: [true, true, false],
+				selection_context: ["playground", "api_config", "eval_setup", "batch_config"],
+			},
+		},
+		{
+			event: "dashboard viewed",
+			weight: 5,
+			properties: {
+				dashboard_section: ["usage", "billing", "api_keys", "models", "evals", "logs"],
+				time_range: ["1h", "24h", "7d", "30d"],
+			},
+		},
+		{
+			event: "docs searched",
+			weight: 4,
+			properties: {
+				search_query_category: ["api_reference", "quickstart", "pricing", "models", "tool_use", "batch_api", "caching", "errors"],
+				results_found: u.weighNumRange(0, 50, 0.5, 8),
+				clicked_result: [true, true, true, false],
+			},
+		},
+		{
+			event: "member invited",
+			weight: 2,
+			properties: {
+				invite_role: ["admin", "developer", "developer", "billing", "viewer"],
+				invite_method: ["email", "email", "sso", "link"],
+			},
+		},
+		{
+			event: "webhook configured",
+			weight: 1,
+			properties: {
+				webhook_event: ["usage_alert", "rate_limit", "batch_complete", "eval_complete", "billing_threshold"],
+				delivery_method: ["https", "https", "slack", "email"],
+			},
+		},
+		{
+			event: "playground session",
+			weight: 4,
+			properties: {
+				model: ["sonnet-4", "sonnet-4", "haiku-4", "opus-4-6"],
+				turns: u.weighNumRange(1, 30, 0.4, 5),
+				shared: [false, false, false, true],
+				tokens_used: u.weighNumRange(100, 20000, 0.3, 3000),
+			},
+		},
+		{
+			event: "account deactivated",
+			weight: 1,
+			isChurnEvent: true,
+			returnLikelihood: 0.1,
+			isStrictEvent: true,
+			properties: {
+				reason: ["cost", "switched_provider", "project_ended", "rate_limits", "no_longer_needed", "performance"],
+			},
+		},
+	],
+
+	// -- Funnels (3) ------------------------------------------
+	funnels: [
+		{
+			name: "Onboarding",
+			sequence: ["organization created", "api key created", "api call"],
+			conversionRate: 70,
+			order: "sequential",
+			isFirstFunnel: true,
+			timeToConvert: 48,
+			weight: 3,
+		},
+		{
+			name: "API to Eval Pipeline",
+			sequence: ["api call", "tool use call", "eval job"],
+			conversionRate: 45,
+			order: "sequential",
+			timeToConvert: 168,
+			weight: 5,
+		},
+		{
+			name: "Usage to Billing",
+			sequence: ["api call", "billing payment"],
+			conversionRate: 30,
+			order: "sequential",
+			timeToConvert: 336,
+			weight: 2,
+		},
+	],
+
+	// -- SuperProps --------------------------------------------
+	superProps: {
+		api_tier: ["Free", "Free", "Build", "Build", "Enterprise"],
+		primary_use_case: ["chatbot", "code_generation", "data_extraction", "content_creation", "agents"],
+		sdk_language: ["python", "typescript", "java", "go", "curl"],
+	},
+
+	// -- UserProps ---------------------------------------------
+	userProps: {
+		api_tier: ["Free", "Free", "Build", "Build", "Enterprise"],
+		primary_use_case: ["chatbot", "code_generation", "data_extraction", "content_creation", "agents"],
+		sdk_language: ["python", "typescript", "java", "go", "curl"],
+		monthly_spend: u.weighNumRange(0, 50000, 0.2, 200),
+		total_api_calls: u.weighNumRange(0, 500000, 0.2, 10000),
+		preferred_model: ["sonnet-4", "sonnet-4", "haiku-4", "opus-4-6"],
+		has_eval_pipeline: [false],
+		hit_rate_limit_early: [false],
+		has_early_eval: [false],
+	},
+
+	// -- Hook Function ----------------------------------------
+	hook: function (record, type, meta) {
+		// ─────────────────────────────────────────────────────────
+		// Hook #6: OUTAGE DAY (event)
+		// Days 40-41: 40% of api calls get is_error=true with
+		// service error types
+		// ─────────────────────────────────────────────────────────
+		if (type === "event") {
+			if (record.event === "api call") {
+				const eventTime = dayjs(record.time);
+				const dayInDataset = eventTime.diff(DATASET_START, "days", true);
+
+				// Hook #6: Outage day errors
+				if (dayInDataset >= 40 && dayInDataset < 42) {
+					if (chance.bool({ likelihood: 40 })) {
+						record.is_error = true;
+						record.error_type = chance.pickone([
+							"service_overloaded",
+							"internal_server_error",
+							"gateway_timeout",
+						]);
+						record.latency_ms = Math.floor((record.latency_ms || 1500) * 3);
+					}
+				}
+
+				// Hook #2: Model migration wave (event portion)
+				// After day 60, 35% of Build/Enterprise users switch to opus-4-7
+				if (dayInDataset >= 60) {
+					if (
+						(record.api_tier === "Build" || record.api_tier === "Enterprise") &&
+						chance.bool({ likelihood: 35 })
+					) {
+						record.model = "opus-4-7";
+					}
+				}
+			}
+
+			return record;
+		}
+
+		// ─────────────────────────────────────────────────────────
+		// Hook: USER PROFILE ENRICHMENT (user)
+		// Tag user profiles for discoverability
+		// ─────────────────────────────────────────────────────────
+		if (type === "user") {
+			// Defaults for hook-driven user properties
+			record.hit_rate_limit_early = false;
+			record.has_early_eval = false;
+			record.has_eval_pipeline = false;
+		}
+
+		// ─────────────────────────────────────────────────────────
+		// EVERYTHING HOOKS
+		// ─────────────────────────────────────────────────────────
+		if (type === "everything") {
+			let events = record;
+			if (!events.length) return record;
+			const profile = meta && meta.profile ? meta.profile : {};
+
+			// Stamp superProps from profile for consistency
+			events.forEach(e => {
+				if (profile.api_tier) e.api_tier = profile.api_tier;
+				if (profile.primary_use_case) e.primary_use_case = profile.primary_use_case;
+				if (profile.sdk_language) e.sdk_language = profile.sdk_language;
+			});
+
+			// Determine first event time for relative day calculations
+			const sortedByTime = [...events].sort((a, b) => dayjs(a.time).valueOf() - dayjs(b.time).valueOf());
+			const firstEventTime = sortedByTime.length > 0 ? dayjs(sortedByTime[0].time) : DATASET_START;
+
+			// ─────────────────────────────────────────────────────
+			// Hook #5: TIER-BASED CONTEXT WINDOW (SUBSCRIPTION TIER)
+			// Scale context_window and input_tokens by tier
+			// ─────────────────────────────────────────────────────
+			const tier = profile.api_tier || "Free";
+			const contextWindow = tier === "Enterprise" ? 2000000 : tier === "Build" ? 1000000 : 200000;
+			const inputMultiplier = tier === "Enterprise" ? 4 : tier === "Build" ? 2 : 1;
+
+			events.forEach(e => {
+				if (e.event === "api call") {
+					e.context_window = contextWindow;
+					e.input_tokens = Math.floor((e.input_tokens || 2000) * inputMultiplier);
+				}
+			});
+
+			// ─────────────────────────────────────────────────────
+			// Hook #1: PROMPT CACHING ADOPTION (CONVERSION)
+			// Users with any cache_enabled=true get 70% cost reduction
+			// on all subsequent api calls
+			// ─────────────────────────────────────────────────────
+			// ~25% of users have caching enabled on at least one event
+			const userId = events[0] && events[0].user_id;
+			const idHash = String(userId || "").split("").reduce((acc, c) => acc + c.charCodeAt(0), 0);
+			const isCacheUser = (idHash % 4) === 0;
+
+			if (isCacheUser) {
+				let cacheActivated = false;
+				// Activate caching on events after the first 20% of user events
+				const activationPoint = Math.floor(events.length * 0.2);
+				events.forEach((e, idx) => {
+					if (e.event === "api call") {
+						if (idx >= activationPoint) {
+							cacheActivated = true;
+						}
+						if (cacheActivated) {
+							e.cache_enabled = true;
+							e.cost_usd = Math.round((e.cost_usd || 0.01) * 0.3 * 10000) / 10000;
+						}
+					}
+				});
+			}
+
+			// ─────────────────────────────────────────────────────
+			// Hook #2: MODEL MIGRATION WAVE (everything portion)
+			// opus-4-7 users get 1.5x tokens_used
+			// (model assignment done in event hook above)
+			// ─────────────────────────────────────────────────────
+			events.forEach(e => {
+				if (e.event === "api call" && e.model === "opus-4-7") {
+					e.tokens_used = Math.floor((e.tokens_used || 2500) * 1.5);
+				}
+			});
+
+			// ─────────────────────────────────────────────────────
+			// Hook #3: AGENTIC LOOP POWER USERS (BEHAVIORS TOGETHER)
+			// Users with tool use + multi_turn get 8x tokens, 3x events
+			// ─────────────────────────────────────────────────────
+			const hasToolUse = events.some(e => e.event === "tool use call");
+			const hasMultiTurn = events.some(e => e.event === "api call" && e.multi_turn === true);
+			const isAgenticUser = hasToolUse && hasMultiTurn;
+
+			if (isAgenticUser) {
+				// Mark all api calls as agentic and boost tokens
+				events.forEach(e => {
+					if (e.event === "api call") {
+						e.is_agentic_user = true;
+						e.tokens_used = Math.floor((e.tokens_used || 2500) * 8);
+					}
+				});
+
+				// Inject 3x extra api call events by cloning existing ones
+				const apiCalls = events.filter(e => e.event === "api call");
+				const extraCount = apiCalls.length * 2; // 2 extra per existing = 3x total
+				for (let i = 0; i < extraCount; i++) {
+					const template = apiCalls[i % apiCalls.length];
+					if (template) {
+						events.push({
+							...template,
+							time: dayjs(template.time).add(chance.integer({ min: 1, max: 120 }), "minutes").toISOString(),
+							user_id: template.user_id,
+							is_agentic_user: true,
+							multi_turn: true,
+						});
+					}
+				}
+			}
+
+			// ─────────────────────────────────────────────────────
+			// Hook #4: RATE LIMIT CHURN
+			// >=5 rate limit errors in first 7 days -> remove 60% of
+			// events after week 1
+			// ─────────────────────────────────────────────────────
+			const firstWeekEnd = firstEventTime.add(7, "days");
+			const earlyRateLimits = events.filter(e =>
+				e.event === "rate limit error" &&
+				dayjs(e.time).isBefore(firstWeekEnd)
+			).length;
+
+			if (earlyRateLimits >= 5) {
+				// Tag the user profile
+				if (profile) profile.hit_rate_limit_early = true;
+
+				// Remove 60% of events after week 1
+				events = events.filter(e => {
+					if (dayjs(e.time).isAfter(firstWeekEnd)) {
+						return chance.bool({ likelihood: 40 }); // keep 40% = remove 60%
+					}
+					return true;
+				});
+			}
+
+			// ─────────────────────────────────────────────────────
+			// Hook #7: BATCH API DISCOUNT (PURCHASE VALUE)
+			// Batch users get 50% lower cost_per_token, 2x tokens_used
+			// ─────────────────────────────────────────────────────
+			const isBatchUser = events.some(e => e.event === "batch job submitted");
+
+			if (isBatchUser) {
+				events.forEach(e => {
+					if (e.event === "api call") {
+						e.is_batch_user = true;
+						e.cost_per_token = Math.round((e.cost_per_token || 0.00001) * 0.5 * 10000000) / 10000000;
+						e.tokens_used = Math.floor((e.tokens_used || 2500) * 2);
+					}
+				});
+			}
+
+			// ─────────────────────────────────────────────────────
+			// Hook #8: EVAL-DRIVEN RETENTION
+			// Early eval users (first 7 days) get 75% D30 retention
+			// Non-eval users get only 25% D30 retention (remove events)
+			// ─────────────────────────────────────────────────────
+			const hasEarlyEval = events.some(e =>
+				e.event === "eval job" &&
+				dayjs(e.time).isBefore(firstWeekEnd)
+			);
+
+			if (hasEarlyEval) {
+				// Tag user profile
+				if (profile) {
+					profile.has_early_eval = true;
+					profile.has_eval_pipeline = true;
+				}
+				// Early eval users keep all their events (high retention)
+			} else {
+				// Non-eval users: remove 75% of events after day 30
+				const day30 = firstEventTime.add(30, "days");
+				events = events.filter(e => {
+					if (dayjs(e.time).isAfter(day30)) {
+						return chance.bool({ likelihood: 25 }); // keep 25%
+					}
+					return true;
+				});
+			}
+
+			return events;
+		}
+
+		return record;
+	},
+};
+
+export default config;