@ak--47/dungeon-master 1.5.4 → 1.6.0

package/dungeons/vertical/community/community.sql

@@ -0,0 +1,216 @@
+-- ============================================================
+-- community.js — v1.6 human-inspection queries (DuckDB)
+--
+-- Every query is keyed to a story id in community.js's `stories` export;
+-- the machine-checked verdicts come from:
+--   node scripts/verify-stories.mjs dungeons/vertical/community/community.js --data-prefix verify-community
+-- Generate first:
+--   node scripts/verify-runner.mjs dungeons/vertical/community/community.js verify-community
+-- Run this file:
+--   duckdb -c ".read dungeons/vertical/community/community.sql"
+-- ============================================================
+
+-- ── identity-resolution prelude ─────────────────────────────
+-- avgDevicePerUser: 2 + account created is both isAuthEvent and
+-- isFirstEvent, so born users auth on their first event; the device-pool
+-- resolve is belt-and-braces for any device-only edge.
+CREATE OR REPLACE VIEW users AS
+SELECT * FROM read_json_auto('data/verify-community-USERS*.json', sample_size=-1, union_by_name=true);
+
+CREATE OR REPLACE VIEW device_map AS
+-- profiles store the device pool under the legacy "anonymousIds" key
+SELECT unnest("anonymousIds") AS device_id, distinct_id FROM users;
+
+CREATE OR REPLACE VIEW ev AS
+-- ::VARCHAR casts — user_id sniffs as UUID, device_id as VARCHAR; DuckDB
+-- refuses to coalesce mixed types
+SELECT coalesce(m.distinct_id::VARCHAR, e.user_id::VARCHAR, e.device_id::VARCHAR) AS uid,
+       e.time::TIMESTAMP AS t,
+       e.*
+FROM read_json_auto('data/verify-community-EVENTS*.json', sample_size=-1, union_by_name=true) e
+LEFT JOIN device_map m ON e.device_id = m.device_id;
+
+-- Per-user counts. Deletions-only hooks (H6 lurker churn, H8 free-tier
+-- comment drop, H10 over-publisher upvote drop, future-time guard) make
+-- hook-time cohort recovery ONE-SIDED: output count >= threshold implies
+-- hook-time count >= threshold, but not the reverse. Stories handle this
+-- with output-implies-hook cohort choices; read these tables the same way.
+CREATE OR REPLACE VIEW per_user AS
+SELECT uid,
+  count(*) FILTER (WHERE event = 'article published') AS pubs,
+  count(*) FILTER (WHERE event = 'article edited') AS edits,
+  count(*) FILTER (WHERE event = 'comment posted') AS comments,
+  count(*) FILTER (WHERE event = 'upvote given') AS upvotes,
+  count(*) FILTER (WHERE event = 'app session') AS sessions,
+  count(*) FILTER (WHERE event = 'discussion posted') AS discussions,
+  count(*) AS n
+FROM ev GROUP BY 1;
+
+CREATE OR REPLACE VIEW first_ev AS
+SELECT uid, min(t) AS f FROM ev GROUP BY 1;
+
+
+-- ── H1-weekend-word-count ───────────────────────────────────
+-- Sat/Sun 'article published' word_count ×1.5 (floored). 'wiki page
+-- created' carries the same prop but is untouched — the placebo arm.
+-- dayofweek(): Sunday = 0 (matches JS getUTCDay).
+SELECT event,
+  CASE WHEN dayofweek(t) IN (0, 6) THEN 'weekend' ELSE 'weekday' END AS bucket,
+  count(*) AS n, round(avg(word_count), 1) AS avg_wc
+FROM ev WHERE event IN ('article published', 'wiki page created')
+GROUP BY 1, 2 ORDER BY 1, 2;
+-- read: published weekend/weekday ≈ 1.5; wiki ≈ 1.0
+
+
+-- ── H2-trending-gaming-window ───────────────────────────────
+-- days 35-50 EXCLUSIVE (hook uses isAfter/isBefore): gaming-hub users'
+-- 'article viewed' view_count ×2. Event content_hub is stamped from the
+-- profile BEFORE H2 runs, so it selects exactly the treated users.
+SELECT CASE WHEN content_hub = 'gaming' THEN 'gaming' ELSE 'other' END AS hub,
+  CASE WHEN t > TIMESTAMP '2026-02-05' AND t < TIMESTAMP '2026-02-20' THEN 'in_window' ELSE 'outside' END AS bucket,
+  count(*) AS n, round(avg(view_count), 1) AS avg_vc
+FROM ev WHERE event = 'article viewed'
+GROUP BY 1, 2 ORDER BY 1, 2;
+-- read: gaming in/out ≈ 2.0; other in/out ≈ 1.0
+
+
+-- ── H3-power-creator-upvotes ────────────────────────────────
+-- >20 publishes at hook time → upvote_count ×3 (floor(3w)=3w exact on
+-- integers). Publishes never deleted → output ≥21 IMPLIES treatment.
+SELECT CASE WHEN p.pubs > 20 THEN 'power' WHEN p.pubs <= 1 THEN 'low' END AS arm,
+  count(DISTINCT p.uid) AS users, round(avg(e.upvote_count), 2) AS avg_uc,
+  round(count(*) FILTER (WHERE e.upvote_count % 3 = 0)::DOUBLE / count(*), 4) AS mod3_share
+FROM per_user p JOIN ev e ON e.uid = p.uid AND e.event = 'upvote given'
+WHERE p.pubs > 20 OR p.pubs <= 1
+GROUP BY 1 ORDER BY 1;
+-- read: power/low avg_uc ≈ 3.0; power mod3_share ≈ 1.0, low ≈ 0.2-0.35
+
+
+-- ── H4-discussion-depth ─────────────────────────────────────
+-- active_contributor (role 'contributor') clones 50% of surviving comments
+-- as is_reply=true. Raw counts are persona-confounded (1.5× vs 0.3×
+-- multipliers) AND conversion-gated (comments are funnel steps;
+-- contributor conversionModifier 1.0 vs reader 0.5), so the raw
+-- comments-per-session DD over-estimates (~2.1) while the
+-- discussion-calibrated DD under-estimates (discussions sit deeper in
+-- their funnel — EL step 4 vs comment step 3 — so the calibrator
+-- over-corrects). The two bracket the 1.5 knob with sign-known biases.
+SELECT u.role, count(DISTINCT u.distinct_id) AS users,
+  round(sum(p.comments)::DOUBLE / nullif(sum(p.sessions), 0), 4) AS comments_per_session,
+  round(sum(p.discussions)::DOUBLE / nullif(sum(p.sessions), 0), 4) AS discussions_per_session
+FROM users u JOIN per_user p ON p.uid = u.distinct_id::VARCHAR
+WHERE u.role IN ('contributor', 'reader')
+GROUP BY 1 ORDER BY 1;
+-- read: corrected DD = (comments DD ÷ discussions DD) ≤ 1.5 ≤ raw comments DD
+
+-- reply-share composition: clones are always replies → contributor share
+-- shifts from the organic 2/3 to (2/3 + 0.5)/1.5 ≈ 0.778 (drop-invariant)
+SELECT u.role,
+  round(count(*) FILTER (WHERE e.is_reply = true)::DOUBLE / count(*), 4) AS reply_share
+FROM users u JOIN ev e ON e.uid = u.distinct_id::VARCHAR AND e.event = 'comment posted'
+WHERE u.role IN ('contributor', 'reader')
+GROUP BY 1 ORDER BY 1;
+-- read: contributor ≈ 0.778, reader ≈ 0.667
+
+
+-- ── H5-edit-war ─────────────────────────────────────────────
+-- >5 edits at hook time → ALL edit_quality redrawn U[1.0, 2.0]. Edits
+-- never deleted → output ≥6 IMPLIES treatment → redraw is EXACT on the
+-- war arm: avg 1.5, zero values above 2.0.
+SELECT CASE WHEN p.edits > 5 THEN 'war' WHEN p.edits BETWEEN 1 AND 4 THEN 'calm' END AS arm,
+  count(DISTINCT p.uid) AS users, round(avg(e.edit_quality), 3) AS avg_q,
+  count(*) FILTER (WHERE e.edit_quality > 2.0) AS over_cap
+FROM per_user p JOIN ev e ON e.uid = p.uid AND e.event = 'article edited'
+WHERE p.edits > 5 OR p.edits BETWEEN 1 AND 4
+GROUP BY 1 ORDER BY 1;
+-- read: war avg_q ≈ 1.5 with over_cap = 0; calm avg_q ≈ 3 (organic 1-5)
+
+
+-- ── H6-lurker-churn ─────────────────────────────────────────
+-- <5 events at hook time → 60% of post-day-10 events dropped (keep 0.4).
+-- Output n∈[2,4] implies treated; output n∈[5,8] implies untreated.
+-- n=1 users carry no churn information (post=0 and days-5-10=0 by
+-- construction) and are excluded; the control band is ADJACENT ([5,8],
+-- not [6,10]) because the front-loading calibration transfers better
+-- between closer activity levels. Arms restricted to role 'reader'.
+-- Raw post/pre ratio is confounded by organic front-loading, so calibrate
+-- on the pre-cutoff half-split (days 0-5 vs 5-10 — never touched by H6).
+SELECT CASE WHEN p.n BETWEEN 2 AND 4 THEN 'tiny' WHEN p.n BETWEEN 5 AND 8 THEN 'small' END AS arm,
+  count(DISTINCT p.uid) AS users,
+  round(sum(CASE WHEN e.t > fe.f + INTERVAL 10 DAY THEN 1 ELSE 0 END)::DOUBLE
+    / nullif(sum(CASE WHEN e.t <= fe.f + INTERVAL 10 DAY THEN 1 ELSE 0 END), 0), 4) AS rho,
+  round(sum(CASE WHEN e.t > fe.f + INTERVAL 5 DAY AND e.t <= fe.f + INTERVAL 10 DAY THEN 1 ELSE 0 END)::DOUBLE
+    / nullif(sum(CASE WHEN e.t <= fe.f + INTERVAL 5 DAY THEN 1 ELSE 0 END), 0), 4) AS rho_pre
+FROM per_user p
+JOIN first_ev fe ON fe.uid = p.uid
+JOIN ev e ON e.uid = p.uid
+JOIN users u ON u.distinct_id::VARCHAR = p.uid AND u.role = 'reader'
+WHERE (p.n BETWEEN 2 AND 4 OR p.n BETWEEN 5 AND 8)
+  AND fe.f < TIMESTAMP '2026-04-15'   -- exclude births without a post window
+GROUP BY 1 ORDER BY 1;
+-- read: (rho_tiny/rho_small) ÷ (rho_pre_tiny/rho_pre_small) ≈ 0.4 keep rate
+
+
+-- ── H7-creator-profiles ─────────────────────────────────────
+-- user hook overwrites per role: creator art U[50,200] rep U[80,100];
+-- moderator U[10,50]/U[40,70]; contributor U[1,15]/U[15,50]; reader 0/U[0,20].
+SELECT role, count(*) AS users,
+  min(reputation_score) AS min_rep, round(avg(reputation_score), 2) AS avg_rep, max(reputation_score) AS max_rep,
+  min(articles_created) AS min_art, max(articles_created) AS max_art
+FROM users GROUP BY 1 ORDER BY 1;
+-- read: ranges exact per role; creator avg_rep ≈ 90
+
+
+-- ── H8-pro-content-lift ─────────────────────────────────────
+-- free/non-supporter tiers drop 65% of ALL 'comment posted' (keep 0.35).
+-- Tier ⊥ persona → H4's clone factor cancels in the pooled ratio.
+SELECT CASE WHEN e.subscription_tier IN ('pro', 'supporter') THEN 'paid' ELSE 'free' END AS arm,
+  count(DISTINCT e.uid) AS users,
+  round(count(*) FILTER (WHERE e.event = 'comment posted')::DOUBLE
+    / nullif(count(*) FILTER (WHERE e.event = 'app session'), 0), 4) AS comments_per_session
+FROM ev e GROUP BY 1 ORDER BY 1;
+-- read: free/paid comments-per-session ≈ 0.35 (the exact keep rate)
+
+
+-- ── H9-content-ttc ──────────────────────────────────────────
+-- funnel-post scales Content Creation gaps: pro/supporter ×0.77, free
+-- ×1.25 (v1.6 scopes the hook to Content Creation only).
+-- CAUTION: cross-event TTC SQL here is the documented greedy-single-pass
+-- limitation — it pairs views/publishes across funnel instances and buries
+-- the signal. The story asserts TTC through the Mixpanel-aligned emulator
+-- (timeToConvert, 60h window = 48h generative × 1.25 max stretch); trust
+-- the story verdict, not ad-hoc pair SQL.
+SELECT u.subscription_tier AS tier, count(*) AS published_events
+FROM ev e JOIN users u ON u.distinct_id::VARCHAR = e.uid
+WHERE e.event = 'article published' GROUP BY 1 ORDER BY 1;
+
+
+-- ── H10-article-magic-number ────────────────────────────────
+-- sweet spot 2-5 publishes → upvote_count ×1.35 (rounded); 6+ publishes →
+-- creator burnout from day 60 (2026-03-02): 40% of upvote events dropped
+-- after the cutoff. Value read is persona-clean (iid draw).
+SELECT CASE WHEN p.pubs BETWEEN 2 AND 5 THEN 'sweet' WHEN p.pubs <= 1 THEN 'low'
+            WHEN p.pubs >= 6 THEN 'over' END AS arm,
+  count(DISTINCT p.uid) AS users, round(avg(e.upvote_count), 3) AS avg_uc
+FROM per_user p JOIN ev e ON e.uid = p.uid AND e.event = 'upvote given'
+GROUP BY 1 ORDER BY 1;
+-- read: sweet/low avg_uc ≈ 1.35-1.4 (integer rounding drifts it up)
+
+-- volume read: publish count is intrinsically coupled to activity, so NO
+-- cross-arm level comparison works (organic upvote share differs 23-58%
+-- across publish bands whatever the denominator). The calendar edge makes
+-- it a difference-in-differences: each arm's own after/before
+-- upvotes-per-session ratio cancels its activity composition (measured
+-- arm-invariant to ~0.1% on untreated data). Contributors only.
+SELECT CASE WHEN p.pubs BETWEEN 2 AND 5 THEN 'sweet' WHEN p.pubs >= 6 THEN 'over' END AS arm,
+  CASE WHEN e.t >= TIMESTAMP '2026-03-02' THEN 'after' ELSE 'before' END AS period,
+  count(DISTINCT p.uid) AS users,
+  count(*) FILTER (WHERE e.event = 'upvote given') AS ups,
+  count(*) FILTER (WHERE e.event = 'app session') AS sessions
+FROM per_user p
+JOIN users u ON u.distinct_id::VARCHAR = p.uid AND u.role = 'contributor'
+JOIN ev e ON e.uid = p.uid
+WHERE p.pubs >= 2
+GROUP BY 1, 2 ORDER BY 1, 2;
+-- read: DiD = (over after/before ups-per-session) ÷ (sweet after/before)
+--       ≈ 0.6 (the keep rate)

package/dungeons/vertical/community/community.verify.mjs

@@ -0,0 +1,72 @@
+#!/usr/bin/env node
+/**
+ * community.verify.mjs — thin wrapper around the story runner.
+ *
+ * All verification logic lives in the `stories` export of ./community.js;
+ * this script just streams the shards and delegates. It is equivalent to:
+ *
+ *   node scripts/verify-stories.mjs dungeons/vertical/community/community.js --data-prefix verify-community
+ *
+ * Generate first:
+ *   node scripts/verify-runner.mjs dungeons/vertical/community/community.js verify-community
+ * Run:
+ *   node dungeons/vertical/community/community.verify.mjs [prefix]   # default verify-community
+ */
+import fs from 'node:fs';
+import path from 'node:path';
+import readline from 'node:readline';
+import { execFile } from 'node:child_process';
+import { promisify } from 'node:util';
+import { buildIdentityMap, evaluateStories, VERDICT_RANK } from '@ak--47/dungeon-master/verify';
+import config, { stories } from './community.js';
+
+const PREFIX = process.argv[2] || 'verify-community';
+const prefixPath = path.join('data', PREFIX);
+
+async function loadShards(suffix) {
+	// streaming load: full-fidelity EVENTS shards exceed the readFileSync cap
+	const dir = path.dirname(prefixPath), base = path.basename(prefixPath);
+	const out = [];
+	if (!fs.existsSync(dir)) return out;
+	for (const f of fs.readdirSync(dir).filter(f => f.startsWith(`${base}-${suffix}`) && f.endsWith('.json')).sort()) {
+		const rl = readline.createInterface({ input: fs.createReadStream(path.join(dir, f)), crlfDelay: Infinity });
+		for await (const line of rl) {
+			if (line.trim()) out.push(JSON.parse(line));
+		}
+	}
+	return out;
+}
+
+const events = await loadShards('EVENTS');
+const profiles = await loadShards('USERS');
+if (!events.length) {
+	console.error(`no shards at ${prefixPath}-EVENTS*.json — run: node scripts/verify-runner.mjs dungeons/vertical/community/community.js ${PREFIX}`);
+	process.exit(1);
+}
+console.log(`community — events=${events.length} users=${profiles.length} (${prefixPath})`);
+
+const execFileP = promisify(execFile);
+const runSql = async (sql) => {
+	const { stdout } = await execFileP('duckdb', ['-json', '-c', sql.replaceAll('{{PREFIX}}', prefixPath)], { maxBuffer: 64 * 1024 * 1024 });
+	return stdout.trim() ? JSON.parse(stdout) : [];
+};
+
+// funnels passed raw (unvalidated) — the H8/H9 emulator stories carry their
+// own explicit conversion window (48h Content Creation generative window ×
+// the 1.25 free-tier stretch = 60h, covering the stretched support), so no
+// funnel-default threading is needed
+const results = await evaluateStories(stories, events, {
+	profiles,
+	funnels: config.funnels,
+	identityMap: buildIdentityMap(profiles),
+	runSql,
+});
+
+let worst = 'NAILED';
+for (const r of results) {
+	console.log(`${r.verdict.padEnd(7)} ${r.id}`);
+	for (const a of r.assertions) console.log(`        ${a.verdict.padEnd(7)} ${a.detail}`);
+	if (VERDICT_RANK[r.verdict] < VERDICT_RANK[worst]) worst = r.verdict;
+}
+console.log(`\nworst verdict: ${worst}`);
+process.exit(VERDICT_RANK[worst] >= VERDICT_RANK.STRONG ? 0 : 1);