diff --git a/app/server/src/confluence/__tests__/confluenceRack.test.ts b/app/server/src/confluence/__tests__/confluenceRack.test.ts
new file mode 100644
index 0000000..5b9bf36
--- /dev/null
+++ b/app/server/src/confluence/__tests__/confluenceRack.test.ts
@@ -0,0 +1,169 @@
+// Investor Flow — confluenceRack.test.ts
+// Pure-logic tests for rack evaluation, redundancy discounting, and
+// picture-change detection.
+
+import { describe, it } from 'node:test';
+import assert from 'node:assert/strict';
+
+import {
+  evaluateRack,
+  labelQuality,
+  detectPictureChange,
+  flippedSlots,
+  slotBody,
+} from '../confluenceRack.ts';
+import type { SlotAssessment } from '../confluenceRack.ts';
+import { confluenceslotById, CONFLUENCE_SLOTS } from '../confluenceSlots.ts';
+
+const fired = (id: string): SlotAssessment => ({ id, state: 'fired' });
+const notFired = (id: string): SlotAssessment => ({ id, state: 'not-fired' });
+
+describe('confluence catalog', () => {
+  it('defines exactly 34 slots across all six families', () => {
+    assert.equal(CONFLUENCE_SLOTS.length, 34);
+    const families = new Set(CONFLUENCE_SLOTS.map((s) => s.family));
+    assert.deepEqual([...families].sort(), ['flows', 'institutional', 'macro', 'seasonal', 'sentiment', 'technical']);
+  });
+
+  it('every defined slot has a primary-rule-safe explain sentence', () => {
+    const directivePattern = /\b(should|recommend|buy now|sell now|go long|go short|recommendation|buy the|sell the|long position|short position)\b/i;
+    for (const s of CONFLUENCE_SLOTS) {
+      assert.ok(s.explain.length > 10, `${s.id} explain note too short`);
+      assert.ok(!directivePattern.test(s.explain), `${s.id} explain note reads like a directive: ${s.explain}`);
+    }
+  });
+});
+
+describe('slotBody', () => {
+  it('exit slots map to bear evidence, bull/bear stay put', () => {
+    assert.equal(slotBody(confluenceslotById.get('goldenCross')!), 'bull');
+    assert.equal(slotBody(confluenceslotById.get('deathCross')!), 'bear');
+    assert.equal(slotBody(confluenceslotById.get('consumerSentimentHigh')!), 'bear');
+  });
+});
+
+describe('labelQuality', () => {
+  it('returns sparse below the minimum evidence threshold', () => {
+    assert.equal(labelQuality(0.5, 0.2, 0.7), 'sparse');
+  });
+
+  it('labels strong-bullish from a one-sided heavy bull weighting', () => {
+    assert.equal(labelQuality(7, 0.5, 7.5), 'strong-bullish');
+  });
+
+  it('labels strong-bearish from heavy bear + exit evidence', () => {
+    assert.equal(labelQuality(0.5, 6, 6.5), 'strong-bearish');
+  });
+
+  it('labels moderate-bullish between strong and weak', () => {
+    assert.equal(labelQuality(2.5, 0.5, 3), 'moderate-bullish');
+  });
+
+  it('labels weak-bullish when barely directional', () => {
+    assert.equal(labelQuality(1.7, 0.7, 2.4), 'weak-bullish');
+  });
+
+  it('labels mixed when neither side holds 60% of evidence', () => {
+    assert.equal(labelQuality(1.2, 1.1, 2.3), 'mixed');
+  });
+});
+
+describe('evaluateRack', () => {
+  it('applies redundancy discount within a redundancy group', () => {
+    // goldenCross + trendAlignment + pullbackToEMA21 all fire (all same group).
+    // groupWeight = 1 + 0.5 + 0.25 = 1.75, not 3.0.
+    const ev = evaluateRack('PLTR', '2026-01-05', [
+      fired('goldenCross'),
+      fired('trendAlignment'),
+      fired('pullbackToEMA21'),
+    ]);
+    assert.equal(ev.bullCount, 3);
+    assert.ok(Math.abs(ev.bullEvidence - 1.75) < 1e-9, `bullEvidence=${ev.bullEvidence}`);
+  });
+
+  it('counts distinct families as independent evidence', () => {
+    // One technical + one institutional + one seasonal + one flows: 4 independent facts.
+    const ev = evaluateRack('PLTR', '2026-01-05', [
+      fired('goldenCross'),
+      fired('instNetActivePositive'),
+      fired('seasonalFavorableMonth'),
+      fired('etfFlowPositive'),
+    ]);
+    assert.equal(ev.bullEvidence, 4);
+    assert.equal(ev.bullCount, 4);
+  });
+
+  it('counts exit slots as bear evidence', () => {
+    const ev = evaluateRack('PLTR', '2026-01-05', [
+      fired('deathCross'),
+      fired('rsiOverbought'),
+      fired('insiderInformedSell'),
+    ]);
+    assert.equal(ev.bearCount, 3);
+    assert.equal(ev.bullCount, 0);
+    assert.equal(ev.quality, 'moderate-bearish');
+  });
+
+  it('ignores unknown slot ids but still counts assessed coverage', () => {
+    const ev = evaluateRack('PLTR', '2026-01-05', [
+      fired('goldenCross'),
+      fired('notARealSlot'),
+      notFired('relVolume'),
+    ]);
+    assert.equal(ev.assessedCount, 2);
+    assert.equal(ev.bullCount, 1);
+  });
+
+  it('mixed picture when bull and bear evidence are both present and balanced', () => {
+    const ev = evaluateRack('NVDA', '2026-01-05', [
+      fired('goldenCross'),
+      fired('instNetActivePositive'),
+      fired('deathCross'),
+      fired('insiderInformedSell'),
+    ]);
+    assert.equal(ev.quality, 'mixed');
+  });
+});
+
+describe('detectPictureChange', () => {
+  it('reports improved when quality rank rises', () => {
+    const prev = evaluateRack('PLTR', '2026-01-05', [fired('deathCross')]);
+    const curr = evaluateRack('PLTR', '2026-01-07', [fired('goldenCross')]);
+    const change = detectPictureChange(prev, curr);
+    assert.equal(change.changed, true);
+    assert.equal(change.changeType, 'improved');
+  });
+
+  it('reports deteriorated when quality rank falls', () => {
+    const prev = evaluateRack('PLTR', '2026-01-05', [fired('goldenCross'), fired('instNetActivePositive')]);
+    const curr = evaluateRack('PLTR', '2026-01-07', [fired('deathCross'), fired('insiderInformedSell')]);
+    const change = detectPictureChange(prev, curr);
+    assert.equal(change.changed, true);
+    assert.equal(change.changeType, 'deteriorated');
+  });
+
+  it('reports unchanged for identical evaluations', () => {
+    const ev = evaluateRack('PLTR', '2026-01-05', [fired('goldenCross')]);
+    const change = detectPictureChange(ev, ev);
+    assert.equal(change.changed, false);
+    assert.equal(change.changeType, 'unchanged');
+    assert.deepEqual(change.flippedSlotIds, []);
+  });
+
+  it('flag lists flipped slot ids', () => {
+    const prev = evaluateRack('PLTR', '2026-01-05', [fired('goldenCross'), notFired('relVolume')]);
+    const curr = evaluateRack('PLTR', '2026-01-07', [notFired('goldenCross'), fired('relVolume')]);
+    const change = detectPictureChange(prev, curr);
+    assert.deepEqual(change.flippedSlotIds.sort(), ['goldenCross', 'relVolume']);
+  });
+});
+
+describe('flippedSlots', () => {
+  it('finds flips including newly-added and removed slots', () => {
+    const flips = flippedSlots(
+      [fired('goldenCross'), notFired('relVolume')],
+      [notFired('goldenCross'), fired('relVolume'), fired('cotPositioning')],
+    );
+    assert.deepEqual(flips.sort(), ['cotPositioning', 'goldenCross', 'relVolume']);
+  });
+});
\ No newline at end of file
diff --git a/app/server/src/confluence/confluenceLibrary.ts b/app/server/src/confluence/confluenceLibrary.ts
new file mode 100644
index 0000000..f140019
--- /dev/null
+++ b/app/server/src/confluence/confluenceLibrary.ts
@@ -0,0 +1,88 @@
+// Investor Flow — Confluence Library (M22, slice 2)
+//
+// Redundancy modeling for the Confluence Signal Engine. Correlated slots that
+// measure the same underlying phenomenon ("golden cross" and "trend alignment"
+// both read the same moving-average structure) are grouped so a rack does not
+// over-count them as independent evidence.
+//
+// Pure data + accessors. NO I/O. ADR-0007: these are evidence-accounting rules,
+// never recommendations.
+
+import { CONFLUENCE_SLOTS, type ConfluenceSlot, type SlotFamily } from './confluenceSlots.ts';
+
+/**
+ * A redundancy group: slots that corroborate each other because they consume
+ * overlapping inputs. Within a group, the first firing slot contributes full
+ * evidence and each additional firing slot contributes a discounted margin
+ * (`REDUNDANCY_DISCOUNT`, applied multiplicatively) so three correlated
+ * bullish slots are not counted as three independent bullish facts.
+ */
+export interface RedundancyGroup {
+  family: SlotFamily;
+  slots: string[];
+}
+
+/** Default discount applied per additional firing slot within a group. */
+export const REDUNDANCY_DISCOUNT = 0.5;
+
+/** Evidence redundancy groups keyed by the phenomenon they all measure. */
+export const REDUNDANCY_GROUPS: RedundancyGroup[] = [
+  { family: 'technical', slots: ['goldenCross', 'trendAlignment', 'pullbackToEMA21'] },
+  { family: 'technical', slots: ['timeSeriesMomentum12_1', 'relativeStrengthVsSpy', 'momentumForward'] },
+  { family: 'technical', slots: ['macdBullish', 'rsiOversold'] },
+  { family: 'technical', slots: ['macdBearish', 'rsiOverbought'] },
+  { family: 'technical', slots: ['relVolume', 'volumeProfileShelf'] },
+  { family: 'institutional', slots: ['instNetActivePositive', 'insiderInformedBuy30d', 'new13da'] },
+  { family: 'macro', slots: ['ratesRegime', 'macroRegimeUp'] },
+  { family: 'macro', slots: ['consumerSentimentLow', 'breadthThrust'] },
+  { family: 'seasonal', slots: ['seasonalFavorableMonth', 'winterHalfOn', 'electionCycleFavorableYear'] },
+  { family: 'flows', slots: ['etfFlowPositive', 'cotPositioning'] },
+];
+
+/** Map slot id -> the group it belongs to (a slot may be in one group). */
+export const slotGroupIndex: ReadonlyMap<string, RedundancyGroup> = new Map(
+  REDUNDANCY_GROUPS.flatMap((g) => g.slots.map((id) => [id, g] as const)),
+);
+
+/** The group a slot belongs to, if any. */
+export function redundancyGroupFor(slotId: string): RedundancyGroup | undefined {
+  return slotGroupIndex.get(slotId);
+}
+
+/**
+ * Effective evidence weight for a set of firing slots *within one group*.
+ * First firing slot contributes 1.0; each additional contributes
+ * `REDUNDANCY_DISCOUNT` of the previous marginal — so N same-group firings
+ * weigh 1 + 0.5 + 0.25 + ... , never N.
+ */
+export function groupWeight(firingSlotIds: string[]): number {
+  if (firingSlotIds.length === 0) return 0;
+  let w = 0;
+  let margin = 1;
+  for (let i = 0; i < firingSlotIds.length; i++) {
+    w += margin;
+    margin *= REDUNDANCY_DISCOUNT;
+  }
+  return w;
+}
+
+/** A flat description of every slot for explain-note / docs rendering. */
+export function slotCatalog(): { id: string; name: string; family: SlotFamily; body: string; explain: string }[] {
+  return CONFLUENCE_SLOTS.map((s) => ({
+    id: s.id,
+    name: s.name,
+    family: s.family,
+    body: s.body,
+    explain: s.explain,
+  }));
+}
+
+/** Convenience: number of slots per family. */
+export function familyCounts(): Record<SlotFamily, number> {
+  const counts: Record<SlotFamily, number> = { technical: 0, institutional: 0, macro: 0, seasonal: 0, flows: 0, sentiment: 0 };
+  for (const s of CONFLUENCE_SLOTS) counts[s.family] += 1;
+  return counts;
+}
+
+/** Re-export the slot type for consumers of the library. */
+export type { ConfluenceSlot } from './confluenceSlots.ts';
\ No newline at end of file
diff --git a/app/server/src/confluence/confluenceRack.ts b/app/server/src/confluence/confluenceRack.ts
new file mode 100644
index 0000000..48cb1b5
--- /dev/null
+++ b/app/server/src/confluence/confluenceRack.ts
@@ -0,0 +1,237 @@
+// Investor Flow — Confluence Rack (M22, slice 2)
+//
+// The aggregator that turns per-slot firing states into a redundancy-aware
+// "picture quality" assessment for a symbol. It:
+//   • buckets fired slots into bullish vs bearish evidence (exit slots count as
+//     bear evidence),
+//   • applies redundancy-group discounting so correlated slots are not
+//     triple-counted,
+//   • labels the overall picture with an evidence-based quality tier,
+//   • detects picture *changes* between consecutive evaluations for the
+//     confluence-change alert producer.
+//
+// Pure: consumes slot assessments, emits a labeled evaluation. NO I/O.
+// ADR-0007: the output is a description of the picture ("strong-bullish",
+// "mixed") — evidence, never a directive to buy or sell.
+
+import { CONFLUENCE_SLOTS, type ConfluenceSlot } from './confluenceSlots.ts';
+import { groupWeight, redundancyGroupFor } from './confluenceLibrary.ts';
+
+/** Per-slot assessment fed into a rack evaluation. */
+export interface SlotAssessment {
+  id: string;
+  /** fired = the slot's evidence is present; not-fired = data seen, evidence absent. */
+  state: 'fired' | 'not-fired';
+  /** Optional 0..1 confidence of the evaluation itself. */
+  confidence?: number;
+  /** Optional free-text note for explain output. */
+  note?: string;
+}
+
+/** A symbol's confluence evaluation for one as-of timestamp. */
+export interface ConfluenceEvaluation {
+  symbol: string;
+  /** ISO date (YYYY-MM-DD) the evaluation is for. */
+  asOf: string;
+  assessments: SlotAssessment[];
+  /** Redundancy-discounted bullish evidence weight. */
+  bullEvidence: number;
+  /** Redundancy-discounted bearish (bear + exit) evidence weight. */
+  bearEvidence: number;
+  /** Raw counts (before redundancy discounting). */
+  bullCount: number;
+  bearCount: number;
+  /** Slots assessed (with data); sheds light on coverage. */
+  assessedCount: number;
+  /** Net bullish evidence (bull - bear). Positive = more bullish evidence. */
+  netEvidence: number;
+  /** Total recognized evidence (bull + bear). */
+  totalEvidence: number;
+  /** Evidence-based picture label. */
+  quality: PictureQuality;
+}
+
+/** Evidence-based picture labels, ordered from most to least constructive. */
+export type PictureQuality =
+  | 'strong-bullish'
+  | 'moderate-bullish'
+  | 'weak-bullish'
+  | 'mixed'
+  | 'weak-bearish'
+  | 'moderate-bearish'
+  | 'strong-bearish'
+  | 'sparse';
+
+/** Ordering for change detection / tier comparison. */
+export const QUALITY_RANK: Record<PictureQuality, number> = {
+  'strong-bullish': 7,
+  'moderate-bullish': 6,
+  'weak-bullish': 5,
+  'mixed': 4,
+  'weak-bearish': 3,
+  'moderate-bearish': 2,
+  'strong-bearish': 1,
+  'sparse': 0,
+};
+
+/** Minimum total evidence before the picture is meaningful rather than 'sparse'. */
+export const MIN_TOTAL_EVIDENCE = 1.0;
+/** Evidence ratio that separates "bullish/bearish" from "mixed". */
+export const DIRECTION_RATIO = 0.6;
+/** Total evidence marking a "strong" vs "moderate" picture. */
+export const STRONG_EVIDENCE = 4.0;
+/** Total evidence marking a "moderate" vs "weak" picture. */
+export const MODERATE_EVIDENCE = 2.0;
+
+/** Which side a slot's firing evidence serves. */
+export function slotBody(slot: ConfluenceSlot): 'bull' | 'bear' {
+  return slot.body === 'exit' ? 'bear' : slot.body;
+}
+
+/** Is `id` a known slot? */
+function isKnownSlot(id: string): boolean {
+  return CONFLUENCE_SLOTS.some((s) => s.id === id);
+}
+
+/**
+ * Evaluate a rack: bucket the fired assessments into bull/bear evidence,
+ * discount per redundancy group, and label the picture.
+ */
+export function evaluateRack(
+  symbol: string,
+  asOf: string,
+  assessments: SlotAssessment[],
+): ConfluenceEvaluation {
+  const slotById = new Map(CONFLUENCE_SLOTS.map((s) => [s.id, s]));
+
+  const firedBull: string[] = [];
+  const firedBear: string[] = [];
+  let assessedCount = 0;
+
+  for (const a of assessments) {
+    if (!isKnownSlot(a.id)) continue;
+    assessedCount += 1;
+    if (a.state !== 'fired') continue;
+    const body = slotBody(slotById.get(a.id)!);
+    if (body === 'bull') firedBull.push(a.id);
+    else firedBear.push(a.id);
+  }
+
+  // Group fired slots by redundancy group; ungrouped slots get singleton buckets.
+  const bucket = (id: string): string => {
+    const g = redundancyGroupFor(id);
+    return g ? `g:${g.slots.join(',')}` : `solo:${id}`;
+  };
+
+  const bullGroups = new Map<string, string[]>();
+  const bearGroups = new Map<string, string[]>();
+  for (const id of firedBull) {
+    const key = bucket(id);
+    const arr = bullGroups.get(key) ?? [];
+    arr.push(id);
+    bullGroups.set(key, arr);
+  }
+  for (const id of firedBear) {
+    const key = bucket(id);
+    const arr = bearGroups.get(key) ?? [];
+    arr.push(id);
+    bearGroups.set(key, arr);
+  }
+
+  let bullEvidence = 0;
+  for (const ids of bullGroups.values()) bullEvidence += groupWeight(ids);
+  let bearEvidence = 0;
+  for (const ids of bearGroups.values()) bearEvidence += groupWeight(ids);
+
+  const netEvidence = bullEvidence - bearEvidence;
+  const totalEvidence = bullEvidence + bearEvidence;
+  const quality = labelQuality(bullEvidence, bearEvidence, totalEvidence);
+
+  return {
+    symbol,
+    asOf,
+    assessments,
+    bullEvidence,
+    bearEvidence,
+    bullCount: firedBull.length,
+    bearCount: firedBear.length,
+    assessedCount,
+    netEvidence,
+    totalEvidence,
+    quality,
+  };
+}
+
+/** Label the picture from the discounted evidence totals. Pure. */
+export function labelQuality(bullEvidence: number, bearEvidence: number, totalEvidence: number): PictureQuality {
+  if (totalEvidence < MIN_TOTAL_EVIDENCE) return 'sparse';
+
+  const ratio = bullEvidence / totalEvidence;
+  if (ratio >= DIRECTION_RATIO) return magnitudeLabel(bullEvidence, 'bullish');
+  if (ratio <= 1 - DIRECTION_RATIO) return magnitudeLabel(bearEvidence, 'bearish');
+  return 'mixed';
+}
+
+function magnitudeLabel(strongSideEvidence: number, side: 'bullish' | 'bearish'): PictureQuality {
+  if (strongSideEvidence >= STRONG_EVIDENCE) return side === 'bullish' ? 'strong-bullish' : 'strong-bearish';
+  if (strongSideEvidence >= MODERATE_EVIDENCE) return side === 'bullish' ? 'moderate-bullish' : 'moderate-bearish';
+  return side === 'bullish' ? 'weak-bullish' : 'weak-bearish';
+}
+
+/** A detected change between two evaluations. */
+export interface ConfluenceChange {
+  changed: boolean;
+  /** Direction of the picture change from the consumer's perspective. */
+  changeType: 'improved' | 'deteriorated' | 'unchanged';
+  current: PictureQuality;
+  previous: PictureQuality | null;
+  netEvidenceShift: number;
+  /** Slot ids whose firing state flipped between the two evaluations. */
+  flippedSlotIds: string[];
+}
+
+/** Minimum net-evidence shift that counts as a *noteworthy* change. */
+export const MIN_CHANGE_SHIFT = 0.35;
+
+/**
+ * Detect whether the picture changed meaningfully between two consecutive
+ * evaluations of the same symbol. Used by the confluence-change alert producer.
+ * Pure.
+ */
+export function detectPictureChange(
+  previous: ConfluenceEvaluation | null,
+  current: ConfluenceEvaluation,
+): ConfluenceChange {
+  if (previous === null) {
+    return { changed: false, changeType: 'unchanged', current: current.quality, previous: null, netEvidenceShift: 0, flippedSlotIds: [] };
+  }
+
+  const flippedSlotIds = flippedSlots(previous.assessments, current.assessments);
+  const netEvidenceShift = current.netEvidence - previous.netEvidence;
+
+  const rankShift = QUALITY_RANK[current.quality] - QUALITY_RANK[previous.quality];
+  if (rankShift > 0 || (rankShift === 0 && netEvidenceShift >= MIN_CHANGE_SHIFT)) {
+    return { changed: true, changeType: 'improved', current: current.quality, previous: previous.quality, netEvidenceShift, flippedSlotIds };
+  }
+  if (rankShift < 0 || (rankShift === 0 && netEvidenceShift <= -MIN_CHANGE_SHIFT)) {
+    return { changed: true, changeType: 'deteriorated', current: current.quality, previous: previous.quality, netEvidenceShift, flippedSlotIds };
+  }
+  return { changed: false, changeType: 'unchanged', current: current.quality, previous: previous.quality, netEvidenceShift, flippedSlotIds };
+}
+
+/** Fired-state flips between two assessment sets. Pure. */
+export function flippedSlots(prev: SlotAssessment[], curr: SlotAssessment[]): string[] {
+  const prevMap = new Map(prev.map((a) => [a.id, a.state]));
+  const currMap = new Map(curr.map((a) => [a.id, a.state]));
+  const flipped: string[] = [];
+  for (const id of prevMap.keys()) {
+    if (currMap.has(id) && prevMap.get(id) !== currMap.get(id)) flipped.push(id);
+  }
+  for (const id of currMap.keys()) {
+    if (!prevMap.has(id)) flipped.push(id);
+  }
+  return flipped;
+}
+
+/** Marker re-export so the module documents the exit→bear convention. */
+export { isBearEvidence } from './confluenceSlots.ts';
\ No newline at end of file
diff --git a/app/server/src/confluence/confluenceSlots.ts b/app/server/src/confluence/confluenceSlots.ts
new file mode 100644
index 0000000..420d68e
--- /dev/null
+++ b/app/server/src/confluence/confluenceSlots.ts
@@ -0,0 +1,116 @@
+// Investor Flow — Confluence Slot Catalog (M22, slice 2)
+//
+// The 34-slot confluence inventory for the Confluence Signal Engine. Each slot is
+// a named, independently-evaluable check whose *firing* state contributes bullish
+// or bearish evidence about a symbol's entry/exit quality.
+//
+// Pure data + retrieval helpers: NO I/O. ADR-0007: slots describe "quality of the
+// picture" — never a buy/sell recommendation. Each slot has an ADR-safe explain
+// note phrased as evidence ("relative strength is in the upper decile") rather
+// than a directive ("buy").
+
+/** Slot body: which side of the picture a firing slot supports. */
+export type SlotBody = 'bull' | 'bear' | 'exit';
+
+/**
+ * "exit" is a subclass of bear evidence: it indicates weakening conditions for an
+ * existing position rather than a fresh short thesis. Exit slots are still counted
+ * as bear evidence (risk-off side of the picture).
+ */
+export const isBearEvidence = (body: SlotBody): boolean => body === 'bear' || body === 'exit';
+
+/** The seven confluence families (independent evidence axes). */
+export type SlotFamily =
+  | 'technical'
+  | 'institutional'
+  | 'macro'
+  | 'seasonal'
+  | 'flows'
+  | 'sentiment';
+
+/** Candle granularity a slot consumes (daily is the engine default). */
+export type SlotGranularity = '1d' | '1wk';
+
+/**
+ * A confluence slot definition. `id` is the stable machine key; `name` plain
+ * English; `explain` is the ADR-0007-safe evidence sentence surfaced to users.
+ */
+export interface ConfluenceSlot {
+  id: string;
+  name: string;
+  family: SlotFamily;
+  body: SlotBody;
+  /** Default candle granularity the slot evaluates on. */
+  granularity: SlotGranularity;
+  /** Short evidence sentence shown in UI / explain notes. */
+  explain: string;
+}
+
+/** Complete 34-slot confluence catalog in evaluation order. */
+export const CONFLUENCE_SLOTS: ConfluenceSlot[] = [
+  // ---------------------------------------------------------------- technical
+  { id: 'goldenCross', name: 'Golden Cross', family: 'technical', body: 'bull', granularity: '1wk', explain: 'The 50-window average has crossed above the 200-window average, a widely-watched trend-quality marker.' },
+  { id: 'deathCross', name: 'Death Cross', family: 'technical', body: 'exit', granularity: '1wk', explain: 'The 50-window average has crossed below the 200-window average, a commonly-cited trend-weakening marker.' },
+  { id: 'trendAlignment', name: 'Trend Alignment', family: 'technical', body: 'bull', granularity: '1wk', explain: 'Price sits above both the 50- and 200-window averages, indicating an aligned long-term trend.' },
+  { id: 'timeSeriesMomentum12_1', name: '12-1 Momentum', family: 'technical', body: 'bull', granularity: '1d', explain: 'The 12-month-return-minus-1-month gauge is positive, a period-validated trend-following signal.' },
+  { id: 'relativeStrengthVsSpy', name: 'Relative Strength vs SPY', family: 'technical', body: 'bull', granularity: '1d', explain: 'The symbol is outperforming the S&P 500 over the 12-1 momentum window.' },
+  { id: 'rsiOversold', name: 'RSI Oversold', family: 'technical', body: 'bull', granularity: '1d', explain: 'Momentum gauge oversold, a bounce-prone condition after sustained weakness.' },
+  { id: 'rsiOverbought', name: 'RSI Overbought', family: 'technical', body: 'exit', granularity: '1d', explain: 'Momentum gauge overbought, a stretched condition after sustained strength.' },
+  { id: 'macdBullish', name: 'MACD Bullish', family: 'technical', body: 'bull', granularity: '1d', explain: 'The MACD line sits above its signal line, indicating positive short-term momentum.' },
+  { id: 'macdBearish', name: 'MACD Bearish', family: 'technical', body: 'exit', granularity: '1d', explain: 'The MACD line sits below its signal line, indicating negative short-term momentum.' },
+  { id: 'relVolume', name: 'Relative Volume', family: 'technical', body: 'bull', granularity: '1d', explain: 'Current volume is notably above its recent average, increasing the confidence contribution of concurrent signals.' },
+  { id: 'volumeProfileShelf', name: 'Volume-Profile Shelf', family: 'technical', body: 'bull', granularity: '1d', explain: 'Price is trading near a high-volume shelf of the volume profile, historically a supported level.' },
+  { id: 'fibLevelCluster', name: 'Fib Level Cluster', family: 'technical', body: 'bull', granularity: '1d', explain: 'Price is near a Fibonacci level that confluences with other technical levels, a commonly-watched confluence.' },
+  { id: 'volatilityRegimeLow', name: 'Low-Volatility Regime', family: 'technical', body: 'bull', granularity: '1d', explain: 'Realized volatility is in a low percentile versus its own history, historically favorable to trend continuation.' },
+  { id: 'pullbackToEMA21', name: 'Pullback to EMA-21', family: 'technical', body: 'bull', granularity: '1d', explain: 'Price has pulled back toward a rising short-term average within a broader uptrend.' },
+  { id: 'momentumForward', name: 'Forward Momentum', family: 'technical', body: 'bull', granularity: '1d', explain: 'Short-horizon rate of change is positive on a daily basis.' },
+
+  // ------------------------------------------------------------- institutional
+  { id: 'instNetActivePositive', name: 'Active Funds Net Positive', family: 'institutional', body: 'bull', granularity: '1d', explain: 'Net institutional flows from active managers are positive in the latest observable quarter.' },
+  { id: 'insiderInformedBuy30d', name: 'Informed Insider Buys (30d)', family: 'institutional', body: 'bull', granularity: '1d', explain: 'Informed insiders have been net buyers over the trailing 30 days, the strongest insider-feedback class.' },
+  { id: 'insiderInformedSell', name: 'Informed Insider Selling', family: 'institutional', body: 'exit', granularity: '1d', explain: 'Informed insiders have been net sellers over the trailing window, a caution signal.' },
+  { id: 'new13da', name: 'New 13D/A Disclosure', family: 'institutional', body: 'bull', granularity: '1d', explain: 'A fresh Schedule 13D/A was disclosed, signaling an activist or concentrated position just formed.' },
+  { id: 'instBuyZoneProximity', name: 'Institutional Buy-Zone Proximity', family: 'institutional', body: 'bull', granularity: '1d', explain: 'Price is near the estimated institutional accumulation zone from the flow engine estimation model.' },
+
+  // -------------------------------------------------------------------- macro
+  { id: 'consumerSentimentLow', name: 'Consumer Sentiment Low', family: 'macro', body: 'bull', granularity: '1wk', explain: 'The consumer-sentiment gauge is in its low historical band, a contrarian backdrop historically followed by higher forward returns.' },
+  { id: 'consumerSentimentHigh', name: 'Consumer Sentiment High', family: 'macro', body: 'exit', granularity: '1wk', explain: 'The consumer-sentiment gauge is in its high historical band, a backdrop historically followed by lower forward returns.' },
+  { id: 'ratesRegime', name: 'Falling-Rates Regime', family: 'macro', body: 'bull', granularity: '1wk', explain: 'Interest-rate expectations are on a falling trajectory, historically supportive of equities.' },
+  { id: 'macroRegimeUp', name: 'Macro Regime Up', family: 'macro', body: 'bull', granularity: '1wk', explain: 'Leading macroeconomic indicators are on a rising trajectory.' },
+  { id: 'breadthThrust', name: 'Breadth Thrust', family: 'macro', body: 'bull', granularity: '1wk', explain: 'Market breadth has surged from oversold, a historically constructive regime marker.' },
+
+  // -------------------------------------------------------------- seasonal
+  { id: 'seasonalFavorableMonth', name: 'Seasonally Favorable Month', family: 'seasonal', body: 'bull', granularity: '1d', explain: 'The calendar month is historically favorable for the symbol, a mild cyclical tailwind.' },
+  { id: 'winterHalfOn', name: 'Winter Half Indicator', family: 'seasonal', body: 'bull', granularity: '1d', explain: 'The November-April half of the year is historically favorable, a mild cyclical tailwind.' },
+  { id: 'electionCycleFavorableYear', name: 'Election-Cycle Favorable Year', family: 'seasonal', body: 'bull', granularity: '1d', explain: 'The current year of the presidential cycle is historically above-average for equities.' },
+  { id: 'nearTurnOfMonth', name: 'Turn-of-Month', family: 'seasonal', body: 'bull', granularity: '1d', explain: 'The calendar is near the turn of the month, historically a benign pocket of liquidity.' },
+  { id: 'nearQuarterEnd', name: 'Near Quarter-End', family: 'seasonal', body: 'bull', granularity: '1d', explain: 'The calendar is near quarter-end, historically associated with institutional window dressing.' },
+
+  // -------------------------------------------------------------------- flows
+  { id: 'etfFlowPositive', name: 'ETF Inflows', family: 'flows', body: 'bull', granularity: '1d', explain: 'The symbol is seeing net positive ETF inflows, evidence of systematic buying pressure.' },
+  { id: 'cotPositioning', name: 'COT Positioning', family: 'flows', body: 'bull', granularity: '1wk', explain: 'CFTC futures positioning is in a supportive band, evidence of aligned speculative flows.' },
+  { id: 'rotationIncipient', name: 'Incipient Sector Rotation', family: 'flows', body: 'bull', granularity: '1d', explain: 'Cross-sectional relative-strength analysis flags the symbol as an emerging rotation target.' },
+
+  // ---------------------------------------------------------------- sentiment
+  { id: 'commentatorSentiment', name: 'Informed Commentator Sentiment', family: 'sentiment', body: 'bull', granularity: '1d', explain: 'Informed commentators tracked via the configured sentiment source are net-positive on the symbol in the measurement window.' },
+];
+
+/** Indexed by slot id for O(1) lookup. */
+export const confluenceslotById: ReadonlyMap<string, ConfluenceSlot> = new Map(
+  CONFLUENCE_SLOTS.map((s) => [s.id, s]),
+);
+
+/** All slot ids, stable evaluation order. */
+export const CONFLUENCE_SLOT_IDS: readonly string[] = CONFLUENCE_SLOTS.map((s) => s.id);
+
+/** Type guard: is `id` a known confluence slot? */
+export function isConfluenceSlot(id: string): id is string {
+  return confluenceslotById.has(id);
+}
+
+/** Slots grouped by family. */
+export function slotsByFamily(family: SlotFamily): ConfluenceSlot[] {
+  return CONFLUENCE_SLOTS.filter((s) => s.family === family);
+}
+
+/** DSL helper removed — catalog written literally above for full control. */
\ No newline at end of file
diff --git a/app/server/src/confluence/sentimentSource.ts b/app/server/src/confluence/sentimentSource.ts
new file mode 100644
index 0000000..b2aa676
--- /dev/null
+++ b/app/server/src/confluence/sentimentSource.ts
@@ -0,0 +1,55 @@
+// Investor Flow — SentimentSource interface (M22, slice 2)
+//
+// Abstraction over "informed commentator sentiment". Phase 1 wires tweet
+// sentiment from the existing X cookie adapter; later phases can swap in other
+// sources (X radar, curated news sentiment) behind the same interface.
+//
+// The rack consumes this so the `commentatorSentiment` slot is evaluable without
+// coupling the confluence engine to any specific social provider. ADR-0007 / ADR-0009:
+// sentiment here is an evidence input (with sample size + window), sourced only
+// through rate-limited adapters — never a recommendation.
+
+/**
+ * A windowed, sample-sized sentiment measurement for one symbol.
+ * `netSentiment` is in [-1, 1]: +1 = fully bullish, -1 = fully bearish.
+ * `sampleSize` and `collectedAt` keep the measurement honest: thin windows are
+ * low-authority evidence and the rack can discount them.
+ */
+export interface CommentatorSentiment {
+  symbol: string;
+  /** Trading days the measurement window spans. */
+  windowDays: number;
+  /** Number of individual comments/tweets sampled in the window. */
+  sampleSize: number;
+  /** Bullish fraction minus bearish fraction in [-1, 1]. */
+  netSentiment: number;
+  bullCount: number;
+  bearCount: number;
+  neutralCount: number;
+  /** ISO timestamp the window closed. */
+  collectedAt: string;
+}
+
+/**
+ * Any provider that can produce a sentiment measurement for a symbol.
+ * Returning `null` means the source has no usable data (not an error).
+ */
+export interface SentimentSource {
+  readonly kind: 'commentatorSentiment';
+  /** Minimum sample size before the rack treats a measurement as evidence. */
+  readonly minSampleSize: number;
+  /** Fetch the latest windowed measurement for `symbol`. Safe to call often; implementers throttle. */
+  fetch(symbol: string, opts?: { windowDays?: number }): Promise<CommentatorSentiment | null>;
+}
+
+/** Convenience: does a measurement carry enough evidence to be admitted? */
+export function sentimentIsUsable(src: SentimentSource, s: CommentatorSentiment | null): s is CommentatorSentiment {
+  return s !== null && s.sampleSize >= src.minSampleSize && s.netSentiment !== 0;
+}
+
+/** Directional read of a usable sentiment measurement. */
+export function sentimentDirection(s: CommentatorSentiment): 'bull' | 'bear' | 'neutral' {
+  if (s.netSentiment > 0) return 'bull';
+  if (s.netSentiment < 0) return 'bear';
+  return 'neutral';
+}
\ No newline at end of file