# ReNoUn — Structural Pattern Analysis Engine (Full Reference)

> ReNoUn is a 17-channel structural analysis MCP server for AI conversations and financial risk management. It measures Recurrence, Novelty, and Unity patterns without analyzing semantic content. Patent Pending #63/923,592.

## What ReNoUn Does

ReNoUn is a structural observability engine. It measures HOW conversations and financial data behave structurally — not WHAT they contain. It detects loops, breakthroughs, convergence, and disorder across 17 independent channels.

For conversations: feed utterances as [{speaker, text}] arrays. ReNoUn returns a Dialectical Health Score (0.0-1.0), constellation patterns, per-channel breakdowns, and actionable recommendations.

For finance: feed OHLCV candle data (open, high, low, close, volume). ReNoUn returns the same structural analysis plus an exposure scalar (0.0-1.0) that reduces position sizing during structural disorder.

## The 17 Channels

### Recurrence (Re1-Re5) — Pattern Repetition
- Re1 Lexical: vocabulary recycling rate
- Re2 Syntactic: sentence structure repetition
- Re3 Rhythmic: turn length and pacing consistency
- Re4 Turn-Taking: predictability of speaker alternation
- Re5 Self-Interruption: repeated self-correction patterns

### Novelty (No1-No6) — New Information
- No1 Lexical: new vocabulary entering the conversation
- No2 Syntactic: new sentence structures appearing
- No3 Rhythmic: sudden pacing changes
- No4 Turn-Taking: unexpected interaction shifts
- No5 Self-Interruption: new correction patterns
- No6 Global Vocabulary Rarity: statistically rare words

### Unity (Un1-Un6) — Coherence
- Un1 Lexical Cohesion: vocabulary linking across turns
- Un2 Syntactic Cohesion: consistent grammar patterns
- Un3 Rhythmic Cohesion: pacing consistency over time
- Un4 Interactional Cohesion: stable interaction patterns
- Un5 Anaphoric Cohesion: building on prior turns
- Un6 Global Structural Symmetry: first-half/second-half mirroring

All channels produce values between 0.0 and 1.0. Higher = more of that quality.

## The 8 Constellation Patterns

1. CLOSED_LOOP — High recurrence, low novelty, high unity. Conversation stuck in a repetitive cycle.
2. HIGH_SYMMETRY — Balanced but rigid structure. Predictable but not productive.
3. PATTERN_BREAK — Recurrence drops, novelty spikes. A disruption in established patterns.
4. CONVERGENCE — Unity rising across channels. Productive movement toward integration.
5. SCATTERING — Low recurrence, low unity, high novelty. Structure is fragmenting.
6. REPEATED_DISRUPTION — Multiple novelty spikes without recovery. Escalating instability.
7. DIP_AND_RECOVERY — Temporary disruption followed by restored stability. Resilience.
8. SURFACE_VARIATION — Surface-level novelty without deep structural change.

## Finance Module

The finance module applies the same 17-channel engine to OHLCV candle data. Key outputs:

- DHS: structural health of price action (0.0-1.0)
- Exposure scalar: recommended position sizing multiplier (0.0-1.0)
- Constellation: current structural state of the market
- Stress level: structural stress intensity
- Crash regime: boolean flag for extreme structural disorder

### Validated Results
- 100% bounded regime accuracy (126+ graded), 245+ total predictions
- Every prediction is public, timestamped, and graded — no other crypto signal service does this
- Average improvement: 5.7 percentage points
- Black swan events caught: COVID crash (Mar 2020), LUNA collapse (May 2022), FTX contagion (Nov 2022), China mining ban (May 2021) — 7/8 events validated
- Cost: approximately 0.1 Sharpe ratio
- Assets tested: BTC, ETH, SOL, BNB, ADA, DOGE, XRP, AVAX, DOT
- Timeframes tested: 1m, 5m, 1h, 4h, 1d

### Exposure Scalar Details
- v2 uses asymmetric EMA smoothing: alpha_down=0.6 (fast de-risk), alpha_up=0.3 (slow recovery)
- Constellation persistence scoring: 1 window = 0.5x weight, 2 windows = 0.8x, 3+ = 1.0x
- Floor of 0.20: exposure never goes to zero

## MCP Tools API

### renoun_analyze
Input: {utterances: [{speaker: string, text: string}], weights?: number[], tags?: object[], weighting_mode?: "weight"|"exclude"|"segment"}
Output: {dhs: number, loop: number, channels: object, constellations: array, recommendations: array}

### renoun_health_check
Input: {utterances: [{speaker: string, text: string}]}
Output: {dhs: number, assessment: string, dominant_constellation: string, summary: string}

### renoun_compare
Input: {utterances_a?: array, utterances_b?: array, result_a?: object, result_b?: object}
Output: {delta_dhs: number, trend: string, constellation_transition: string, top_channel_shifts: array}

### renoun_pattern_query
Input: {action: "save"|"list"|"query"|"trend", session_name?: string, result?: object, tags?: string[]}
Output: varies by action

### renoun_steer
Input: {action: "add_turns"|"get_status"|"clear_session"|"list_sessions", session_id?: string, utterances?: array}
Output: {signals: array, buffer_size: number, windows_analyzed: number}

### renoun_finance_analyze
Input: {klines: [{open, high, low, close, volume}], symbol?: string, timeframe?: "1m"|"5m"|"15m"|"1h"|"4h"|"1d"}
Output: {dhs: number, constellations: array, stress: number, exposure: {raw_v1: number, smoothed_v2: number}, crash_regime: boolean}

## Installation

pip install renoun-mcp

Or run directly:
python -m renoun_mcp

## Pricing

- Metered (default): 50 free calls/day, $0.02/call beyond that, self-provisioning, no credit card
- Enterprise: Contact us for custom limits and SLAs

## Links

- Website: https://harrisoncollab.com
- GitHub: https://github.com/98lukehall/renoun-mcp
- PyPI: https://pypi.org/project/renoun-mcp/
- Pricing: https://harrisoncollab.com/pricing.html
- Contact: 98lukehall@gmail.com
- Author: Luke Hall, Harrison Collab
- Patent: Pending #63/923,592
- Version: 1.3.1