convez.ai
PluginAutomaçãoAvançado
yogsoth-ai

Convergence — Motor de Decisão Universal

yogsoth-ai/convergence

Transforma dados em decisões validadas via 6 campanhas de convergência paralelas

Ver no GitHub
InicialPluginScore · de 0 a 100

Plugin que transforma conjuntos desestruturados de candidatos em seleções ranqueadas, portfolios balanceados e decisões validadas através de 6 campanhas de convergência paralelas. Implementa métodos rigorosos de scoring multi-critério, ranking por comparação pareada, consenso estruturado, avaliação de viabilidade, otimização de portfolio e verificação adversarial.

Instalação

$npx convez-ai yogsoth-ai/convergence

1 comando no terminal: detecta o formato do repositório (marketplace ou plugin avulso) e instala via Claude Code. Requer Node 18+ e o Claude Code instalado.

Métricas

Estrelas
2
Forks
0
Acompanhando
0
Saúde
50%
Último commit
há 5 meses
Versão
—
Issues abertas
0
Contribuidores
1

Casos de uso

  • Seleção de candidatos ou projetos contra múltiplos critérios ponderados
  • Resolução de desacordos entre perspectivas divergentes usando Delphi e síntese dialética
  • Otimização de portfolio balanceado com Markowitz, Knapsack e Real Options
  • Avaliação de viabilidade técnica usando TRL, NASSS, Stage-Gate e TRIZ
  • Verificação adversarial de decisões via Red Teaming e Pré-mortem

Veredito

Ferramenta rigorosa e modular para decisões complexas que exigem múltiplas perspectivas e validação adversarial. Adequada para pesquisa, P&D e seleção estratégica — complexidade justificada pelo rigor que entrega, não por marketing.

README

A tradução para o português deste README ainda está na fila.

Decisions are not made in the moment of choosing — they are made in the structure of evaluation. A bad framework produces confident wrong answers. A good framework surfaces the right trade-offs and lets the decision survive scrutiny.

🎯 Convergence

Universal Convergence Engine for Research Decision-Making

🧭 Part of the De-Anthropocentric Research Engine. This repository is one of nine composable research packages that make up DARE — the full autonomous research-orchestration system. DARE bundles this package together with the others into a single self-contained clone, unified under one orchestrator. To use these skills as intended — with the spec-driven orchestrator and cross-package routing — clone the main repository rather than this repo alone.

Transforms unstructured candidate sets into ranked selections, balanced portfolios, and validated decisions through 6 parallel convergence campaigns spanning scoring, ranking, consensus, feasibility, optimization, and adversarial verification.


⚡ What It Does

  • 📊 Multi-criteria scoring — score and rank candidates against weighted criteria using AHP, TOPSIS, ELECTRE, PROMETHEE, and non-compensatory screening methods
  • 🔄 Pairwise ranking — produce global rankings via systematic pairwise comparisons with consistency checking and convergence detection
  • 🤝 Structured consensus — resolve disagreements between multiple perspectives using Delphi, Nominal Group Technique, and dialectical synthesis
  • 🔬 Feasibility assessment — evaluate real-world viability using TRL, NASSS, Stage-Gate, TRIZ, TOC, and parametric estimation
  • 💼 Portfolio optimization — select balanced combinations using Markowitz, Knapsack, Pareto, Real Options, MAP-Elites, and minimax regret
  • ⚔️ Steel-manning — adversarial verification via Devil's Advocacy, Pre-mortem, Red Teaming, and Dialectical Inquiry

🏗️ Architecture

ENTRY.md (campaign router)
  → Campaign (6): self-contained convergence paradigm
    → Strategy: selected by convergence intent/scenario
      → Tactic: multi-step orchestration pattern
        → SOP: single operation (subagent or import)

Campaign Routing

SignalCampaign
score/rank candidates against multiple criteriamulti-criteria-scoring
produce global ranking via pairwise comparisonspairwise-ranking
multiple perspectives disagree, need convergencestructured-consensus
assess feasibility/readiness of candidatesfeasibility-assessment
select a balanced portfolio from candidatesportfolio-optimization
verify rejected candidates, stress-test winnerssteel-manning

Multi-Campaign Orchestration

CC composes campaigns autonomously:

  • Serial: scoring → steel-manning → portfolio
  • Parallel: scoring + pairwise → take intersection
  • Backtrack: steel-manning rejects → re-enter scoring
  • Nested: consensus determines weights → scoring uses those weights
  • Skip: if only 3 candidates with clear criteria, pairwise alone suffices

📦 Manifest

CampaignStrategiesTacticsSOPs
multi-criteria-scoring5310
pairwise-ranking539
structured-consensus5310
feasibility-assessment5310
portfolio-optimization5310
steel-manning539
Total301858

Plus 3 shared SOPs (saturation-detection, sensitivity-analysis, multi-stakeholder-simulation) and 5 import SOPs (web-search, web-research, paper-overview, paper-search, paper-research).


🔌 Dependencies

DependencyWhat It Provides
web-browsingweb-search + web-research (import SOPs)
literature-enginepaper-overview + paper-search + paper-research (import SOPs)
subagent-spawningSubagent dispatch conventions
context-managementcontext-init + context-checkpoint
wiki-vault MCPKnowledge persistence to graph

🧪 Testing

# Integration test scenarios (7 end-to-end)
tests/integration-prompt.md

Each scenario validates routing, output structure, budget enforcement, and hard-gate compliance.


📄 License

Apache-2.0


A component of the De-Anthropocentric Research Engine, part of the Yogsoth AI ecosystem. Built by Pthahnix.

Ver o original no GitHub

Bom para

Relacionados