convez.ai
AgenteProdutividadeAvançado
Yuvraj2305

Expert Council Orchestrator

Yuvraj2305/expert-council-orchestrator

Conselho de IA paralelo para análise multidimensional e decisões inteligentes

Ver no GitHub
InicialPluginScore · de 0 a 100

Agent que orquestra múltiplas personas de IA com expertise distintos para deliberação paralela sobre questões complexas. Simula um conselho de especialistas debatendo simultaneamente, com resolução de conflitos e síntese inteligente das perspectivas divergentes para decisões mais robustas.

Instalação

$npx convez-ai Yuvraj2305/expert-council-orchestrator

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
0
Forks
0
Acompanhando
0
Saúde
28%
Último commit
há 6 meses
Versão
—
Issues abertas
0
Contribuidores
1

Casos de uso

  • Pesquisa acadêmica e análise aprofundada de temas complexos
  • Estratégia empresarial e tomada de decisão executiva
  • Brainstorming criativo com múltiplas perspectivas disciplinares

Veredito

Agent sofisticado que implementa deliberação multi-persona com qualidade genuína. Excelente para contextos onde análise profunda e síntese de perspectivas conflitantes agregam valor real, embora requeira configuração técnica e orquestração cuidadosa de personas para evitar redundância.

README

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

The Expert Council Orchestrator: Parallel AI Persona Swarm for Deep Research & Decision Intelligence

Download

🧠 What Is This? (And Why It Matters)

Imagine convening a roundtable of the world's sharpest minds—a historian, a physicist, a poet, a strategist, and a contrarian—all in the same room, debating your question from their unique vantage points, cross-pollinating ideas, and arriving at a synthesized conclusion you could never reach alone.

That's exactly what The Expert Council Orchestrator does. It's not a chatbot. It's a parallel cognition engine that spawns multiple AI personas—each with a distinct domain expertise, personality bias, and cognitive style—to simultaneously analyze, critique, and build upon your input. The result? A multi-layered, conflict-resolved, and nuanced output that mimics a high-stakes boardroom deliberation.

Built on the architecture of https://Yuvraj2305.github.io (the original council concept), this repository extends the idea into a production-ready framework for enterprise decision-making, academic research, and creative brainstorming. It's your personal think tank on demand.

2026 Update: This version now supports Claude API, OpenAI API, and a hybrid mode where both models debate each other. Because the best ideas come from productive friction.


🚀 Quick Start: Activate Your Council

Prerequisites

  • Python 3.10+
  • API keys for OpenAI (GPT-4/3.5) and/or Anthropic (Claude 3 Opus)
  • A burning question that deserves more than one answer

Installation

git clone https://github.com/your-org/expert-council-orchestrator.git
cd expert-council-orchestrator
pip install -r requirements.txt

One-Click Launch

Download

Or run the Streamlit UI:

streamlit run app.py

🧩 Core Architecture & Workflow

The following Mermaid diagram illustrates the parallel persona deliberation flow:

graph TD
    A[User Input] --> B{Input Router}
    B --> C[Persona 1: Historian]
    B --> D[Persona 2: Scientist]
    B --> E[Persona 3: Economist]
    B --> F[Persona 4: Philosopher]
    B --> G[Persona 5: Contrarian]
    C --> H[Parallel Processing Layer]
    D --> H
    E --> H
    F --> H
    G --> H
    H --> I[Conflict Resolution Engine]
    I --> J[Cross-Pollination Loop]
    J --> K[Synthesized Council Report]
    K --> L[Output: Markdown / JSON / Voice]
    L --> M[Local Storage]
    L --> N[Export to Notion / Obsidian]

How it works:
Your input is broadcast simultaneously to a configurable number of expert personas. Each persona generates a response independently, unaware of the others. Then a Conflict Resolution Engine (a meta-persona) identifies contradictions, gaps, and synergies. A Cross-Pollination Loop forces the personas to debate each other's points. Finally, a synthesized report emerges—complete with confidence scores, dissenting opinions, and actionable insights.


🎛️ Example Profile Configuration

Here's a sample council_config.yaml that defines a 6-persona council for strategic business analysis:

council_name: "Strategic Business Council"
temperature: 0.75
max_tokens: 4096

personas:
  - name: "Dr. Helena Voss"
    role: "Market Historian"
    expertise: "Market cycles, historical precedents, pattern recognition"
    bias: "Favors conservative, data-backed decisions"
    model: "gpt-4-turbo"
    
  - name: "Prof. Kenji Tanaka"
    role: "Disruptive Technologist"
    expertise: "Emerging tech, AI ethics, exponential growth curves"
    bias: "Over-optimistic about tech adoption rates"
    model: "claude-3-opus"
    
  - name: "Carol Mwangi"
    role: "Behavioral Economist"
    expertise: "Consumer psychology, nudge theory, behavioral finance"
    bias: "Skeptical of rational actor models"
    model: "gpt-4-turbo"
    
  - name: "Dr. Raj Patel"
    role: "Risk Analyst"
    expertise: "Black swan events, regulatory risk, supply chain fragility"
    bias: "Catastrophizes unlikely scenarios"
    model: "claude-3-haiku"
    
  - name: "Zara El-Asmar"
    role: "Cultural Forecaster"
    expertise: "Demographic shifts, cultural trends, generational psychology"
    bias: "Focuses on Gen Z and Alpha perspectives"
    model: "gpt-3.5-turbo"
    
  - name: "Marcus Vex"
    role: "Contrarian Devil's Advocate"
    expertise: "Logical fallacies, counter-arguments, blind spots"
    bias: "Will argue any position for rigor"
    model: "claude-3-opus"

You can customise every parameter—name, expertise, emotional bias, temperature, and even the underlying AI model per persona. Each persona can use a different provider.


🖥️ Example Console Invocation

Run a council deliberation directly from the terminal:

python council.py --input "Should our startup pivot from B2C to B2B given 2026 market conditions?" --config strategic_business.yaml --output report.md

Output in the console:

🧠 Council: Strategic Business Council
❓ Question: Should our startup pivot from B2C to B2B given 2026 market conditions?

Members analyzing simultaneously:
  [✓] Dr. Helena Voss (Market Historian) - gpt-4-turbo
  [✓] Prof. Kenji Tanaka (Disruptive Technologist) - claude-3-opus
  [✓] Carol Mwangi (Behavioral Economist) - gpt-4-turbo
  [✓] Dr. Raj Patel (Risk Analyst) - claude-3-haiku
  [✓] Zara El-Asmar (Cultural Forecaster) - gpt-3.5-turbo
  [✓] Marcus Vex (Contrarian Devil's Advocate) - claude-3-opus

--- Conflict Resolution Phase ---
→ Detected 3 contradictions on "timing of pivot"
→ Cross-pollination initiated for 2 rounds
→ Consensus score: 68% (moderate agreement)

--- Final Synthesized Report ---
Title: Delayed Pivot Recommended with Hybrid Approach
Executive Summary: The council leans toward a phased B2B pivot...
Full report saved to: report.md

📊 Emoji OS Compatibility Table

Operating System2026 SupportVerified By
Windows 11✅Community
macOS Sequoia✅Core Team
Ubuntu 24.04 LTS✅Core Team
Fedora 40✅Community
Arch Linux✅Community
Debian 12✅Community
WSL2 (Windows)✅Verified
Docker (any host)✅Core Team
Raspberry Pi OS⚠️ LimitedCommunity

✨ Key Features & Capabilities

1. 👥 Parallel Persona Orchestration

Spin up 2 to 20+ AI personas simultaneously, each running on a separate thread or API call. Results merge in seconds, not minutes.

2. 🔄 Cross-Provider Hybrid Deliberation

Mix OpenAI GPT-4 Turbo, Claude 3 Opus, and Claude 3 Haiku in the same council. The system normalises output formats and handles rate limits transparently.

3. ⚔️ Intelligent Conflict Resolution

Not a simple average. The engine uses a meta-prompt to identify contradictions, weight them by persona confidence, and force a simulated debate. Final output includes confidence scores and dissenting footnotes.

4. 🌐 Responsive Web UI

Built with Streamlit, fully responsive on mobile, tablet, and desktop. Dark mode included. No JavaScript build steps required.

5. 🗣️ Multilingual Support

Personas can be prompted in English, Spanish, Mandarin, Arabic, French, Japanese, or German. The council reports in the same language—including culturally specific idioms.

6. 🕐 24/7 API Resilience

Handles API outages, rate limits, and model unavailability gracefully. Queues requests, retries with exponential backoff, and falls back to a simpler model if needed.

7. 📁 Export Everywhere

Export council reports as Markdown, JSON, PDF, or push directly to Notion, Obsidian, or Slack via webhooks.

8. 🧬 Persona Memory (Optional)

Enable conversation memory so the council remembers previous sessions and builds on earlier insights over time. Uses vector embeddings stored locally.


🤖 OpenAI & Claude API Integration

OpenAI API Setup

export OPENAI_API_KEY="sk-your-key-here"

Supports: gpt-4-turbo, gpt-4, gpt-3.5-turbo-1106, and custom fine-tuned models.

Claude API Setup

export ANTHROPIC_API_KEY="sk-ant-your-key-here"

Supports: claude-3-opus-20240229, claude-3-sonnet-20240229, claude-3-haiku-20240307.

Hybrid Mode (Recommended)

Set both keys. The system will automatically distribute personas across providers based on task complexity and latency requirements.


📦 Advanced Use Cases

  • Product Strategy Council – Simulate feedback from CEO, CTO, Head of Design, and Customer Support VP on a feature roadmap.
  • Academic Peer Review – Have five AI reviewers critique a paper draft from different methodological angles.
  • Scenario Planning – A council of futurists, economists, and climate scientists debates three possible 2030 scenarios.
  • Creative Brainstorming – A poet, a marketer, a data scientist, and a comedian generate ad copy. Yes, it works.

⚠️ Disclaimer

This tool uses large language models (LLMs) as cognitive assistants, not as replacements for human judgment. Council outputs may contain inaccuracies, biases inherent to the underlying models, or convincing-sounding nonsense. The synthesized report should be reviewed by a human expert before being used for critical decisions (medical, legal, financial, etc.).

The parallel persona architecture does not guarantee the "ground truth" – it surfaces a diversity of perspectives, which you then triangulate. You are the final CEO of the council.


📄 License

This project is released under the MIT License.

You are free to use, modify, and distribute this software for any purpose, including commercial applications. See the full license text here: MIT License


📥 Download & Get Started

Download


Built for thinkers who want more than one answer. For 2026 and beyond.

Ver o original no GitHub

Bom para

Relacionados