Multi-Agent Consortium

This example demonstrates how multiple specialized agents can collaborate using MAIF to produce a comprehensive artifact with full version history, content tracking, and forensic analysis capabilities.

Overview

The Multi-Agent Consortium example simulates a complex planning scenario: "How do I walk from California to Nepal in a meaningful way - where I have infinite ability to swim, and don't need to sleep".

Participating Agents

The consortium includes:

1. GeographyAgent: Analyzes terrain and routes.
2. CulturalAgent: Provides cultural insights and meaningful experiences.
3. LogisticsAgent: Handles practical considerations.
4. SafetyAgent: Assesses risks and safety measures.
5. CoordinatorAgent: Orchestrates the collaboration and synthesizes results.

Enhanced Features

• Version History Tracking: All content changes are tracked with full version history.
• Content Evolution: Iterative refinement of contributions based on feedback.
• Cross-Agent Dependencies: Management of dependencies between different agents' outputs.
• Forensic Analysis: Analysis of collaboration patterns and contribution history.
• Privacy & Security: Granular privacy controls and security verifications.
• Semantic Embeddings: Searchability through semantic understanding.

Running the Demo

To run the demo:

python examples/multi_agent_consortium_demo.py

Implementation Details

Each agent is implemented as a subclass of BaseAgent, which handles MAIF integration:

class BaseAgent:
    def __init__(self, agent_id, agent_type, specialization, shared_maif=None):
        # ... initialization ...
        self.maif = shared_maif if shared_maif is not None else create_maif(agent_id, enable_privacy=True)

    def contribute(self, query, context=None):
        # ... generate contribution ...
        self._store_contribution(contribution)
        return contribution

The agents collaborate by sharing a shared_maif instance or by exchanging data through the coordinator.