This toolkit addresses the creation and orchestration of autonomous, evolving AI agent ecosystems for AI-First strategies. It targets developers building complex, adaptive multi-agent applications that require dynamic response, learning, and collaboration, offering a structured approach to managing agent lifecycles and interactions.

How It Works

EAT employs a goal-oriented orchestration model centered around a SystemAgent that acts as a central coordinator. It leverages a SmartLibrary for semantic search and versioning of agents and tools, enhanced by a dual embedding strategy for task-aware retrieval. Communication and discovery are managed via a dual SmartAgentBus (System and Data buses), enabling microservice-like interactions. Components can adapt and improve over time through an evolution mechanism, with optional multi-level human-in-the-loop review for governance and safety.

Quick Start & Requirements

• Install: Clone the repository, create a virtual environment, and install dependencies via pip install -r requirements.txt followed by pip install -e ..
• Prerequisites: Python 3.x, API keys for LLM providers (e.g., OpenAI).
• Demo: Run python examples/invoice_processing/architect_zero_comprehensive_demo.py.
• Configuration: Edit .env.example to .env and add API keys.
• Docs: Architecture overview available at docs/ARCHITECTURE.md.

Highlighted Details

• SystemAgent Orchestrator: A central ReAct agent managing component lifecycle (search, create, evolve) and task execution.
• SmartLibrary with Dual Embedding: Semantic search using Content Embeddings (raw content) and Applicability Embeddings (task relevance), enabling task-aware retrieval.
• SmartAgentBus (Dual Bus): Manages agent registration/discovery (System Bus) and capability-based requests (Data Bus).
• Intent Review System: Optional multi-level human-in-the-loop review for design, component selection, and execution plans.
• Component Evolution: Mechanisms to adapt agents/tools based on feedback and new requirements.

Maintenance & Community

• Contributors: Matias Molinas, Ismael Faro.
• Dependencies: BeeAI Framework, OpenAI Agents SDK, ChromaDB, LiteLLM.
• Roadmap: Focus on enhanced context/memory, smart caching, improved evolution, refined dual embedding, and expanded testing.

Licensing & Compatibility

• License: Apache License Version 2.0.
• Compatibility: Permissive license suitable for commercial use and integration with closed-source systems.

Limitations & Caveats

The docs/ARCHITECTURE.md and several technical reference sections are marked as "Action Required: Create/Update," indicating incomplete or outdated documentation for key architectural components and concepts. The primary quick start is a single comprehensive demo, with other use cases requiring deeper exploration of the examples/ directory.