abi-core-ai 0.1.0b107

Agent-Based Infrastructure Core - Runtime and CLI

ABI-Core 🤖

ABI-Core-AI — The foundation for building Agent-Based Infrastructure (ABI) — a new architectural paradigm where intelligent agents collaborate through semantic context, policy-driven governance, and modular orchestration.

Agent-Based Infrastructure Core — A comprehensive framework for building, deploying, and managing AI agent systems with semantic layers, orchestration, and security policies.

---

🧭 Core Philosophy

ABI-Core is built on three fundamental principles:

1. Semantic Interoperability — Agents must share meaning, not just data.
2. Distributed Intelligence — No single model owns the truth; collaboration is the substrate.
3. Governed Autonomy — Security and compliance must evolve as fast as intelligence itself.

⚠️ Beta Release: This is a beta version. APIs may change and some features are experimental.

---

🚀 Quick Start

Installation

pip install abi-core-ai

Create Your First Project

# Create a new ABI project with semantic layer
abi-core create project my-ai-system --with-semantic-layer

# Navigate to your project
cd my-ai-system

# Provision models (automatically starts services and downloads models)
abi-core provision-models

# Create an agent
abi-core add agent my-agent --description "My first AI agent"

# Create an agent card for semantic discovery
abi-core add agent-card my-agent --description "General purpose AI assistant" --url http://localhost:8000

# Run your project
abi-core run

📖 Need help? Check out our complete documentation with guides, examples, and API reference.

---

🔧 Model Serving Options

ABI-Core supports two model serving strategies for Ollama:

Centralized (Recommended for Production)

A single shared Ollama service serves all agents:

• ✅ Lower resource usage — One Ollama instance for all agents
• ✅ Easier model management — Centralized model updates
• ✅ Faster agent startup — No need to start individual Ollama instances
• ✅ Centralized caching — Shared model cache across agents

abi-core create project my-app --model-serving centralized

Distributed (Default)

Each agent has its own Ollama instance:

• ✅ Complete isolation — Each agent has independent models
• ✅ Independent versions — Different model versions per agent
• ✅ Development friendly — Easy to test different configurations
• ⚠️ Higher resource usage — Multiple Ollama instances

abi-core create project my-app --model-serving distributed
# or simply (distributed is default)
abi-core create project my-app

Note: Guardian service always maintains its own Ollama instance for security isolation, regardless of the chosen mode.

---

🎯 What is ABI-Core?

ABI-Core-AI is a production-ready framework for building Agent-Based Infrastructure systems that combine:

• 🤖 AI Agents — LangChain-powered agents with A2A (Agent-to-Agent) communication
• 🧠 Semantic Layer — Vector embeddings and distributed knowledge management
• 🔒 Security — OPA-based policy enforcement and access control
• 🌐 Web Interfaces — FastAPI-based REST APIs and real-time dashboards
• 📦 Containerization — Docker-ready deployments with orchestration

---

🏗️ Architecture

┌─────────────────┐    ┌─────────────────┐    ┌─────────────────┐
│   AI Agents     │◄──►│ Semantic Layer  │◄──►│   Guardian      │
│                 │    │                 │    │   Security      │
│ • LangChain     │    │ • Vector DB     │    │ • OPA Policies  │
│ • A2A Protocol  │    │ • Embeddings    │    │ • Access Control│
│ • Custom Logic  │    │ • Knowledge     │    │ • Monitoring    │
└─────────────────┘    └─────────────────┘    └─────────────────┘
         │                       │                       │
         └───────────────────────┼───────────────────────┘
                                 │
                    ┌─────────────────┐
                    │  Web Interface  │
                    │                 │
                    │ • FastAPI       │
                    │ • Real-time UI  │
                    │ • Monitoring    │
                    └─────────────────┘

---

📋 Features

🤖 Agent System

• Multi-Agent Architecture — Create specialized agents for different tasks
• A2A Communication — Agents can communicate and collaborate
• LangChain Integration — Leverage the full LangChain ecosystem
• Custom Tools — Extend agents with domain-specific capabilities

🧠 Semantic Layer

• Agent Discovery — MCP-based agent finding and routing
• Vector Storage — Weaviate-based semantic search (automatically configured)
• Agent Cards — Structured agent metadata and capabilities
• Access Validation — OPA-integrated security for semantic access
• Embedding Mesh — Distributed embedding computation and caching
• Context Awareness — Agents understand semantic relationships
• Auto-Configuration — Weaviate vector database included automatically

🔒 Security & Governance

• Policy Engine — Open Policy Agent (OPA) integration
• Access Control — Fine-grained permissions and roles
• Audit Logging — Complete activity tracking
• Compliance — Built-in security best practices

🌐 Web & APIs

• REST APIs — FastAPI-based service endpoints
• Real-time Updates — WebSocket support for live data
• Admin Dashboard — Monitor and manage your agent system
• Custom UIs — Build domain-specific interfaces

---

🛠️ CLI Commands

Project Management

# Create new projects with optional services and model serving strategy
abi-core create project <name> [--domain <domain>] [--with-semantic-layer] [--with-guardian] [--model-serving centralized|distributed]
abi-core provision-models          # Download and configure LLM models (auto-starts services)
abi-core status                    # Check project status
abi-core run                       # Start all services
abi-core info                      # Show project information

Agent Development

# Create and manage agents
abi-core add agent <name> [--description <desc>] [--model <model>] [--with-web-interface]
abi-core remove agent <name>       # Remove an agent
abi-core info agents               # List all agents

Services Management

# Add services to existing projects
abi-core add service semantic-layer [--name <name>] [--domain <domain>]
abi-core add service guardian [--name <name>] [--domain <domain>]
abi-core add service guardian-native [--name <name>] [--domain <domain>]

# Quick service shortcuts
abi-core add semantic-layer [--domain <domain>]    # Add semantic layer directly
abi-core remove service <name>                     # Remove any service

Agent Cards & Semantic Layer

# Manage agent cards for semantic discovery
abi-core add agent-card <name> [--description <desc>] [--model <model>] [--url <url>] [--tasks <tasks>]
abi-core add policies <name> [--domain <domain>]   # Add security policies

Examples

# Create a finance project with centralized model serving (recommended for production)
abi-core create project fintech-ai --domain finance --with-semantic-layer --with-guardian --model-serving centralized
cd fintech-ai

# Provision models (starts Ollama and downloads qwen2.5:3b + embeddings)
abi-core provision-models

# Add a specialized trading agent (automatically uses centralized Ollama)
abi-core add agent trader --description "AI trading assistant" --model qwen2.5:3b

# Create agent card for semantic discovery
abi-core add agent-card trader --description "Execute trading operations" --url http://localhost:8001 --tasks "trade,analyze,risk-assessment"

# Add semantic layer to existing project (Weaviate included automatically)
abi-core add semantic-layer --domain finance

# Create a development project with distributed model serving (each agent has own Ollama)
abi-core create project dev-project --model-serving distributed
cd dev-project

# Provision models (starts all agents with their Ollama instances + main Ollama for embeddings)
abi-core provision-models

# Remove components when needed
abi-core remove service semantic_layer
abi-core remove agent trader

---

📁 Project Structure

When you create a new project, you get:

my-project/
├── agents/                 # Your AI agents
│   └── my-agent/
│       ├── agent.py        # Agent implementation
│       ├── main.py         # Entry point
│       ├── models.py       # Data models
│       └── agent_cards/    # Agent cards for semantic discovery
├── services/               # Supporting services
│   ├── semantic_layer/     # AI agent discovery & routing
│   │   └── layer/
│   │       ├── mcp_server/ # MCP server for agent communication
│   │       └── embedding_mesh/ # Vector embeddings & search
│   └── guardian/           # Security & policy enforcement
├── agent_cards/            # Global agent cards directory
├── config/                 # Configuration files
│   └── config.py
├── compose.yaml            # Container orchestration
├── .abi/                   # ABI project metadata
│   └── runtime.yaml
└── README.md               # Project documentation

---

🔧 Configuration

ABI-Core uses environment variables and YAML configuration files:

# .abi/runtime.yaml
agents:
  my-agent:
    model: "qwen2.5:3b"
    port: 8000
    
semantic_layer:
  provider: "weaviate"
  host: "localhost:8080"
  
security:
  opa_enabled: true
  policies_path: "./policies"

---

🚀 Deployment

Docker (Recommended)

docker-compose up --build
docker-compose up --scale my-agent=3

Kubernetes

abi-core-ai deploy kubernetes
kubectl apply -f ./k8s/

---

🧪 Examples

Simple Agent

from abi_core_ai.agent.agent import AbiAgent
from abi_core_ai.common.utils import abi_logging

class MyAgent(AbiAgent):
    def __init__(self):
        super().__init__(
            agent_name='my-agent',
            description='A helpful AI assistant'
        )
    
    async def stream(self, query: str, context_id: str, task_id: str):
        abi_logging(f"Processing: {query}")
        response = await self.llm.ainvoke(query)
        yield {
            'content': response.content,
            'response_type': 'text',
            'is_task_completed': True
        }

Agent Communication

await self.send_message(
    target_agent="agent-b",
    message="Process this data",
    data={"items": [1, 2, 3]}
)

---

📚 Documentation

📖 Full Documentation: https://abi-core.readthedocs.io

• Getting Started - Installation and quick start
• Quick Start Guide - Get running in 5 minutes
• Models Guide - Model selection and provisioning
• FAQ - Frequently asked questions
• Architecture - System design and concepts

---

🤝 Contributing

We welcome contributions! This is a beta release, so your feedback is especially valuable.

Development Setup

git clone https://github.com/Joselo-zn/abi-core-ai
cd abi-core-ai
uv sync --dev

Running Tests

uv run pytest

---

📄 License

Apache 2.0 License — see LICENSE for details.

---

🆘 Support

• Issues — GitHub Issues
• Discussions — GitHub Discussions
• Email — [e-mail]

---

🗺️ Roadmap

Milestone	Description	Status
v0.2.0	Enhanced agent orchestration	🔜 In progress
v0.3.0	Advanced semantic search	🧠 Planned
v0.4.0	Multi-cloud deployment	🧩 Planned
v1.0.0	Production-ready stable release	🏁 Target Q3 2026

---

Built with ❤️ by José Luis Martínez
Creator of ABI (Agent-Based Infrastructure) — redefining how intelligent systems interconnect.

✨ From Curiosity to Creation: A Personal Note

I first saw a computer in 1995. My dad had received a Windows 3.11 machine as payment for a job. I was fascinated. At the time, I wanted to study robotics — but when I touched that machine, everything changed.

I didn't understand what the Internet was, and I had no idea where to go… but even in that confusion, I felt something big. When I wrote my first Visual C++ program in 1999, I felt like a hacker. When I built my first web page, full of GIFs, I was flying.

Nobody taught me. I just read manuals. And now, years later, that journey continues — not just as a coder, but as the creator of ABI. This is for the kids like me, then and now.