agent-orchestration-lib 0.1.0

A framework for building composable, observable, and maintainable agent workflows

Agent Orchestration Library

A framework for building composable, observable, and maintainable agent workflows with Python.

Features

• ExecutionContext: Dependency injection container for managing shared state and services
• EventEmitter: Event-driven architecture with pub/sub pattern for notifications
• AgentBlock: Base class for agent implementations with validation and error handling
• RetryStrategy: Configurable retry logic with exponential backoff and LLM fallbacks
• Flow: Multi-agent workflow orchestration with sequential, parallel, and conditional execution

Installation

pip install agent-orchestration-lib

Quick Start

from agent_lib import ExecutionContext, AgentBlock, EventEmitter
from pydantic import BaseModel

# Define your input/output models
class GreetingInput(BaseModel):
    name: str

class GreetingOutput(BaseModel):
    message: str

# Create an agent
class GreetingAgent(AgentBlock[GreetingInput, GreetingOutput]):
    def get_input_model(self):
        return GreetingInput

    def get_output_model(self):
        return GreetingOutput

    async def execute_impl(self, input_data: GreetingInput) -> GreetingOutput:
        return GreetingOutput(message=f"Hello, {input_data.name}!")

# Execute the agent
async def main():
    context = ExecutionContext()
    events = EventEmitter()

    agent = GreetingAgent("greeting", context, events)
    result = await agent.execute(GreetingInput(name="World"))
    print(result.message)  # Output: Hello, World!

Documentation

Full documentation available at: https://agent-orchestration-lib.readthedocs.io

Development

# Clone the repository
git clone https://github.com/yourusername/agent-orchestration-lib.git
cd agent-orchestration-lib

# Install development dependencies
pip install -e ".[dev]"

# Run tests
pytest

# Run type checking
mypy src/agent_lib

# Format code
black src tests
ruff check src tests

Architecture

This library was designed based on patterns extracted from production agent systems. See the architecture documentation for detailed design decisions.

Key design principles:

• Type Safety: Full Pydantic validation for inputs and outputs
• Observability: Event emission for all significant operations
• Composability: Agents can be combined into complex workflows
• Testability: Each component is independently testable
• Flexibility: Extensible through dependency injection

License

MIT License - see LICENSE file for details.

Contributing

Contributions welcome! Please read CONTRIBUTING.md for guidelines.

Project Status

Alpha - API may change. Not recommended for production use yet.

Current version: 0.1.0