ezyagent 0.0.1

agent framework designed to make AI agent development simple

EasyAgent: A Modern, Simple, and Powerful Agent Framework

Overview

EasyAgent is a next-generation agent framework designed to make AI agent development simple, observable, and reliable. By learning from the limitations of existing frameworks, we've created a solution that prioritizes developer experience while providing enterprise-grade features.

🌟 Key Features

Simple Yet Powerful

from ezyagent import Agent

# Create an agent in one line
agent = Agent("gpt-4")

# Start chatting
response = await agent.chat("Explain quantum computing")


# Add tools easily
@agent.tool
async def search(query: str) -> str:
    """Search the web."""
    return await web_search(query)

First-Class Async Support

# Stream responses
async with Agent() as agent:
    async for message in agent.stream_chat("Write a long story"):
        print(message)
        
# Parallel operations
async def process_queries(queries: List[str]):
    async with Agent() as agent:
        tasks = [agent.chat(q) for q in queries]
        responses = await asyncio.gather(*tasks)

Advanced Logging & Observability

# Comprehensive logging setup
agent.logger.configure(
    format="json",
    outputs={
        "console": {"level": "INFO"},
        "file": {
            "path": "agent.log",
            "level": "DEBUG"
        },
        "cloudwatch": {
            "group": "agents",
            "stream": "production"
        }
    },
    metrics=["tokens", "latency", "costs"],
    trace_requests=True
)

# Access logs and metrics
print(agent.logger.get_metrics())
print(agent.logger.get_recent_traces())

Robust Error Handling

try:
    response = await agent.chat("Complex query")
except AgentError as e:
    print(f"Error Type: {e.error_type}")
    print(f"Provider Error: {e.provider_error}")
    print(f"Context: {e.context}")
    print(f"How to fix: {e.remediation}")
    print(f"Debug trace: {e.debug_info}")

Intelligent State Management

# Built-in memory and state management
agent.memory.save_state("user_preferences", preferences)
agent.memory.add_context("User is a developer")

# Access conversation history
history = agent.memory.get_chat_history()
context = agent.memory.get_relevant_context("query")

# Persistent storage
await agent.memory.save_to_disk("agent_state.json")
await agent.memory.load_from_disk("agent_state.json")

Universal Provider Support

# Easy provider switching
agent = Agent(provider="openai")
agent = Agent(provider="anthropic")
agent = Agent(provider="ollama")

# Multiple providers with fallback
agent = Agent(
    providers=["anthropic", "openai"],
    fallback_strategy="sequential"
)

# Custom provider configuration
agent = Agent(
    provider="openai",
    config={
        "max_retries": 3,
        "timeout": 30,
        "rate_limit": 100
    }
)

Why EasyAgent?

Problems with Existing Frameworks

1. Langchain

• ❌ Complex setup and steep learning curve
• ❌ Confusing abstractions
• ❌ Poor error handling
• ❌ Limited async support
• ✅ Extensive tool ecosystem
• ✅ Good documentation

2. AutoGen

• ❌ Complex configuration
• ❌ Limited logging
• ❌ Difficult debugging
• ✅ Good multi-agent support
• ✅ Built-in caching

3. Pydantic-AI

• ❌ Limited provider support
• ❌ Basic logging
• ❌ No state management
• ✅ Strong type validation
• ✅ Clean data structures

4. LlamaIndex

• ❌ Complex for simple uses
• ❌ Heavy resource usage
• ❌ Confusing documentation
• ✅ Great RAG support
• ✅ Good data ingestion

5. PhiData

• ❌ Limited features
• ❌ Basic logging
• ❌ Limited providers
• ✅ Simple API
• ✅ Clean implementation

EasyAgent's Solutions

1. Development Experience

• One-line setup
• Clear, concise API
• Comprehensive documentation
• Type hints everywhere
• Informative error messages
• IDE autocomplete support

2. Observability

• Structured logging
• Request tracing
• Cost tracking
• Performance metrics
• Debug mode
• Custom metric support

3. Reliability

• Automatic retries
• Smart rate limiting
• Provider fallbacks
• Error recovery strategies
• Validation checks

4. Flexibility

• Easy extension
• Custom tools
• Provider agnostic
• State management
• Memory systems
• Custom implementations

5. Performance

• Async by default
• Efficient resource usage
• Built-in caching
• Streaming support
• Parallel operations

Architecture

easyagent/
├── core/
│   ├── agent.py          # Base agent classes
│   ├── memory.py         # State management
│   ├── tools.py          # Tool management
│   └── providers/        # LLM providers
├── logging/
│   ├── logger.py         # Logging core
│   ├── formatters.py     # Log formatters
│   ├── handlers.py       # Output handlers
│   └── monitors.py       # Metrics
├── utils/
│   ├── errors.py         # Error handling
│   ├── validation.py     # Input validation
│   └── helpers.py        # Utilities
└── examples/             # Usage examples

Installation

pip install easyagent

Quick Start

from ezyagent import Agent

# Create an agent
agent = Agent("gpt-4")

# Enable logging
agent.logger.configure(format="json", outputs=["console"])


# Add a tool
@agent.tool
async def search(query: str) -> str:
    """Search the web."""
    return await web_search(query)


# Chat with the agent
async def main():
    response = await agent.chat("Find recent news about AI")
    print(response)


# Run
asyncio.run(main())

Documentation

Full documentation is available at docs.easyagent.dev

License

MIT License - feel free to use in your own projects!

Contributing

We welcome contributions! See CONTRIBUTING.md for guidelines.