hermes-ai 0.3.7

Hermes is a lightweight, powerful abstraction layer over LlamaIndex that simplifies building production-ready AI agents with essential utilities and multi-agent capabilities.

Hermes AI Agent Framework

Current Version: 0.3.7

🚀 Smart LlamaIndex Abstraction for AI Agents

🌟 Hermes is a lightweight, powerful abstraction layer over LlamaIndex that simplifies building production-ready AI agents with essential utilities and multi-agent capabilities. 🚀

✨ Features

• 🛠️ Automatic Tool Integration - Convert Python functions to tools effortlessly
• 🤖 Multi-Agent Orchestration - Coordinate multiple specialized agents
• 🧠 Enhanced Prompt Management - Built-in Chain of Thought and context awareness
• 💬 Smart Memory Management - Automatic chat history handling with configurable limits
• 🔌 Multi-Provider Support - OpenAI, Azure, Anthropic, and more
• 📊 Text Processing Utilities - Keyword extraction, formatting, and enhancement

🚀 Quick Start

pip install hermes-ai

import asyncio
from hermes.core import Agent

async def main():
    # Create a simple tool
    def get_weather(location: str) -> str:
        """Get weather information for a location."""
        return f"Weather in {location}: Sunny, 25°C"

    # Initialize your agent
    agent = Agent(
        provider="openai",
        model="gpt-4o-mini",
        name="WeatherAssistant",
        description="Helps with weather information",
        prompt="Provide accurate and helpful weather updates.",
        tools=[get_weather]
    )

    # Execute the agent
    response = await agent.execute(input_data="What's the weather in Tokyo?")
    print(response)

asyncio.run(main())

Output:

Using the weather tool, I can see that the weather in Tokyo is Sunny with a temperature of 25°C. It's a beautiful day!

💬 Simple Chat Example

import asyncio
from hermes.core import Agent

async def main():
    # Initialize agent
    agent = Agent(
        provider="openai",
        model="gpt-4o-mini"
    )
    
    # Ask a question
    response = await agent.execute(input_data="Qual é a capital do Brasil?")
    print(response)
    # Output: A capital do Brasil é Brasília. Ela foi oficialmente 
    # inaugurada em 21 de abril de 1960...

asyncio.run(main())

🌐 Web Interface Example

import asyncio
from hermes.core import Agent
from hermes.web import hermes_web

async def main():
    agent = Agent(
        provider="openai",
        model="gpt-4o-mini"
    )

    await hermes_web(port=8000, agent=agent)

if __name__ == "__main__":
    asyncio.run(main())

🔗 Web Interface Repository: Hermes-Web

🖼️ Screenshots Web Interface

Main chat interface with conversation history

Visualization of tools used by agents

Message input interface with validation

Loading indicator during processing

🏗️ Multi-Agent System

import asyncio
from hermes.core import Agent

async def main():
    # Define specialist tools
    def get_market_info(query: str) -> str:
        """Get current market conditions and trends."""
        return "Market is bullish today with major indices up 2%. USD at R$ 5.10."
    
    def calculate_investment(amount: float, period: int) -> str:
        """Calculate investment returns with recommendations."""
        expected_return = amount * (1 + 0.12) ** (period / 12)
        return f"Investment of R$ {amount:,.2f} for {period} months: Expected return R$ {expected_return:,.2f}"
    
    # Create specialized agents
    market_agent = Agent(
        provider="openai",
        model="gpt-4o-mini",
        name="MarketAnalyst",
        description="Expert in market analysis and quotes",
        tools=[get_market_info]
    )

    investment_agent = Agent(
        provider="openai",
        model="gpt-4o-mini",
        name="InvestmentAdvisor", 
        description="Expert in investment planning",
        tools=[calculate_investment]
    )

    # Coordinator agent that delegates to specialists
    coordinator = Agent(
        provider="openai",
        model="gpt-4o-mini",
        name="Coordinator",
        description="Routes questions to appropriate specialists",
        prompt="""Analyze user questions and delegate to the right expert:
        - Market questions → MarketAnalyst
        - Investment questions → InvestmentAdvisor
        - Complex questions → consult both""",
        tools=[market_agent, investment_agent]
    )
    
    # Execute coordinated task
    result = await coordinator.execute(
        input_data="Should I invest R$ 10000 for 12 months? What's the market like?"
    )
    print(result)

asyncio.run(main())

🔧 Core Components

Agent Configuration

Agent(
    provider="openai",           # LLM provider
    model="gpt-4o-mini",        # Model name
    name="Assistant",           # Agent identity
    description="Helpful AI",   # Agent purpose
    prompt="Behavior guidelines", # System prompt
    tools=[function1, function2], # Available tools
    temperature=0.7,            # Creativity control
    max_chat_history_length=20  # Memory management
)

Built-in Utilities

• Automatic tool conversion from Python functions
• Enhanced prompt formatting with current context
• Chat history management with configurable limits
• Input enhancement with keyword extraction
• Multi-provider LLM support

📁 Project Structure

hermes/
├── core.py          # Main Agent class
├── utils.py         # Text processing utilities
├── tools.py         # Tool creation helpers
└── providers.py     # LLM provider configurations

🎯 Use Cases

• Customer Support Agents
• Multi-Domain Expert Systems
• Research Assistants
• Data Analysis Tools
• Content Generation Systems

🔮 Roadmap

• Vector database integration
• Advanced memory backends
• Streaming responses
• Plugin system
• Web interface

💡 Contributing

We welcome contributions! Please see our Contributing Guide for details.

📄 License

MIT License - see LICENSE file for details.

---

Build intelligent agents faster with Hermes - The messenger of AI capabilities