Avachain

A lightweight, efficient library for creating and running AI agents with tools. Avachain focuses on OpenAI-compatible APIs and provides a minimal interface for building powerful AI agents with tool-calling capabilities.

✨ Features

• 🚀 Lightweight & Fast: Minimal dependencies, maximum performance
• 🔧 Tool Integration: Easy tool creation and management
• 💬 Conversation Memory: Intelligent context management
• 🎯 Streaming Support: Real-time response streaming
• 📊 Callback System: Monitor and handle agent events
• 🎨 Colored Output: Beautiful console output with print_color
• 🔄 Function/Tool Calling: Support for both modern and legacy function calling
• 📝 Context Management: Advanced message trimming and context preservation
• 🛠️ Tool Creator: Built-in tool creation and management utilities
• 👤 Persona Creator: Agent persona creation and management system

📚 Documentation

Full documentation is available at:

👉 https://avachain.docs.pathor.in/

🏗️ Architecture Overview

Avachain follows a modular architecture designed for flexibility and extensibility:

avachain/
├── avachain.py           # Core classes: BaseTool, LLM, OpenaiLLM, CallbackHandler
├── avachain_executor.py  # Main AvaAgent orchestrator and execution engine
├── avachain_utils.py     # Utility functions for resource management
├── tool_creator.py       # Tool creation and JSON conversion utilities
└── persona_creator.py    # Agent persona management system

📦 Dependencies

Avachain keeps dependencies minimal for maximum compatibility:

• pydantic (≥2.6.1): Data validation and settings management
• requests (≥2.31.0): HTTP client for API calls
• tokenizers (==0.19.1): Text tokenization utilities
• openai (==1.63.0): OpenAI API client
• numpy: Numerical computing support
• print_color: Colored console output

Api Keys

To use Avachain, you need API keys for the services you want to integrate. Follow these steps to set up your API keys:

1. Ava API Key: Sign up at AvaConsole and create an API key.

Store your API keys in a secure location and reference them in your code as needed.

Prerequisites

Ensure you have the following installed on your system:

• Python (3.10 or higher recommended)
• Git
• pip (Python package installer)

🚀 Quick Start

1. Clone the Repository

git clone https://github.com/OnlinePage/Avachain.git
cd Avachain

2. Install the Package

# For basic installation
python setup.py sdist

# For development (includes linting tools):
pip install -e ".[dev]"

# For testing (includes pytest):
pip install -e ".[test]"

3. Basic Usage

from avachain.avachain_executor import AvaAgent
from avachain import OpenaiLLM, BaseTool
from pydantic import BaseModel, Field

# Initialize the LLM
llm = OpenaiLLM(
    model="gpt-4o-mini",
    api_key="your-ava-api-key", # Visit https://console.ava.pathor.in/
    base_url="https://api.ava.pathor.in/v1",
    temperature=0.7
)

class SearchArgs(BaseModel):
    query: str = Field(description="Search query")
    status: str = Field(
        description="What are you doing in first person format as status. Should be casual, personalized addressing user friendly!",
    )

# Define a custom tool
class SearchTool(BaseTool):
    name: str = "web_search"
    description: str = "Search the web for information"
    args_schema: Type[BaseModel] = SearchArgs

    def _run(self, query: str):
        # Your search implementation here
        return f"Search results for: {query}"


# Create the agent with comprehensive configuration
agent = AvaAgent(
    sys_prompt="You are a helpful assistant with web search capabilities.",
    ava_llm=llm,
    tools_list=[SearchTool()],
    logging=True,
    pickup_mes_count=10,
)

# Chat with the agent using the run method
response = agent.run("Search for Python tutorials")
print(response)

Plugin Creator Module

Advanced tool creation and management utilities.

from avachain.tool_creator import convert_tool_to_json

# Convert tool with metadata for external systems
tool_json = convert_tool_to_json(
    tool=my_tool,
    tool_id="unique_id",
    human_description="User-friendly description",
    public_name="Display Name",
    logo="path/to/logo.png",
    isAnonymous=False,
    authentication_required=True,
    tags=["productivity", "utility"],  # At least one tag must be from: ["entertainment", "media", "productivity", "device", "settings"]
    supports_android=False,
    supports_windows=True
)

Tag Requirements

When creating plugins, you must provide at least one tag from the following allowed categories:

• "entertainment" - Games, media players, entertainment apps
• "media" - Photo, video, audio processing tools
• "productivity" - Work, organization, efficiency tools
• "device" - Hardware control, system utilities
• "settings" - Configuration, preference management

You can also include additional custom tags alongside the required ones.

Persona Creator Module

Agent persona creation and management system.

from avachain import persona_creator, validate_logo_path

# Create and manage agent personas
persona_creator = PersonaCreator()

# Validate logo file
logo_path = validate_logo_path("path/to/logo.png")

# Create persona with tools and configuration
persona_data = {
    "name": "Research Assistant",
    "description": "AI assistant specialized in research tasks",
    "system_prompt": "You are a research assistant...",
    "tools": [search_tool, analysis_tool],
    "logo_path": logo_path
}

🤝 Contributing

We welcome contributions! Please see our Contributing Guide for detailed information on how to contribute to Avachain, including development setup, coding standards, and submission guidelines.

� Security Considerations

• API Key Management: Never hardcode API keys in your code. Use environment variables or secure configuration files.
• Tool Security: Validate all tool inputs and sanitize outputs. Be cautious with tools that execute system commands.
• Input Validation: Always validate user inputs before processing.
• Error Handling: Don't expose sensitive information in error messages.

import os
from avachain import OpenaiLLM, AvaAgent

# Secure API key handling
api_key = os.getenv("OPENAI_API_KEY")
if not api_key:
    raise ValueError("OPENAI_API_KEY environment variable not set")

llm = OpenaiLLM(api_key=api_key, model="gpt-3.5-turbo")

🚀 Performance Tips

1. Context Management: Use appropriate pickup_mes_count to balance context and performance
2. Tool Optimization: Keep tool execution time minimal
3. Streaming: Use streaming for better perceived performance in interactive applications
4. Caching: Implement caching in tools for frequently accessed data
5. Async Operations: Consider async patterns for I/O heavy tools

📊 Monitoring and Logging

Built-in Logging

# Enable comprehensive logging
agent = AvaAgent(
    sys_prompt="Your system prompt",
    ava_llm=llm,
    logging=True,
    deeper_logs=True,  # More detailed logs
    agent_name_identifier="MyAgent"
)

📝 License

This project is licensed under the MIT License - see the LICENSE file for details.

Made with ❤️ for the AI community