A simple but powerful agent framework that utilizes LLMs for task planning and execution using the ReACT (Reasoning, Acting, and Observing) framework.

• Intelligent Planning: Breaks down complex tasks into smaller, manageable pieces
• ReACT Framework: Implements Reasoning → Acting → Observing cycle for iterative problem solving
• Tool Integration: Built-in tools for web search, browser automation, and more
• LLM Integration: Uses OpenAI's GPT models for reasoning and planning
• Extensible: Easy to add new tools and capabilities

1. Clone the repository
2. Install dependencies:

   pip install -r requirements.txt

3. Set up environment variables:

   cp env_example.txt .env
   # Edit .env with your OpenAI API key

# Install the package
pip install -e .

# Set up environment
cp env_example.txt .env
# Edit .env with your OpenAI API key

from miniagent import Agent

# Initialize the agent
agent = Agent()

# Give the agent a task
result = agent.run("Research the latest developments in AI and summarize the key findings")
print(result)

# Run a simple task
python -m miniagent.cli "Search for the current weather in New York"

# Run demos
python -m miniagent.cli --demo

# List available tools
python -m miniagent.cli --list-tools

1. Planner: Breaks down tasks into subtasks
2. ReACT Engine: Manages the reasoning, acting, and observing cycle
3. Tool Manager: Handles tool execution and results
4. LLM Interface: Communicates with language models

• Web Search: Search the internet for information
• Browser Automation: Navigate and interact with web pages
• File Operations: Read and write files
• Text Processing: Analyze and summarize text

agent = Agent()
result = agent.run("Find the current weather in New York")

agent = Agent()
result = agent.run("""
    Research the top 5 AI companies in 2024, 
    analyze their business models, and create a summary report
""")

agent = Agent()
result = agent.run("""
    Create a file with information about Python programming,
    then read and analyze the content
""")

agent = Agent()
result = agent.run("""
    Navigate to https://example.com,
    extract the main content, and summarize it
""")

The framework can be configured through environment variables:

• OPENAI_API_KEY: Your OpenAI API key (required)
• OPENAI_MODEL: Model to use (default: gpt-4)
• MAX_ITERATIONS: Maximum ReACT iterations (default: 10)
• TEMPERATURE: LLM temperature for creativity (default: 0.1)
• ENABLE_WEB_SEARCH: Enable web search tool (default: true)
• ENABLE_BROWSER: Enable browser automation (default: true)
• ENABLE_FILE_OPS: Enable file operations (default: true)
• BROWSER_HEADLESS: Run browser in headless mode (default: true)
• BROWSER_TIMEOUT: Browser timeout in seconds (default: 30)

from miniagent.tools import BaseTool

class MyCustomTool(BaseTool):
    name = "my_custom_tool"
    description = "Description of what this tool does"
    
    def execute(self, **kwargs):
        # Your tool logic here
        return {"result": "tool output"}

from miniagent.planner import BasePlanner

class MyCustomPlanner(BasePlanner):
    def plan(self, task: str) -> List[str]:
        # Your planning logic here
        return ["step1", "step2", "step3"]

Run the test suite to verify everything works:

python test_agent.py

To set up the development environment:

# Install in development mode
pip install -e .

# Run tests
python test_agent.py

# Run demos
python demo.py

MIT License