My First Agent

An AI-powered code agent built with Python and Google's Gemini 2.5 Flash. This agent can understand natural language instructions and autonomously perform file operations, read code, and execute Python scripts.

Features

• File Exploration - List files and directories with metadata
• File Reading - Read and analyze file contents
• Python Execution - Run Python scripts and capture output
• File Writing - Create or modify files
• Agentic Loop - Automatically chains multiple operations to complete complex tasks

Requirements

• Python >= 3.13
• Google Gemini API key

Installation

1. Clone the repository:

git clone https://github.com/sanchezcodes/my-first-agent.git
cd my-first-agent

2. Install dependencies using uv:

uv sync

3. Create a .env file with your Gemini API key:

GEMINI_API_KEY='your-api-key-here'

Usage

Run the agent with a natural language instruction:

python main.py "your instruction here"

Examples

# List all files in the working directory
python main.py "List all files in the working directory"

# Run the calculator
python main.py "Run the calculator with expression 3 + 5"

# Create a new file
python main.py "Create a file called hello.txt with the content 'Hello World!'"

# Run tests and get results
python main.py "Run all tests in the calculator and tell me the results"

Verbose Mode

Add --verbose to see detailed information about token usage and function calls:

python main.py "your instruction" --verbose

Project Structure

my-first-agent/
├── main.py              # Agent entry point
├── call_function.py     # Function routing and execution
├── config.py            # Configuration constants
├── prompts.py           # System prompts for the LLM
├── functions/           # Tool implementations
│   ├── get_files_info.py
│   ├── get_file_content.py
│   ├── run_python_file.py
│   └── write_file.py
└── calculator/          # Sample application for testing
    ├── main.py
    ├── tests.py
    └── pkg/
        ├── calculator.py
        └── render.py

Available Tools

Tool	Description
get_files_info	Lists files and directories with size and type information
get_file_content	Reads the content of a file (up to 10,000 characters)
run_python_file	Executes a Python script with optional arguments
write_file	Creates or overwrites a file with specified content

Configuration

Edit config.py to customize:

MAX_CHARS = 10000          # Max characters to read from files
WORKING_DIR = "./calculator"  # Restricted working directory
MAX_ITERS = 20             # Max agent loop iterations

Security

The agent includes several safety measures:

• Path validation to prevent directory traversal
• Working directory sandboxing
• Process execution timeout (30 seconds)
• File size limits

License

MIT