AI Coding Agent

This repository contains an AI-powered coding agent that leverages Google's Gemini LLM to answer questions, plan function calls, and interact with a Python codebase. The agent can list files, read file contents, execute Python scripts, and write files, all within a secure working directory.

Features

• Natural Language Interface: Ask coding questions or make requests in plain English.
• Function Calling: The agent can plan and execute function calls such as reading files, writing files, and running Python scripts.
• Extensible Functions: Modular function system (see the functions/ directory) for file operations and code execution.
• Calculator Example: Includes a sample calculator/ project for demonstration and testing.
• Verbose Mode: Optional verbose output for debugging and transparency.

Key Directories

• functions/: Contains modules for file operations and code execution:
  • get_files_info.py
  • get_file_content.py
  • write_file.py
  • run_python_file.py
  • search_in_file.py
• calculator/: Example Python project with its own main.py and tests.py.

Example working Calculator Bug Fix (Operator Precedence)

The calculator app was initially producing incorrect results due to a bug in the operator precedence handling. Specifically, expressions like "3 + 5 * 2" were evaluated as 16 instead of the correct result 13. The issue was caused by the + operator being evaluated before the * operator.

Example Usage

Here’s how the calculator now works:

1. Basic Calculation (3 + 5)

   (venv) levyashvin@Yashvin:~/bootdev/ai-agent$ python calculator/main.py "3 + 5"

   Output:

   ┌─────────┐
   │  3 + 5  │
   │         │
   │  =      │
   │         │
   │  8      │
   └─────────┘
   

2. Incorrect Calculation (Before Fix - 3 + 5 * 2)

   (venv) levyashvin@Yashvin:~/bootdev/ai-agent$ python calculator/main.py "3 + 5 * 2"

   Output (Incorrect):

   ┌─────────────┐
   │  3 + 5 * 2  │
   │             │
   │  =          │
   │             │
   │  16         │
   └─────────────┘
   

3. Asking Gemini to fix the issue

   (venv) levyashvin@Yashvin:~/bootdev/ai-agent$ python main.py "can you fix the bug in the calculator it is giving 3 + 5 * 2 = 16 when it should be 13"

   Output:

   - Calling function: get_files_info
   - Calling function: get_file_content
   - Calling function: get_file_content
   - Calling function: write_file
   - Calling function: run_python_file
   <class 'subprocess.CompletedProcess'>
   CompletedProcess(args=['python', '/home/levyashvin/bootdev/ai-agent/calculator/tests.py'], returncode=0, stdout='', stderr='.........\n----------------------------------------------------------------------\nRan 9 tests in 0.001s\n\nOK\n')
   Great! The tests passed, which means the issue with operator precedence has been resolved, and the calculator should now correctly evaluate expressions like "3 + 5 * 2".
   

4. Correct Calculation After the Fix

   Now, when running the same expression:

   (venv) levyashvin@Yashvin:~/bootdev/ai-agent$ python calculator/main.py "3 + 5 * 2"

   Output (Correct):

   ┌─────────────┐
   │  3 + 5 * 2  │
   │             │
   │  =          │
   │             │
   │  13         │
   └─────────────┘
   

Summary:

• The bug in the operator precedence was fixed by adjusting the calculation logic in the calculator app.
• The multiplication operator (*) is now correctly evaluated before addition (+), resulting in the correct output for expressions like 3 + 5 * 2.
• Tests were run and passed successfully to verify that the bug was fixed.

Setup

1. Clone the repository:

   git clone https://github.com/levyashvin/ai-agent/
   cd ai-agent

2. Install dependencies:

   pip install -r requirements.txt

3. Set up your environment:

   • Create a .env file in the root directory with your Gemini API key:

     GEMINI_API_KEY=your_api_key_here
     

Usage

Run the agent with a natural language prompt:

python main.py "How does the calculator render an output?"

• Add --verbose to enable verbose output:

  python main.py "List all files in the calculator directory" --verbose

Configuration

Edit config.py to adjust:

• Model version (MODEL)
• System prompt (SYSPROMPT)
• Working directory (WORKDIR)
• Max iterations, timeout, and verbosity

Testing

Run the included tests for function modules:

python tests.py

Extending

To add new capabilities, implement a new function in the functions/ directory and register its schema in call_function.py.