AI_Bot (Tool-Using Coding Agent)

AI_Bot is a command-line AI coding assistant built in Python that can interact with a local project through a controlled set of tools. It supports both one-shot prompts and an interactive REPL mode, allowing the assistant to inspect files, write changes, execute Python scripts, and run automated tests.

This project goes beyond a basic chatbot by implementing a functional agent loop that can decide when to call tools, use tool output as context, and continue until a final response is produced.

This project is intended to be run on Linux/WSL (recommended) for the most consistent behavior and best compatibility.

Key Features (Beyond the Base Boot.dev Project)

Interactive REPL Mode

AI_Bot includes a REPL chat mode that keeps conversation context across multiple turns.

• Start the bot once and interact continuously
• Exit cleanly when finished

Tool Calling (Local Function Execution)

The assistant can call local tools to perform real operations inside the project:

• List files and directories
• Read file contents
• Write or overwrite files
• Execute Python files with optional arguments
• Run pytest test suites

Multi-Step Agent Loop

Instead of responding once and stopping, AI_Bot can:

1. Evaluate the request
2. Call the appropriate tool(s)
3. Interpret the results
4. Continue tool usage until the task is completed (or a safety limit is reached)

Test Runner Integration (run_tests)

AI_Bot includes a run_tests tool that executes pytest and returns:

• Exit code
• Command used
• STDOUT / STDERR output

It also supports selecting which tests to run by path (example: tests vs calculator/tests).

Working Directory Injection for Safety

Tool calls automatically receive a controlled working directory to reduce risk of unsafe path access. This keeps file operations scoped to expected locations.

Project Structure

AI_Bot/
├─ .venv/
├─ calculator/
│  ├─ pkg/
│  │  ├─ calculator.py
│  │  ├─ morelorem.txt
│  │  └─ render.py
│  ├─ tests/
│  │  ├─ test_calculator.py
│  │  └─ test_smoke.py
│  ├─ lorem.txt
│  ├─ main.py
│  ├─ main.txt
│  └─ tests.py
│
├─ functions/
│  ├─ get_file_content.py
│  ├─ get_files_info.py
│  ├─ run_python_file.py
│  ├─ run_tests.py
│  └─ write_file.py
│
├─ tests/
│  ├─ test_call_functions.py
│  ├─ test_get_file_content.py
│  ├─ test_get_files_info.py
│  ├─ test_loop_prompt.py
│  ├─ test_prompts.py
│  ├─ test_run_python_file.py
│  ├─ test_smoke.py
│  └─ test_write_file.py
│
├─ .env
├─ .gitignore
├─ .python-version
├─ call_function.py
├─ config.py
├─ main.py
├─ prompts.py
├─ pyproject.toml
├─ README.md
└─ uv.lock

Requirements

• Python 3.12+
• uv
• Linux or WSL (recommended)
• OpenAI API key

Setup

1. Create a .env file

In the project root:

OPENAI_API_KEY=your_key_here

2. Install dependencies

uv sync

Usage

One-Shot Mode

Run a single prompt and exit:

uv run main.py "list the files in the root"

REPL Mode

Start an interactive session:

uv run main.py

Example REPL prompts:

• list the files in the root
• read the contents of main.py
• write 'hello' to main.txt
• run tests
• run tests in calculator/tests

Notes

This project demonstrates a practical tool-using agent workflow in a controlled local environment. The main focus is on building a functional coding agent with multi-step execution, test running, and safe project interaction rather than a simple chat-only bot.