CLI Agent Example

Build an interactive CLI coding assistant using LocalBackend and the console toolset.

Quick Start

Bash

cd examples/local_cli
pip install pydantic-ai-backend[console]
export OPENAI_API_KEY=your-key
python cli_agent.py

# Include hidden files (e.g., .env) when searching
python cli_agent.py --include-hidden

Basic Implementation

Python

import asyncio
from dataclasses import dataclass
from pydantic_ai import Agent
from pydantic_ai_backends import LocalBackend, create_console_toolset, get_console_system_prompt

@dataclass
class Deps:
    backend: LocalBackend

# Create backend and toolset
backend = LocalBackend(root_dir=".", enable_execute=True)
toolset = create_console_toolset()

# Create agent
agent = Agent(
    "openai:gpt-4o-mini",
    system_prompt=f"""You are a helpful coding assistant.
{get_console_system_prompt()}
""",
    deps_type=Deps,
)
agent = agent.with_toolset(toolset)

async def main():
    deps = Deps(backend=backend)

    print("CLI Agent ready! Type 'quit' to exit.")

    while True:
        user_input = input("You: ").strip()

        if user_input.lower() in ("quit", "exit"):
            break

        result = await agent.run(user_input, deps=deps)
        print(f"\nAgent: {result.output}\n")

asyncio.run(main())

Features

The agent can:

• List files: "Show me what's in this directory"
• Read files: "Read the main.py file"
• Write files: "Create a hello.py that prints Hello World"
• Edit files: "Change the function name from foo to bar"
• Search: "Find all files containing 'TODO'" (use --include-hidden to search dotfiles)
• Execute: "Run the tests"

Example Session

Text Only

CLI Agent ready! Type 'quit' to exit.

You: Show me the project structure

Agent: Let me list the files in the current directory.

Contents of .:
  src/
  tests/
  README.md
  pyproject.toml

The project has:
- `src/` - Source code
- `tests/` - Test files
- `README.md` - Documentation
- `pyproject.toml` - Project config

You: Create a fibonacci function

Agent: I'll create a fibonacci.py file with the function.

Created fibonacci.py with a fibonacci function that calculates
the nth Fibonacci number using recursion with memoization.

You: Run it with n=10

Agent: Running the script...

Output: 55

The 10th Fibonacci number is 55.

Command Line Options

The full example supports these options:

Bash

# Specify working directory
python cli_agent.py --dir /path/to/project

# Use different model
python cli_agent.py --model anthropic:claude-3-haiku

# Disable shell execution
python cli_agent.py --no-execute

# Restrict file access
python cli_agent.py --restrict

# Include hidden files in searches
python cli_agent.py --include-hidden

# Single task mode
python cli_agent.py --task "Create a hello world script"

Full Example

See examples/local_cli/cli_agent.py for the complete implementation with:

• Command line argument parsing
• Interactive and single-task modes
• Security options (--restrict, --no-execute)
• Help command