AnonCodexCli

An interactive CLI-based AI coding assistant inspired by Cursor.

Features

• Multiple operation modes:
  • Interactive: Ask questions and get responses
  • Autonomous: Let the AI plan and execute tasks
  • Manual: Control each step of the process
• Support for various LLMs:
  • Google Gemini
  • OpenAI (GPT models)
  • Anthropic Claude
  • Local LLMs via Ollama or LM Studio
• File operations and codebase understanding
• Terminal command execution
• Code search capabilities
• Interactive UI with rich terminal formatting

Installation

See INSTALL.md for detailed installation instructions.

1. Clone this repository:

git clone https://github.com/AnonAmit/AnonCodexCli.git
cd AnonCodexCli

2. Install dependencies:

pip install -r requirements.txt

3. Create a .env file with your API keys (see .env.example):

cp .env.example .env

4. Edit the .env file with your actual API keys.

Usage

Run the CLI:

python cli.py

This will present you with mode selection:

1. Interactive
2. Autonomous
3. Manual

Interactive Mode

In interactive mode, you can:

• Ask coding questions
• Request file edits
• Run terminal commands through the assistant

Example:

> Fix the bug in app.py line 56

Autonomous Mode

In autonomous mode, the assistant will:

1. Plan the execution steps
2. Execute each step
3. Test the results
4. Report progress

You can toggle whether to show code in the terminal.

Manual Mode

In manual mode, you control each step the assistant takes.

Supported LLM Models

AnonCodexCli supports multiple LLM providers:

Google Gemini

• gemini-pro

OpenAI

• gpt-3.5-turbo
• gpt-4
• gpt-4o

Anthropic Claude

• claude-3-opus
• claude-3-sonnet

Local LLMs

• Ollama: llama3, mistral, mixtral
• LM Studio: custom

Model Selection

When running the CLI, you'll be prompted to select a model. You can also specify a model using the --model flag:

python cli.py --model gemini-pro

Testing Models

You can test all supported models to verify API connections:

python test_models.py

Configuration

Configure AnonCodexCli through the .env file:

• DEFAULT_MODEL: Default LLM to use
• DEFAULT_MODE: Default operation mode
• MAX_TOKENS: Maximum tokens for LLM responses
• HISTORY_PATH: Path to store interaction history
• DEBUG: Enable debug logging

Command Line Arguments

usage: cli.py [-h] [--mode {interactive,autonomous,manual}] [--model MODEL] [--debug] [--query QUERY]

AnonCodexCli - A CLI-based AI coding assistant

options:
  -h, --help            show this help message and exit
  --mode {interactive,autonomous,manual}, -m {interactive,autonomous,manual}
                        Operation mode (default: interactive)
  --model MODEL         LLM model to use (default: gemini-pro)
  --debug, -d           Enable debug mode
  --query QUERY, -q QUERY
                        Query to process (if provided, runs in non-interactive mode and exits after completion)

TOOL UI SCREENSHOTS

License

MIT License