q - CLI Tool for Querying LLMs

A command-line interface tool for querying Large Language Models (LLMs) with advanced features like context injection, command suggestions, and streaming output.

Features

• 🤖 Support for multiple LLM providers (OpenAI, etc.)
• 📝 Context injection from various sources:
  • Shell history (--hist)
  • Directory listings (--here)
  • File contents (--file)
• 💡 Command suggestions mode (--cmd)
• 🔄 Optional streaming output (--stream)
• 🎨 Beautiful progress display and colored output
• 💾 Response caching
• 🔁 Automatic retry with exponential backoff
• 🔒 Secure API key management

Quick Install

Requirements:

• Git
• Rust and Cargo (will be automatically installed if missing)

Using Homebrew

brew tap rfushimi/tap
brew install q

Manual Installation

Alternatively, you can install with a single command:

curl -sSL https://raw.githubusercontent.com/rfushimi/q/refs/heads/main/install.sh | bash

This will:

1. Install Rust if needed
2. Clone the repository
3. Build from source
4. Install the binary to ~/.bin/ (default)
5. Add the directory to your PATH if needed

You can customize the installation directory by setting the BIN_DIR environment variable:

BIN_DIR=/usr/local/bin curl -sSL https://raw.githubusercontent.com/rfushimi/q/refs/heads/main/install.sh | bash

Usage

Basic query:

q "What is Rust?"

With context:

# Include shell history context
q --hist "What did I do wrong in my last command?"

# Include current directory listing
q --here "What are the main source files?"

# Include file content
q --file src/main.rs "What does this code do?"

Command suggestions:

q --cmd "How do I find large files?"

Streaming output:

q --stream "Explain quantum computing"

Configuration

API keys are stored in configuration files:

# Set OpenAI API key
q set-key openai YOUR_API_KEY

Options

Options:
  -H, --hist           Include shell history context
  -P, --provider       Select LLM provider [default: gemini]
  -M, --model         Select model name (e.g., gemini-2.0-flash, gpt-3.5-turbo)
  -D, --here           Include current directory listing
  -F, --file <FILE>    Include file content
  -C, --cmd            Get command suggestions
      --stream         Enable streaming output
      --no-cache      Disable response caching
      --retries <N>    Maximum retry attempts [default: 3]
      --debug          Show debug information
  -h, --help          Print help
  -V, --version       Print version

Development

Requirements:

• Rust 1.70 or later
• OpenAI API key (or other supported provider)

Build from source:

git clone https://github.com/rfushimi/q.git
cd q
cargo build --release

# install
cargo install --path . q

Run tests:

cargo test

License

MIT License - see LICENSE for details.