llamatex

Convert natural language descriptions to LaTeX equations in your terminal, in under a second!

Features

llamatex offers two primary modes of operation:

1. Single Instruction Mode: Generate LaTeX for a single equation description.
2. Clipboard Mode: Send LaTeX code from highlighted text into your clipboard.
3. Markdown File Editing Mode: Automatically replace natural language descriptions with LaTeX equations in a markdown file.

For equations that can be rendered in a single line, smoltex opens a new window displaying the rendered equation.

Installation

pip install llamatex

For local generation support:

pip install llamatex[local]

Alternatively, you can build from source using poetry.

poetry install

Usage

Single Instruction mode

llamatex render -p "equation for the area of a circle"

Output:

Latex string: A = \pi r^2

>> completion time: 856.4 ms

Clipboard mode

llamatex uses pynput to listen to your inputs and control the keyboard. To allow clipboard mode, allow your terminal/IDE to be added to Assessibility in System Settings.

Run llamatex in clipboard mode using the following command:

llamatex run

Markdown File Editing Mode

1. Create a markdown file (e.g., equations.md) with placeholders:

# Important Equations

1. Area of a circle:
{"area of a circle"}

2. Einstein's mass-energy equivalence:
{"einstein's famous equation"}

2. Run llmatex:

llamatex render -f equations.md

3. The file will be updated with LaTeX equations:

# Important Equations

1. Area of a circle:
$A = \pi r^2$

2. Einstein's mass-energy equivalence:
$E = mc^2$

Use the hotkey Cmd+Shift+L (on macOS) to process the selected text. Press esc to exit the program.

Configuration

API Setup

1. Get a free API key from Groq console.
2. Set the environment variable

export GROQ_API_KEY=your-api-key-here

Available models:

• llama3: llama3-8b, llama3-70b, llama3.1-8b, llama3.1-70b
• gemma: gemma-7b, gemma2-9b
• mixtral: mixtral-8x7b
• Phi3 (local only): phi-3-mini

Server selection

Choose between groq (default) or llama-cpp (for local generation) using the -s or --server option:

llamatex render -p "pythagorean theorem" -s llama-cpp