eras 0.5.2

A terminal command library that provides a Natural Language Interface for running shell commands.

Unlock the Power of Your Terminal with ERAS 🚀

Welcome to ERAS: the Easily Runnable AI Shell that revolutionizes your command-line experience.

Say goodbye to the hassle of searching for the right command syntax—ERAS transforms your natural language instructions into precise terminal commands instantly.

Why ERAS?

• Seamless Integration : Compatible with Windows, Mac, and Linux, ERAS fits right into your workflow, regardless of your operating system. 🖥️
• Effortless Execution : No more jumping between your terminal and browser. Just type what you need, and ERAS will do the rest. ⚡
• AI-Powered Precision : Harness the power of OpenAI or any OpenAI-spec-compliant model like Llama.cpp server to ensure your commands are executed accurately and efficiently. 🤖

Experience the future of terminal commands with ERAS and elevate your productivity today. 🌟

Demo

Eras Demo

Install

Mac & Linux

pip install eras

If you get an error about externally-managed-environment it's because you have python installed via homebrew, and need to use pipx instead:

pipx

brew install pipx
pipx ensurepath
pipx install eras

With pipx you may need to open a new terminal window or source .zshrc a couple of times.

Windows

pip install eras

You will be prompted for various configuration settings. Once set, close your terminal window and open up a new one.

Use

The first use of Eras will prompt you to enter various configs. See the Config Examples section below.

The first use will also prompt you to allow accessibility features for Terminal.app, which is required to allow Eras to populate the terminal with the shell command.

When you ask Eras a question, it will use AI to create a shell command, then populate your next terminal line with the command, so all you have to do is review and press enter.

Usage Examples

% eras list all files in the current directory

ERAS will then populate the next terminal line with the command:
% ls

All you have to do is confirm and hit enter!

...
% eras create a new file named jason.txt with contents "hello world"
% echo "hello world" > jason.txt
...

Chat Examples

% eras /chat What is the capital of France 
% Paris is the capital of France

Note: if you want to use ? or other special terminal chars you should surround your question in quotes
% eras /chat "What is the capital of France?"
% Paris is the capital of France

Config Examples

On the first run of eras, you will be prompted to enter config values, which are saved as env vars.

To run the config again, you can run eras /config

An API key must be provided when using OpenAI or Llama.cpp (can be anything unless you explicitly set one in Llama.cpp)

OpenAI

For OpenAI you must provide:

• API Key - Grab an OpenAI key from here
• Model name - (defaults to gpt-4) Find any valid one here

Alternative OpenAI Server (e.g. Llama.cpp server)

You will be prompted to enter the alternative base url. e.g. http://127.0.0.1:8080

Llama.cpp Setup Instructions

Find the server setup instructions at: https://github.com/ggerganov/llama.cpp

NVIDIA GPU Instructions

If you are running a NVIDIA GPU, you will want to build the llama server with CUDA enabled, as outlined here

I followed these instructions:

• Download Visual Studio and install CMake tools

  • Go to https://visualstudio.microsoft.com/downloads/ and download “Visual Studio 2022 (Community)”

• Open VisualStudioSetup.exe to install Visual Studio

• Select Desktop development with C++ to install CMake tools

• Download a gguf model:

  • e.g. https://huggingface.co/MaziyarPanahi/Llama-3-8B-Instruct-32k-v0.1-GGUF/blob/main/Llama-3-8B-Instruct-32k-v0.1.Q4_K_M.gguf

Run these commands in Developer Command Prompt (search Windows and it'll come up)

git clone https://github.com/ggerganov/llama.cpp.git
cd llama.cpp
mkdir build
cd build
cmake .. -DLLAMA_CUDA=ON
cmake --build . --config Release

Start the Llama server:

.\bin\Release\server.exe -m C:\shared-drive\llm_models\Llama-3-8B-Instruct-32k-v0.1.Q4_K_M.gguf -ngl 9999 --host 0.0.0.0 --ctx-size 32000