private-gpt/octogen

Octogen is an Open-Source Code Interpreter Agent Framework

中文

Octogen

an open-source code interpreter
一款开源可本地部署的代码解释器

News

• 2023-10-24 🎉 Octogen v0.5.0 has been released 🎉

octogen_demo.mp4

Supported OSs	Supported Interpreters	Supported Dev Enviroment

Getting Started

Requirement

• python 3.10 and above
• pip
• docker 24.0.0 and above, or podman

To deploy Octogen, the user needs permission to run Docker commands.
To use codellama, your host must have at least 8 CPUs and 16 GB of RAM.

Install the octogen on your local computer

1. Install og_up

pip install og_up

2. Set up the Octogen service

og_up

You have the following options to select

• OpenAI , recommanded for daily use
• Azure OpenAI
• CodeLlama,
• Octogen agent services powered by GPT4 and Codellama 34B

The default is using docker as container engine. use podman with flag --use_podman

3. Execute the command og, you will see the following output

Welcome to use octogen❤️ . To ask a programming question, simply type your question and press esc + enter
You can use /help to look for help

[1]🎧>

Development

Prepare the environment

git clone https://github.com/dbpunk-labs/octogen.git
cd octogen
python3 -m venv octogen_venv
source octogen_venv/bin/activate
pip install -r requirements.txt

Run the sandbox including Agent with mock model and Kernel

$ bash start_sandbox.sh
$ og

Welcome to use octogen❤️ . To ask a programming question, simply type your question and press esc + 
enter
Use /help for help

[1]🎧>hello
╭─ 🐙Octogen ─────────────────────────────────────────────────────────────────────────────────────────╮
│                                                                                                     │
│  0 🧠 how can I help you today?                                                                     │
│                                                                                                     │
╰─────────────────────────────────────────────────────────────────────────────────────────────────────╯
[2]🎧>

• To use openai for development, just update the config in the start_sandbox.sh with the example of openai-env.example
• To use azure openai for development, just update the config in the start_sandbox.sh with the example of azure-env.example
• To use codellama for development, just update the config in the start_sandbox.sh with the example of codellama-env.example

Supported API Service

name	type	status	installation
Openai GPT 3.5/4	LLM	✅ fully supported	use og_up then choose the OpenAI
Azure Openai GPT 3.5/4	LLM	✅ fully supported	use og_up then choose the Azure OpenAI
LLama.cpp Server	LLM	✔️ supported	use og_up then choose the CodeLlama
Octopus Agent Service	Code Interpreter	✅ supported	apply api key from octogen.dev then use og_up then choose the Octogen

The internal of local deployment

• Octogen Kernel: The code execution engine, based on notebook kernels.
• Octogen Agent: Manages client requests, uses ReAct to process complex tasks, and stores user-assembled applications.
• Octogen Terminal Cli: Accepts user requests, sends them to the Agent, and renders rich results. Currently supports Discord, iTerm2, and Kitty terminals.

Features

• Automatically execute AI-generated code in a Docker environment.
• Experiment feature, render images in iTerm2 and kitty.
• Upload files with the /up command and you can use it in your prompt
• Experiment feature, assemble code blocks into an application and you can run the code directly by /run command
• Support copying output to the clipboard with /cc command
• Support prompt histories stored in the octopus cli

if you have any feature suggestion. please create a discuession to talk about it

Roadmap

• roadmap for v0.5.0