README_EN.md

---

🍩 Deploy Docs & 🧀 Dev Docs & 📝 中文Readme & 🤝 How to contribute

Any issues with deployment? Submit an Issue to help us fix SLA

The project is Open source reproduction attempt for ChatGpt like, with FunctionCall and ToolCall as the core, supporting multiple messaging platforms.

Using message queue, it can easily handle function requests and support complex plug-in and functional design. Well supported access to file.

Supports multiple model sources and cross-platform message forwarding.

Demo

Unlike previous projects, this project tries to replicate ChatGpt's plugin system based on the messaging platform, implementing some or more features.

Because func call is a feature, it only supports Openai type api, and does not intend to support LLM without func call.

📦 Feature

• 🍪 Complete plug-in development ecosystem, using classic design, can be used after installation through pip
• 📝 Messaging system, no limit on time, no limit on sender, define sender and receiver, completely decoupled logic
• 📎 Route messages, customize message routing, and use routing to determine how to operate.
• 📬 Public open quota/private self-configured backend/proxy token authentication, providing flexible and scalable authentication development solutions
• 🍾 Support middleware interception development, and develop extensions to operate data before and after the process
• 🎵 Refined statistics system, easy to count usage
• 🍰 Support plug-in human-in-the-loop verification, authentication, and plug-in blacklist can be set
• 📦 Improve standard file interaction support, upload/download files
• 🍖 Supports individual configuration of environment keys and provides personal private environment variables for plug-ins
• 🍟 Supports incremental support for large language models, supports multi-platform expansion, and can be adapted by inheriting standard classes
• 🍔 Supports both FunctionCall and ToolCall features to dynamically build the required function classes based on the model

🧀 Introduction to authentication system

The authentication system we use is called Service Provider, which is the service provider. Its role is to assign Endpoint/Key/Model to each sender for authentication. Has a token as a bound OpenKey. The program will call the set Service Provider to read the private Key/configuration Token to obtain the authentication information.

Both the authentication component and the backend need to be implemented by yourself.

🧀 Preview of some plugins

Sticker Converter	Timer Func	Translate

🎬 Platform support

Platform	Support	File System	Tip
Telegram	✅	✅
Discord	✅	✅
Kook	✅	✅	No Support Replies start
Slack	✅	✅	No Support Replies start
QQ	❌
Wechat	❌
Twitter	❌
Matrix	❌
IRC	❌
...			Create issue/pr

📦 Quick Start

Read the 🧀 Deployment Documentation for more information.

Please use pdm run python3 start_sender.py pdm run python3 start_receiver.py to test whether it can run normally.

Performance indicator test (Until 2023/11/1)

Attention, does not include the memory usage of services such as pm2, redis, rabbitmq, mongodb, docker, etc.

Process	Memory Max Head Size	Tester	Client
receiver	120.847MB	python3 -m memray run --live start_receiver.py	telegram
sender	83.669MB	python3 -m memray run --live start_sender.py	telegram

🥣 Docker

Build Hub: sudoskys/llmbot

Automatic Docker/Docker-compose installation

If you are using a brand new server, you can use the following shell to try to automatically install this project.

This script will automatically install the required services and map ports using Docker methods, if you have deployed redis, rabbitmq, mongodb.

Please modify the docker-compose.yml file yourself.

curl -sSL https://raw.githubusercontent.com/LLMKira/Openaibot/main/deploy.sh | bash

Manual Docker-compose installation

git clone https://github.com/LlmKira/Openaibot.git
cd Openaibot
pip install pdm && pdm install -G bot
docker-compose -f docker-compose.yml up -d

Update the image with docker-compose pull.

View Shell in docker, use docker exec -it llmbot /bin/bash, type exit to exit.

🍔 Shell

To manually start using Pm2, you need to install redis, rabbitmq, mongodb by yourself.

git clone https://github.com/LlmKira/Openaibot.git
cd Openaibot
pip install pdm && pdm install -G bot
apt install npm -y && npm install pm2 && pm2 start pm2.json
pm2 monit

Use pm2 restart pm2.json to restart the program.

Recommend using pdm for dependency management, because we use pydantic^1.9.0, in order to prevent version conflicts, we use pdm for dependency management.

🍪 Slash Command

clear - erase chat history
help - show docs
chat - chat
task - chat with function_enable
ask - chat with function_disable
tool - list all functions
set_endpoint - set private key and endpoint
clear_endpoint - erase private key and endpoint
auth - auth a function
env - env for function
token - bind token
token_clear - clear token binding
func_ban - ban a function
func_unban - unban a function
bind - Bind rss platforms
unbind - Unbind rss platforms

💻 How to develop?

For plugin development, please refer to the sample plugins in the plugins directory.

Plugin development please refer to 🧀 Plugin Dev Docs

🤝 We need your help!

This is a long term project and we started the development of the LLM APP very early!

We applied a plugin-like system and search online before GPT3 OpenaiApi was released(davinci-003)

After many iterations, we have worked hard to make this project more standardized, generic, and open.

We can't do it on our own at the moment:

• We need help with the documentation
• Web UI

Feel free to submit a Pull Request or discuss, we'd love to receive your contribution!

📜 Agreement

This project has no relationship with OPENAI/ChatGpt.