---

🍩 Deploy Docs & 🧀 Dev Docs & 🤝 Contribute

Don't hesitate to Star ⭐️, Issue 📝, and PR 🛠️

Python>=3.9

This project uses the ToolCall feature.

It integrates a message queuing and snapshot system, offering plugin mechanisms and authentication prior to plugin execution.

The model adheres to the Openai Format Schema. Please adapt using gateway or one-api independently.

Demo	Vision With Voice

🔨 Roadmap

• Removal of legacy code
• Deletion of metric system
• Deletion of model selection system, unified to OpenAI Schema
• Implementation of a more robust plugin system
• Project structure simplification
• Elimination of the Provider system
• Hook support
• Access to TTS
• Add standalone support for gpt-4-turbo and vision
• Add LLM reference support to the plugin environment. (extract && search in text)

📦 Features

• 🍪 A comprehensive plugin development ecosystem, adopting a classic design, and seamless integration with plugins through pip installation
• 📝 Message system with no time or sender constraints, offering fully decoupled logics
• 📬 Offers Login via a URL mechanism, providing a flexible and expandable authentication development solution
• 🍰 Empowers users to authorize plugin execution. Users can configure plugin environment variables at their discretion
• 📦 Support for plugins to access files
• 🍟 Multi-platform support – extend new platforms by inheriting the base class
• 🍔 Plugins can determine their appearance in new sessions dynamically, preventing performance degradation despite large amounts of plugins

🍔 Login Modes

• Login via url: Use /login token$https://provider.com to Login. The program posts the token to the interface to retrieve configuration information, how to develop this.
• Login: Use /login https://api.com/v1$key$model to login

🧀 Plugin Previews

Sticker Converter	Timer Function	Translate Function

🎬 Platform Support

Platform	Support	File System	Remarks
Telegram	✅	✅
Discord	✅	✅
Kook	✅	✅	Does not support triggering by reply
Slack	✅	✅	Does not support triggering by reply
QQ	❌
Wechat	❌
Twitter	❌
Matrix	❌
IRC	❌
...			Create Issue/PR

📦 Quick Start

Refer to the 🧀 Deployment Document for more information.

📦 One-click Deployment

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

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

📦 Manual Installation

# Install Voice dependencies
apt install ffmpeg
# Install RabbitMQ
docker pull rabbitmq:3.10-management
docker run -d -p 5672:5672 -p 15672:15672 \
  -e RABBITMQ_DEFAULT_USER=admin \
  -e RABBITMQ_DEFAULT_PASS=8a8a8a \
  --hostname myRabbit \
  --name rabbitmq \
  rabbitmq:3.10-management
docker ps -l
# Install Project
git clone https://github.com/LlmKira/Openaibot/
cd Openaibot
pip install pdm
pdm install -G bot
cp .env.exp .env && nano .env
# Test
pdm run python3 start_sender.py
pdm run python3 start_receiver.py
# Host
apt install npm
npm install pm2 -g
pm2 start pm2.json

🥣 Docker

Build Hub: sudoskys/llmbot

Note that if you run this project using Docker, you will start Redis, MongoDB, and RabbitMQ. But if you're running locally, just RabbitMQ

Manual Docker-compose Installation

git clone https://github.com/LlmKira/Openaibot.git
cd Openaibot
cp .env.exp .env&&nano .env
docker-compose -f docker-compose.yml up -d

The Docker configuration file docker-compose.yml contains all databases. In fact, Redis and MongoDB are not required. You can remove these databases yourself and use the local file system.

Update image using docker-compose pull.

Use docker exec -it llmbot /bin/bash to view Shell in Docker, enter exit to exit.

🍪 Slash Commands

clear - Deletes chat records
help - Displays documentation
chat - Conversation
task - Use a function to converse
ask - Disable function-based conversations
tool - Lists all functions
login - Login
auth - Authorize a function
env - Environment variables of the function

💻 How to Develop Plugins?

Refer to the example plugins in the plugins directory and the 🧀 Plugin Development Document for plugin development documentation.

Hooks

Hooks control the EventMessage in sender and receiver. For example, we have voice_hook in built-in hooks.

you can enable it by setting VOICE_REPLY_ME=true in .env.

/env VOICE_REPLY_ME=yes
# must

/env REECHO_VOICE_KEY=<key in dev.reecho.ai>
# not must

use /env VOICE_REPLY_ME=NONE to disable this env.

check the source code in llmkira/extra/voice_hook.py, learn to write your own hooks.

🧀 Sponsor

📜 Notice

This project, named OpenAiBot, signifying "Open Artificial Intelligence Robot", is not officially affiliated with OpenAI.