直觉性设计的全平台Bot，多轮会话管理，多模态(语音/图片)理解，自维护套件，交叉追溯回复。
If you don't have the instant messaging platform you need or you want to develop a new application, you are welcome to contribute to this repository.
You can develop a new Controller by using Event.py.
Compatibility with multiple LLMs and integration with GPT and third-party systems is handled by our llm-kira project on GitHub.
Please submit an issue/discussion if you have a deployment issue rather than emailing me
- It can accurately limit billing, with limits and ID binding. ⚡️
- Supports asynchronous operations and can handle multiple requests simultaneously. 🚀
- Allows for private and group chats, catering to different scenarios. 💬
- Implements chat rate limiting to avoid overly frequent requests. ⏰
- Provides entertainment and interactive features, allowing for proactive engagement with users. 🎉
- Includes blacklists, whitelists, and quota systems to control conversation partners. 🔒
- Designed for full compatibility and strong scalability, adapting to different application scenarios. 🔌
- Features a memory pool that guarantees the storage of context memory for up to 1000 rounds, with dynamic construction. 💾
- Supports management, roles, and custom writing styles, providing more personalized options. 🤖
- Integrates Azure and Whisper local voice assistants, offering more ways for voice interaction. 🗣
- Allows for polling of multiple API keys for easy management, with automatic expiration reminders. 📊
- Supports multimodal interaction, including image Blip comprehension support, speech recognition, and sticker support. 👂😎
- Has a sound content safety system, including removable content safety components and official API filtering of content. 🔒
- Theoretically supports cross-platform access to any chat platform. 🌐
- The intuitive design of the chat allows for cross-replying, retracing replies, trigger-based replies, and the use of stickers for added fun. 😊
- Theoretical abstract design of third-party information injection LLM, supporting real-time content and automatic injection of the latest information to assist in answering. 🕸
- Self-maintaining model framework that supports any LLM model and any external API integration, abstracting and unifying access to GPT3 and GPT3.5. 🤖
Make sure your server has 1GB of RAM and 10GB of free storage.
For Arm architecture servers: curl --proto '=https' --tlsv1.2 -sSf https://sh.rustup.rs | sh (The setup.sh can now
automatically install rust.)
curl -LO https://raw.githubusercontent.com/LLMKira/Openaibot/main/setup.sh && sh setup.shFor Chinese users
curl -LO https://raw.kgithub.com/LLMKira/Openaibot/main/setup.sh && sh setup.sh- set Redis
apt-get install redis
systemctl enable redis.service --now- edit bot config
cp Config/app_exp.toml Config/app.toml
nano Config/app.toml# Comment out which part you don't want to start
# 注释你不需要的部分
# QQ Bot
[Controller.QQ]
master = [114, 514] # QQ number
account = 0 # Bot s QQ number
http_host = 'http://localhost:8080' # Mirai http Server
ws_host = 'http://localhost:8080' # Mirai Websocket Server
verify_key = ""
trigger = false # Proactive response when appropriate
INTRO = "POWER BY OPENAI" # Suffixes for replies
ABOUT = "Created by github.com/LLMKira/Openaibot" # /about
WHITE = "Group NOT in WHITE list" # Whitelist/Blacklist tips
# Proxy set, but does not proxy openai api, only bot
proxy = { status = false, url = "http://127.0.0.1:7890" }
# Telegram Bot
[Controller.Telegram]
master = [114, 514] # User Id @JsonDumpBot
botToken = '' # Bot Token @botfather
trigger = false
INTRO = "POWER BY OPENAI"
ABOUT = "Created by github.com/LLMKira/Openaibot"
WHITE = "Group NOT in WHITE list"
# 设置的代理，只代理 bot openai api->service.json
proxy = { status = false, url = "http://127.0.0.1:7890" }
# 基础对话事件服务器，Web支持或者音箱用&Use by Voice Assistant
[Controller.BaseServer]
host = "127.0.0.1"
port = 9559If you want configure the backend or openai proxy. Please Check Deploy Docs
- Telegram
Make sure the bot is a group admin or privacy mode is turned off.
Our robots can be started in multiple processes.
apt install npm
npm install pm2@latest -g
# or
yarn global add pm2
# test bot
python3 main.py
# run bot
pm2 start pm2.json
pm2 status
# stop bot
pm2 stop pm2.json
pm2 stop xx(id)
pm2 restart x(id)Once configured, send a message and use the /add_white_user command to add your platform ID returned by the bot to the
whitelist and you will be able to talk.
Or use /close_group_white_mode to turn off the bot's group whitelist mode.
In addition to the robot, we also have a voice assistant.
Voice Assistant is a web-dependent voice assistant that you can easily run on small devices through Azure or Openai's recognition services.
- Run BaseEvent Server
# 基础对话事件服务器，Web支持或者音箱用
[Controller.BaseServer]
port = 9559- Run Vits Server
https://github.com/LlmKira/MoeGoe
- Run Assistant
cd Assistant
cat install.md
pip3 install -r requirements.txt
python3 clinet.pyUse /add_api_key Command add OpenaiKey to Config/api_keys.json.
You'll be wondering about our multi-platform ID system. This is how we store your ID in our
json/database: real_id + suffix.
- toml
Use your real ID in app.toml, which is the whitelist prompt without the suffix.
- json/command
When using the user/group authorization command, you need to follow the real ID with the corresponding suffix ID.
| Controller | suffix_id | desc |
|---|---|---|
| 101 | ||
| Telegram | 100 | |
| Api | 103 |
| models | token limit | cost |
|---|---|---|
| gpt-3.5-turbo | 4095 | optimized for chat at 1/10th the cost of text-davinci-003. |
| gpt-3.5-turbo-0301 | 4095 | optimized for chat at 1/10th the cost of text-davinci-003. |
| models | token limit | cost |
|---|---|---|
| code-davinci-002 | 8000 | During this initial limited beta period, Codex usage is free. |
| code-cushman-001 | 2048 | During this initial limited beta period, Codex usage is free. |
| text-davinci-003 | 4000 | $0.0200 /1K tokens |
| text-curie-001 | 2048 | $0.0020 /1K tokens |
| text-babbage-001 | 2048 | $0.0005 /1K tokens |
| text-ada-001 | 2048 | $0.0004 /1K tokens |
To prevent malicious inducement and attacks, we have two measures: OpenAI TosApi filtering and simplified/traditional Chinese keyword filtering.
Simplified/traditional Chinese keyword filtering(Only For Chinese)
If not exist, the sensitive word file will be automatically initialized into Data/Danger.form, and it can be disabled
by clearing it to one line.
OpenAI TosApi filtering
Leave the inspection type array empty to disable the filter.
Please see Service Configuration Guide.
Our llm-kira dependency library is stored in the current package directory when there is no Redis support.
The application itself is stored in Redis for robustness, except for api_keys.json, service.json
and assistants.json.
If you have config.json, the application will automatically initialise this file. And you can update the configuration
to this file using the /config command.
Due to lack of maintainers, some commands only work on some platforms.
chat - talk
write - continue writing
forgetme - reset memory
remind - Scene setting cancel overwrite with short text
voice - voice support
style - set the preferred word
trigger - Admin initiates unsolicited responses
trace - Admin activates automatic tracking of associated channels
cross - whether the Admin starts a cross-response
silent - Admin starts silent error reporting
auto_adjust - automatic optimizer
set_user_cold - set user cooldown
set_group_cold - set group cooldown
set_token_limit - set output limit length
set_input_limit - set input limit length
see_api_key - Several Api keys now
del_api_key - Delete Api key
add_api_key - add Api key
config - get/backup hot configuration file
set_per_user_limit - set normal user limit
set_per_hour_limit - set user hour limit
promote_user_limit - Promote user limit
reset_user_usage - Reset user usage
add_block_group - block group
del_block_group - Unblock group
add_block_user - block user
del_block_user - Unblock user
add_white_group - add whitelist group
add_white_user - add whitelist user
del_white_group - delist whitelist group
del_white_user - remove whitelist user
update_detect - update sensitive words
open_user_white_mode - open user whitelist
open_group_white_mode - open group whitelist
close_user_white_mode - close user whitelist
close_group_white_mode - close group whitelist
open - open the robot
close - close the robot
change_head - set head switch
change_style - set the style switch
help - helpDetails On Deploy Guide
Network Plugins/Proxy Settings/Custom Model Names/Speech Services/Picture Understanding/Censor Configuration Please see Service Configuration Guide
详细接口/服务配置/自定义 请查看文档 Deploy Guide
贴纸设置/代理设置/切换其他模型/语音服务/图片理解/审查配置 请查看 服务配置
- ChatGPT Mirai Bot is a QQ bot based on the ChatGPT Web Api& Edge Api & GPT3.5
https://github.com/lss233/chatgpt-mirai-qq-bot
This project open source and available under
the [AGPL License](https://github.com/LLMKira/Openaibot/blob/main/LICENSE).CLAUSE 说明了如何授权，声明，附加条款等内容。
You wouldn't believe it, but Ai also wrote part of this Readme