GPT-Telegramus

ChatGPT and DALLE implementation as a telegram bot

Support project

Support the project by buying and listening to my music 🎵

Bandcamp
Spotify
Other platforms (Latest album)
YouTube (Updates rarely)
SoundCloud (Updates rarely)

Dependencies

ChatGPT (ChatGPT by OpenAI): https://chat.openai.com/chat
DALLE (DALLE by OpenAI): https://openai.com/dall-e-2/
acheong08/ChatGPT (API): https://github.com/acheong08/ChatGPT
python-telegram-bot (Telegram bot API): https://github.com/python-telegram-bot/python-telegram-bot

Get started

Install Python and pip
Download source code
Install requirements pip install -r requirements.txt --upgrade
For API type 1 (recommended) (revChatGPT API V1, free, uses same model as official website):
1. Create account at OpenAI. Make sure you have access to https://chat.openai.com/
2. Open https://chat.openai.com/api/auth/session
3. Copy value of accessToken into access_token in chatgpt_auth in chatgpt_api_1 in settings.json file
4. Configure proxy if needed
For API type 0 (Official API, more "stupid" model, uses credits):
1. Generate API Key https://platform.openai.com/account/api-keys and paste it into open_ai_api_key in chatgpt_api_0 in settings.json file
For DALL-E, generate API Key https://platform.openai.com/account/api-keys
Type Generated OpenAI API Key into open_ai_api_key in dalle in settings.json file
Create bot at https://t.me/BotFather
Type Bot's token into api_key in telegram in settings.json file
Run main script python main.py

Example settings.json:

{
  "modules": {
    "chatgpt": true,
    "chatgpt_api_type": 1,
    "dalle": true
  },
  
  "chatgpt_api_0": {
    "open_ai_api_key": "sk-2xxxxxxXXxXXXxxXXXxxXXXXXXXXXXXxxxxxxxxxxxxxxxXX",
    "engine": "text-davinci-003",
    "proxy": ""
  },
  
  "chatgpt_api_1": {
    "chatgpt_auth": {
      "email": "",
      "password": "",
      "session_token": "",
      "access_token": "xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx.........."
    },
	
    "proxy": {
      "enabled": true,
      "auto": true,
      "https_only": true,
      "manual_proxy": "http://111.222.123.111:443",
      "check_interval_seconds": 300,
      "check_message": "1+1",
      "check_message_timeout": 240,
      "check_reply_must_include": "2",
      "max_number_of_processes": 5,
      "initialization_timeout": 60,
      "too_many_requests_wait_time_seconds": 600
    }
  },
  
  "dalle": {
    "open_ai_api_key": "sk-2xxxxxxXXxXXXxxXXXxxXXXXXXXXXXXxxxxxxxxxxxxxxxXX",
    "image_size": "512x512",
    "use_proxy": true
  },
  
  "telegram": {
    "api_key": "1234567890:XXXxXxxXxxXXxxXXX-XXXXXXXxXxxxXxXxX",
    "queue_max": 5,
    "show_queue_message": true
  }
}

Bot messages

You can edit telegram bot messages by editing file messages.json. You can add new lines by adding \\n

Note: make shure you don't delete argumensts {0} in message and please restart script to apply changes

Running as service on linux

Install Python and pip
Clone repo
Edit settings.json
Install systemd sudo apt-get install -y systemd

Create new service file sudo nano /etc/systemd/system/gpt-telegramus.service

[Unit]
Description=GPT-Telegramus service
After=multi-user.target

[Service]
Type=simple
Restart=always
WorkingDirectory=YOUR DIRECTORY HERE/GPT-Telegramus
ExecStart=YOUR DIRECTORY HERE/GPT-Telegramus/run.sh
RestartSec=5

[Install]
WantedBy=multi-user.target

Reload systemctl daemon sudo systemctl daemon-reload
Enable service sudo systemctl enable gpt-telegramus.service
Start service sudo systemctl start gpt-telegramus.service

Running in Docker

WARNING: not tested and not recommended

Install Docker
Clone repo

Build container

docker buildx build -t telegramus --load -f Dockerfile .

Run the container

docker run -d -e TELEGRAMUS_SETTINGS_FILE=you_settings_file_location -e TELEGRAMUS_MESSAGES_FILE=you_messages_file_location --name gpt-telegramus --restart on-failure telegramus

Note: You can specify settings, messages and chats file location. (default location is in project folder):

ENV TELEGRAMUS_SETTINGS_FILE "settings.json"
ENV TELEGRAMUS_MESSAGES_FILE "messages.json"
ENV TELEGRAMUS_MESSAGES_FILE "chats.json"

Proxy to bypass OpenAI's geo-blocking

It is possible to bypass geo-blocking. GPT-Telegramus includes automatic proxy-list downloading

Set enabled in proxy in settings.json to true
Restart app and hope for the best.

GPT-Telegramus will have to download the proxy list itself and start trying various proxies (see console for logs). Sometimes trying can take a very long time (Usually up to half of the proxies is at least one that works)

If you have proxy that definitely works you can specify it in manual_proxy in proxy in settings.json. Make sure you set auto to false when using manual_proxy

proxy settings description:

enabled - Whether proxy login is enabled.
auto - Download proxies automatically. Otherwise, use manual_proxy
https_only - Don't include http proxies in list
manual_proxy - Manual proxy server. It must support HTTPS, but you need to type it in http://IP:PORT format
check_interval_seconds - Automatic connection check interval (in seconds)
check_message - This message will be sent as a request
check_message_timeout - How long should a response take?
check_reply_must_include - The response message must contain this text to consider that the check was successful
max_number_of_processes - How many background processes are used for checking proxy
initialization_timeout - How long to wait for initialization with proxy

Telegram bot commands

/start - Shows version
/help - Shows list of commands
/queue - Shows requests queue
/gpt YOUR REQUEST - Request to ChatGPT (or just type your request as message without /gpt command)
/draw YOUR REQUEST - Request to DALLE
/clears - Clears chat history
/restart - Restart chatGPT and Telegram Bot (not tested properly)
Type any message to ask chatGPT (without /gpt command)

Chat history

File chats.json saves conversation_id and parent_id for each telegram chat to prevent history collision

It only works properly if chatgpt_api_type is set to 1

TODO

Make whitelist for users and admin account

LukasCCB/GPT-Telegramus