sorokinvld/tg-ai-connector

This package can be used to connect Telegram bot to AI engines such as OpenAI ChatGPT, Dall-E, Midjourney, Stable Diffusion, etc.

Installation

Docker

https://hub.docker.com/repository/docker/desprit/tg-ai-connector/general

# If you're on Linux
docker run -d --rm -v /path/to/config.toml:/app/bot/config.toml desprit/tg-ai-connector:1.0.0
# If you're on MacOS M1
docker run -d --rm -v /path/to/config.toml:/app/bot/config.toml --platform linux/amd64 desprit/tg-ai-connector:1.0.0

Map log file from the host to the container if needed:

docker run -v /path/to/config.toml:/app/bot/config.toml -v /path/to/log.txt:/app/bot/log.txt desprit/tg-ai-connector:1.0.0
tail -f /path/to/log.txt

Manual

Python 3.10+ is required.

# Install system packages
apt install python3.10-dev python3.10-venv

# Create Virtual environment
python3.10 -m venv venv
# Activate it
source venv/bin/activate
# Install dependencies
pip install -r requirements.txt

Available integrations

1. OpenAI through Official API
   • Dall-E
   • ChatGPT
   • Text Completion
2. Replicate through Replicate
   • Midjourney
   • Stable Diffusion
   • Other networks

Relicate offers many different networks. Initially I was only using image generation but recently discovered a speech-to-text network. Since you need to provide an audio to it, the easiest way I found is to use "reply" in your Telegram chat. Basically you select an audio message, reply to it with /a command (or whatever command you chose for the openai/whisper integration) and optionally provide language, for example /a de.

You may also need to change Bot Privacy settings: https://stackoverflow.com/questions/50204633/allow-bot-to-access-telegram-group-messages

Usage

Authorization

You can whitelist/blacklist users and chats via the config file. Alternatively, set the admin_id and then use bot commands:

/whitelist user_id
/whitelist username
/whitelist chat_id
/blacklist user_id
/blacklist username
/blacklist chat_id

Config

Create config.toml, fill YOUR_TELEGRAM_TOKEN and tokens for integrations:

debug = true

[general]
text_history_ttl = 300 # optional, for how long to store user messages, default 5 minutes
text_history_size = 10 # optional, how many messages from each user to keep

[telegram]
bot_token = "YOUR_TELEGRAM_TOKEN"
admin_id = 111 # optional, id of admin user who can whitelist and blacklist chats and users
allowed_users = [123, 234] # optional, a list of users from which messages are allowed
allowed_chats = [345, 456] # optional, a list of changes from which all messages are allowed

[integrations]

[integrations.openai]
api_key = "OPEN_AI_TOKEN" # set it to enable OpenAI integration
max_tokens = 1000 # max tokens to return by OpenAI text models, default 500
[[integrations.openai.networks]]
name = "completion"
version = "text-davinci-003"
command = "t" # Telegram command to trigger Text Completion requests
type = "text"
[[integrations.openai.networks]]
name = "chat"
version = "gpt-4"
command = "c" # Telegram command to trigger ChatGPT requests
type = "text"
[[integrations.openai.networks]]
name = "image"
version = "dalle"
command = "d" # Telegram command to trigger Dall-E requests
type = "text"

[integrations.replicate]
api_key = "REPLICATE_TOKEN" # set it to enable Replicate integration
[[integrations.replicate.networks]]
name = "tstramer/midjourney-diffusion"
version = "436b051ebd8f68d23e83d22de5e198e0995357afef113768c20f0b6fcef23c8b"
command = "m" # Telegram command to trigger Midjourney requests
type = "image"
[[integrations.replicate.networks]]
name = "stability-ai/stable-diffusion"
version = "f178fa7a1ae43a9a9af01b833b9d2ecf97b1bcb0acfd2b1c1c1c1c1c1c1c1c1c"
command = "s" # Telegram command to trigger Stable Diffusion requests
type = "image"
[[integrations.replicate.networks]]
name = "cjwbw/anything-v3.0"
version = "f410ed4c6a0c3bf8b76747860b3a3c9e4c8b5a827a16eac9dd5ad9642edce9a2"
command = "anything" # Telegram command to trigger requests to cjwbw/anything-v3.0
type = "image"
[[integrations.replicate.networks]]
name = "cjwbw/portraitplus"
version = "629a9fe82c7979c1dab323aedac2c03adaae2e1aecf6be278a51fde0245e20a4"
command = "portraitplus" # Telegram command to trigger requests to cjwbw/portraitplus
type = "image"
[[integrations.replicate.networks]]
name = "openai/whisper"
version = "e39e354773466b955265e969568deb7da217804d8e771ea8c9cd0cef6591f8bc" # Telegram command to trigger requests to openai/whisper speech-to-text model
command = "a"
type = "audio"

Explore Replicate website to find more models.

Running in development

# Start bot
make start
# Check logs
tail -f log.txt

Running on the server

# Start bot
python -m src.bot &
# Stop bot
make stop

Useful Telegram commands

# Test bot is alive
/ping
# Show available commands
/help

ChatGPT and Text Completion stores history of requests which can be manually cleaned using clear command.

/p clear

Troubleshooting

• A request to the Telegram API was unsuccessful. Error code: 409. Description: Conflict: terminated by other getUpdates request; make sure that only one bot instance is running

You have another bot running on the background. Stop it with make stop, if it doesn't work find that process and kill it manually:

ps aux | grep "src.bot"
kill -9 <PID>