emkshv / telegram-llm-assistant

A minimalistic LLM-powered Telegram assistant written in Rust that uses a self-contained Sqlite database and is very easy to install.

A minimalistic Telegram assistant with customizable LLM API providers. It has a very simple installation process: download the binary for your architecture and run it. It will create an Sqlite3 database file, run migrations and start the bot.

Built with:

• Mobot Telegram bot framework
• Sqlx SQL toolkit
• Sqlite self-contained database

Creating a Telegram bot and obtaining a token

Go to BotFather and enter /newbot. Fill in the description and save the token to the TELEGRAM_TOKEN environment variable. To define the commands for the autocomplete: enter /setcommands, select your bot, and then paste:

new - Clear the current context and start a new chat.
get_behavior - Display the current system message that defines the bot's behavior.
set_behavior - Set the new system message for defining the bot's behavior.
get_model - Get the current completion model.
set_model - Set the completion model for your bot.
version - Display the current version.

Running using Docker

Make sure you have Docker & Docker Compose. On desktop, you can use Docker Desktop or OrbStack.

The Docker Compose file expects your environment variables to be loaded:

cp .env.example .envrc
# edit .envrc
source .envrc

Build the image and run the container:

docker build -t telegram-llm-assistant .
docker-compose up

Local development

• Install Rust via RustUp
• cp .env.example .envrc
• Edit .envrc to set environment variables
• Load environment variables from .envrc using direnv, or source .envrc-equivalent in your shell.
• Now you can compile with cargo build

Building from source

• Install Rust via RustUp
• Clone the repository
• Run cargo build --release