Expose LLM-powered assistants through messaging platforms such as Telegram and Discord

🤖 ← 🐈 ← 🌐 ← 📱 ← 🧙‍♂️

---

llm-expose gives you a channel-first CLI workflow: configure providers, attach channels, control pairings, and optionally integrate MCP servers for tool-aware completions.

Features

• Multi-channel support (Telegram and Discord).
• LiteLLM provider support for broad model compatibility.
• Local OpenAI-compatible endpoint support.
• MCP server integration for tool-aware responses.
• Pairing-based access control per channel.
• CLI-first setup and operations.

Installation

Quick Install (One-Liner)

Linux & macOS:

curl -fsSL https://raw.githubusercontent.com/edo0xff/llm-expose/main/scripts/install.sh | bash

Windows (PowerShell as Administrator):

powershell -ExecutionPolicy Bypass -Command "iex (New-Object Net.WebClient).DownloadString('https://raw.githubusercontent.com/edo0xff/llm-expose/main/scripts/install-windows.ps1')"

From PyPI

pip install llm-expose

From source

git clone https://github.com/edo0xff/llm-expose.git
cd llm-expose
pip install -e .

Development install

pip install -e '.[dev]'

See scripts/README.md for detailed installation instructions and troubleshooting.

Quick Start

llm-expose is interactive by default, which is usually the fastest path for humans. Use --no-input for headless automation and add -y when the command can require confirmation.

1. Configure a model:

llm-expose add model

2. Configure a channel (interactive):

llm-expose add channel

3. Pair an allowed user/chat ID:

llm-expose add pair 123456789 --channel my-telegram

4. Start the channel runtime:

llm-expose start

Headless equivalent (CI/scripts):

llm-expose add model --name gpt4o-mini --provider openai --model-id gpt-4o-mini -y --no-input
llm-expose add channel --name my-telegram --client-type telegram --bot-token "123456789:AAExampleTelegramToken" --model-name gpt4o-mini -y --no-input
llm-expose add pair 123456789 --channel my-telegram --no-input
llm-expose start --channel my-telegram -y --no-input

If you are unsure about available options, run:

llm-expose --help
llm-expose add --help
llm-expose start --help

Pairing Model

Incoming chat/channel IDs must be explicitly paired before the service replies.

When an unpaired ID sends a message, the service returns:

This instance is not paired. Run llm-expose add pair <channel-id>

Pairings are stored per channel configuration.

Common pairing commands:

• llm-expose add pair <id> --channel <channel-name>
• llm-expose list pairs
• llm-expose list pairs --channel <channel-name>
• llm-expose delete pair <id> --channel <channel-name>

Configuration Workflow

llm-expose currently uses CLI commands to persist configuration (models, channels, and MCP settings).

Recommended setup order:

1. Add one or more models (llm-expose add model ...).
2. Add one or more channels (llm-expose add channel ...).
3. Add optional MCP servers (llm-expose add mcp ...).
4. Pair allowed IDs (llm-expose add pair ...).
5. Run exposure service (llm-expose start ...).

Development

Run quality checks:

ruff check .
black --check .
mypy llm_expose
pytest

Roadmap

• PyPI release automation.
• Hosted docs site with architecture and API references.
• More channel adapters and provider presets.

Contributing

See CONTRIBUTING.md.

License

MIT. See LICENSE.