Examples & Cookbook

Real-world examples and code snippets to help you get started with llm-expose.

All examples below use the default interactive flow unless noted otherwise. For automation/CI, use the same commands with explicit flags plus --no-input (and -y when confirmation is required).

Basic Setup Examples

Example 1: Simple Telegram Bot with GPT-4o

Complete setup for a basic Telegram bot powered by GPT-4o.

Step-by-stepAll at once (with env var)

# 1. Get your Telegram bot token from @BotFather
# (Create a new bot and copy the token)

# 2. Add the model
llm-expose add model
# Suggested answers:
# - Name: gpt4o
# - Provider: openai
# - Model ID: gpt-4o

# 3. Create the Telegram channel
llm-expose add channel
# Suggested answers:
# - Name: telegram-main
# - Type: telegram
# - Token: <paste your token>
# - Model: gpt4o

# 4. Pair your Telegram Chat ID
# Get your ID from @userinfobot
llm-expose add pair 123456789 --channel telegram-main

# 5. Start the bot!
llm-expose start --channel telegram-main

# Set API key
export OPENAI_API_KEY="sk-..."
export TELEGRAM_BOT_TOKEN="123:ABC..."

# Add model
llm-expose add model --name gpt4o --provider openai --model-id gpt-4o

# Create channel with token from env
llm-expose add channel  # Select Telegram, paste token

# Pair user (replace with your ID)
llm-expose add pair 123456789 --channel telegram-main

# Run!
llm-expose start --channel telegram-main

Example 2: Discord Bot with Claude 3 Haiku

Budget-friendly Discord bot using smaller Claude model.

# Set API key
export ANTHROPIC_API_KEY="sk-ant-..."

# Add model (cheaper, fast)
llm-expose add model
# Suggested answers:
# - Name: claude-haiku
# - Provider: anthropic
# - Model ID: claude-3-haiku-20240307

# Create Discord channel
llm-expose add channel
# Suggested answers:
# - Name: discord-support
# - Type: discord
# - Token: <your bot token>
# - Model: claude-haiku

# Pair users (Discord User IDs)
llm-expose add pair 987654321 --channel discord-support
llm-expose add pair 567890123 --channel discord-support

# Start
llm-expose start --channel discord-support

---

Advanced Examples

Example 3: Telegram Bot with File System Access (MCP)

Telegram bot that can read and write files on your system.

SetupUsage in Telegram

# 1. Set API key
export OPENAI_API_KEY="sk-..."
export TELEGRAM_BOT_TOKEN="123:ABC..."

# 2. Add model
llm-expose add model
# Suggested answers:
# - Name: gpt4o
# - Provider: openai
# - Model ID: gpt-4o

# 3. Add MCP server for file system
llm-expose add mcp
# When prompted:
# - Name: filesystem
# - Type: stdio
# - Command: npx @modelcontextprotocol/server-filesystem /home/user/shared

# 4. Create Telegram channel
llm-expose add channel

# 5. Attach MCP server to channel
llm-expose config set --channel telegram-main --mcp filesystem

# 6. Start!
llm-expose start --channel telegram-main

User: Create a summary file for today
Bot: I'll create a summary file for you...
<bot creates summary.md in shared folder>
Bot: Done! I've created summary.md

User: What files are in the shared folder?
Bot: The files in /home/user/shared are:
- summary.md (created today)
- notes.txt
- data.json

Example 4: Discord Community Server with Role-Based Access

Multiple channels for different roles/departments.

# Setup models
llm-expose add model --name gpt4o-turbo \
  --provider openai \
  --model-id gpt-4o

llm-expose add model --name gpt4o-mini \
  --provider openai \
  --model-id gpt-4o-mini

# Create channels
llm-expose add channel
# • discord-general (gpt4o-mini for cost)

llm-expose add channel
# • discord-support (gpt4o-turbo for better responses)

# Pair support team users
llm-expose add pair 111111111 --channel discord-support  # User 1
llm-expose add pair 222222222 --channel discord-support  # User 2

# Pair all users for general (pair a group chat instead)
llm-expose add pair 333333333 --channel discord-general  # Group/Channel ID

# Start both channels
llm-expose start --channel discord-general &
llm-expose start --channel discord-support

Example 5: Multi-Model Telegram Bot

Different models for different use cases.

# Add multiple models
llm-expose add model --name fast --provider openai --model-id gpt-4o-mini
llm-expose add model --name smart --provider openai --model-id gpt-4o
llm-expose add model --name creative --provider anthropic --model-id claude-3-opus

# Create separate channels for each
llm-expose add channel  # -> telegram-fast (model: fast)
llm-expose add channel  # -> telegram-smart (model: smart)
llm-expose add channel  # -> telegram-creative (model: creative)

# Pair same user to all channels
llm-expose add pair 123456789 --channel telegram-fast
llm-expose add pair 123456789 --channel telegram-smart
llm-expose add pair 123456789 --channel telegram-creative

# Start all three
for channel in telegram-fast telegram-smart telegram-creative; do
  llm-expose start --channel $channel &
done

---

Integration Examples

Example 6: Local Development Setup with Ollama

Use a local LLM for free development.

Install OllamaConfigure llm-expose

# Download Ollama from https://ollama.ai
# Or on Linux:
curl https://ollama.ai/install.sh | sh

# Pull a model
ollama pull llama2:7b
# or
ollama pull mistral

# Start Ollama (it runs on localhost:11434)
ollama serve

# Add Ollama model (OpenAI-compatible endpoint)
llm-expose add model \
  --name llama2-local \
  --provider openai \
  --model-id llama2 \
  --api-base http://localhost:11434/v1

# Create channel
llm-expose add channel
# (Select Telegram or Discord)

# Start (no API key needed!)
llm-expose start --channel my-channel

Example 7: GitHub Issue Resolver Bot

Discord bot that monitors and comments on GitHub issues (with MCP GitHub).

# Setup
export OPENAI_API_KEY="sk-..."
export GITHUB_TOKEN="ghp_..."

# Add model
llm-expose add model --name gpt4o --provider openai --model-id gpt-4o

# Add GitHub MCP server
llm-expose add mcp
# • Name: github
# • Command: npx @modelcontextprotocol/server-github

# Create Discord channel for GitHub updates
llm-expose add channel
# • discord-github
# • Model: gpt4o
# • MCP: github

# Pair GitHub bot user/webhook channel
llm-expose add pair <your-github-bot-channel-id> --channel discord-github

# Now you can:
# - Forward GitHub issues to Discord
# - Use bot to analyze and comment on issues
# - Automated triage with the bot's insights

---

Configuration Management

Example 8: Backup and Version Control Configuration

Keep your configuration in git (securely).

# Create a config backup
mkdir -p ~/backups
cp -r ~/.config/llm-expose ~/backups/llm-expose-backup

# Create a git repo for configuration (without secrets!)
cd ~/projects
git init llm-expose-config
cd llm-expose-config

# Copy config structure (we'll remove secrets next)
cp -r ~/.config/llm-expose . 

# Create .gitignore to exclude secrets
cat > .gitignore << 'EOF'
# Don't commit API keys!
*.json
!models-schema.json
!channels-schema.json
.env
.env.local
EOF

# For deployment, store secrets in environment variables
cat > .env.example << 'EOF'
OPENAI_API_KEY=sk-...
TELEGRAM_BOT_TOKEN=123:ABC...
ANTHROPIC_API_KEY=sk-ant-...
EOF

# Now safe to commit
git add .gitignore .env.example
git commit -m "Initial config structure"

Example 9: Production Deployment with systemd

Run bots as system services that auto-restart.

Create service fileEnable and start

# Create /etc/systemd/system/llm-expose-telegram.service
sudo tee /etc/systemd/system/llm-expose-telegram.service > /dev/null << 'EOF'
[Unit]
Description=LLM Expose Telegram Bot
After=network.target

[Service]
Type=simple
User=llm-user
Group=llm-user
WorkingDirectory=/home/llm-user
Environment="OPENAI_API_KEY=sk-..."
Environment="TELEGRAM_BOT_TOKEN=123:ABC..."
ExecStart=/usr/local/bin/llm-expose start --channel telegram-main
Restart=always
RestartSec=10
StandardOutput=journal
StandardError=journal

[Install]
WantedBy=multi-user.target
EOF

# Enable service (starts on boot)
sudo systemctl enable llm-expose-telegram

# Start service
sudo systemctl start llm-expose-telegram

# Check status
sudo systemctl status llm-expose-telegram

# View logs
sudo journalctl -u llm-expose-telegram -f

---

Specialized Use Cases

Example 10: Personal Knowledge Assistant

Telegram bot with access to your personal wiki/knowledge base.

# Add model
llm-expose add model --name gpt4o --provider openai --model-id gpt-4o

# Add filesystem MCP for your knowledge base
llm-expose add mcp
# • filesystem accessing ~/documents/wiki/

# Create Telegram channel with MCP
llm-expose add channel
# • telegram-assistant
# • Model: gpt4o
# • MCP: filesystem

# Only pair yourself
llm-expose add pair <your-id> --channel telegram-assistant

# Usage: "Summarize my notes on Python" → Bot reads your wiki files and responds

Example 11: Content Creator Assistant

Discord server for content team with tool access.

# Models for different tasks
llm-expose add model --name creative --provider anthropic --model-id claude-3-opus
llm-expose add model --name editor --provider openai --model-id gpt-4o

# Discord channels
llm-expose add channel  # discord-writers (uses creative)
llm-expose add channel  # discord-editors (uses editor)

# Pair team members
llm-expose add pair <writer-id> --channel discord-writers
llm-expose add pair <editor-id> --channel discord-editors
llm-expose add pair <manager-id> --channel discord-writers  # Managers see both
llm-expose add pair <manager-id> --channel discord-editors

# With MCP tools, writers can:
# - Ask bot for writing suggestions
# - Get editing feedback
# - Research topics

---

Troubleshooting Common Setup Issues

See the Troubleshooting Guide for detailed help with:

• Bot not responding
• API key errors
• Configuration issues
• Performance problems
• Deployment issues

---

Tips & Tricks

Pro Tips

Monitoring Multiple BotsQuick Configuration ResetTesting Bot LocallyKeeping Logs for Debugging

# Start all channels at once (in background)
for channel in telegram-main discord-support; do
  llm-expose start --channel $channel > logs/$channel.log 2>&1 &
done

# View all logs
tail -f logs/*.log

# To reset without losing everything, backup first:
cp -r ~/.config/llm-expose ~/.config/llm-expose.backup

# Then carefully update just what you need:
llm-expose config set --channel mychannel --model new-model-id

# Use Ollama or local endpoint for free testing
llm-expose add model --name test --provider openai \
  --model-id llama2 \
  --api-base http://localhost:11434/v1

# No API key needed!

# Start with persistent logging
llm-expose start --channel my-channel \
  > ~/.logs/llm-expose-$(date +%Y%m%d).log 2>&1 &

# Check with:
tail -f ~/.logs/llm-expose-*.log

---

Next Steps

• Explore MCP: Check the MCP Integration Guide for more tool options
• Deploy to production: See Deployment Guide for cloud options
• Advanced configuration: Review Provider Configuration for more models
• Custom development: Build your own MCP servers or llm-expose extensions

---

Have an example you'd like to share? Contribute to the docs!