Personal AI Agent with Telegram & Apple Watch Support

A personal AI assistant powered by Ollama (Llama 3.1) that integrates with Telegram BotFather and provides a REST API for Apple Watch and other clients.

• AI Chat: Natural language conversations with memory context
• Health Tracking: Log calories, workouts, weight, water, and mood
• Reminders: Set and manage reminders with natural language time parsing
• Lists: Manage todo, grocery, wishlist, watchlist, reading, ideas, and goals
• Notes: Save and retrieve notes
• Expenses: Track spending with categories
• Web Search: Search the web via DuckDuckGo
• Daily Summary: Get AI-powered summaries of your day

1. Ollama: Install and run Ollama with Llama 3.1

   # Install Ollama from https://ollama.ai
   ollama pull llama3.1
   ollama serve

2. Python 3.8+: Ensure Python is installed

3. Telegram Bot (optional): Create a bot via BotFather

   • Message @BotFather on Telegram
   • Create a new bot with /newbot
   • Copy the API token

1. Clone or navigate to the project directory

2. Create a virtual environment:

   python -m venv venv
   source venv/bin/activate  # On Windows: venv\Scripts\activate

3. Install dependencies:

   pip install -r requirements.txt

4. Configure environment variables in .env:

   TELEGRAM_TOKEN=your_telegram_bot_token
   ALLOWED_USER_ID=your_telegram_user_id
   OLLAMA_MODEL=llama3.1

   To get your Telegram user ID, message @userinfobot on Telegram.

Run the Telegram bot:

python bot.py

The bot will start polling for messages. Use /commands in Telegram to see all available commands.

Run the API server:

python api.py

The API will be available at http://localhost:5001 (port 5000 is often used by AirPlay on macOS)

GET /health

POST /chat
Content-Type: application/json

{
  "message": "Your message here"
}

POST /log/food
Content-Type: application/json

{
  "food": "ate chicken salad for lunch"
}

POST /log/workout
Content-Type: application/json

{
  "workout": "ran 5km"
}

POST /log/mood
Content-Type: application/json

{
  "mood": "feeling great"
}

POST /log/weight
Content-Type: application/json

{
  "weight": "75"
}

POST /log/water
Content-Type: application/json

{
  "glasses": 2
}

GET /reminders

POST /reminders
Content-Type: application/json

{
  "message": "remind me to call mom at 6pm"
}

GET /summary

GET /logs?category=all  # or: calories, fitness, mood, water, weight

POST /quick
Content-Type: application/json

{
  "action": "morning"  # Options: morning, evening, water, mood_good, mood_bad, workout
}

1. Create Siri Shortcuts that make HTTP requests to your API server

2. Example shortcuts:

   • Morning Check-in: POST /quick with {"action": "morning"}
   • Log Water: POST /log/water with {"glasses": 1}
   • Log Workout: POST /log/workout with {"workout": "completed workout"}
   • Daily Summary: GET /summary

3. Add these shortcuts to your Apple Watch face as complications

For a more integrated experience, you can build a simple watchOS app:

1. Create a new watchOS project in Xcode
2. Use URLSession to make HTTP requests to your API server
3. Design a simple UI with quick-action buttons

Example Swift code for making requests:

import Foundation

func sendMessage(_ message: String) {
    guard let url = URL(string: "http://your-server-ip:5000/chat") else { return }
    
    var request = URLRequest(url: url)
    request.httpMethod = "POST"
    request.setValue("application/json", forHTTPHeaderField: "Content-Type")
    
    let body: [String: String] = ["message": message]
    request.httpBody = try? JSONSerialization.data(withJSONObject: body)
    
    let task = URLSession.shared.dataTask(with: request) { data, _, _ in
        if let data = data,
           let response = try? JSONSerialization.jsonObject(with: data) as? [String: Any],
           let result = response["response"] as? String {
            print(result)
        }
    }
    task.resume()
}

For Apple Watch to access your API:

1. Local Network: Ensure your Mac and Apple Watch are on the same Wi-Fi network
2. Port Forwarding: If accessing from outside your local network, set up port forwarding on your router
3. Dynamic DNS: Use a service like ngrok for testing:

   ngrok http 5001

4. Port: The API runs on port 5001 by default (port 5000 is often used by AirPlay on macOS)
5. HTTPS: For production, use a reverse proxy with SSL (nginx, Caddy, etc.)

• /want [item] - Add to wishlist
• /todo [task] - Add to to-do list
• /grocery [item] - Add to grocery list
• /watchlist [movie/show] - Add to watchlist
• /reading [book] - Add to reading list
• /ideas [idea] - Save an idea
• /goals [goal] - Add a goal
• /showlist [name] - Show a list
• /done [id] - Mark item as done
• /delete [id] - Delete item
• /clearlist [name] - Clear entire list

• /calories [food] - Log food & auto-estimate calories
• /workout [description] - Log a workout
• /weight [kg] - Log your weight
• /water [glasses] - Log water intake
• /mood [how you feel] - Log your mood
• /health - Show today's health summary

• /remind [message] - Save a reminder
• /reminders - Show all reminders
• /note [title] | [content] - Save a note
• /notes - List all notes
• /shownote [id] - View a note

• /spend [amount] [category] [description] - Log an expense
• /expenses - Today's expenses
• /allexpenses - All recent expenses

• /summary - Today's full summary
• /logs - Show all today's logs
• /ask [question] - Ask the AI anything
• /search [query] - Search the web

All data is stored in SQLite database at ~/myagent/memory.db

Tables:

• logs - Food, fitness, mood, water, weight entries
• reminders - Reminders with time and recurrence
• lists - Todo, grocery, wishlist, etc.
• notes - Saved notes
• expenses - Expense tracking
• habits - Habit tracking
• snippets - Text snippets

• Ensure Ollama is running: ollama serve
• Check the model is available: ollama list
• Verify OLLAMA_MODEL in .env matches an installed model

• Verify TELEGRAM_TOKEN is correct
• Check ALLOWED_USER_ID matches your Telegram user ID
• Ensure the bot is running: python bot.py

• Check firewall settings on your Mac
• Ensure devices are on the same network
• Test with curl: curl http://localhost:5000/health
• Use ngrok for external testing: ngrok http 5000

• The API server currently has no authentication - add API keys for production use
• Telegram bot is restricted to a single user ID
• Consider adding HTTPS for production deployments
• Keep your .env file secure and never commit it to version control

MIT License - Feel free to modify and use for personal projects.