adkt

ADK Telegram (adkt)

A Python library that simplifies the integration of Google's Agent Development Kit (ADK) with Telegram bots. It provides a high-level wrapper using python-telegram-bot to quickly deploy ADK-powered agents that can interact via Telegram, including proactively sending messages back to the user.

Key Features

• Simplified Boilerplate: Reduces the setup code needed to connect an ADK agent to a Telegram interface.
• Session Management: Automatically handles ADK sessions per unique Telegram chat ID using InMemorySessionService.
• Runner Management: Manages ADK Runner instances for each active chat.
• Agent-Initiated Messages: Includes a mechanism to provide a specific tool (send_telegram_message_tool) to your ADK agent, allowing it to send messages back to the Telegram chat proactively during its execution flow.
• Built on python-telegram-bot: Leverages the popular and robust python-telegram-bot library (v21+ recommended).
• Configurable: Supports restricting bot access to specific chat IDs and enabling debug logging.

Installation

Install adkt and its direct dependency python-telegram-bot:

pip install adkt python-telegram-bot~=21.0

Prerequisites:

Your agent application code will require additional libraries and setup:

1. Google ADK: You need to have Google's Agent Development Kit installed and configured in your environment. The installation mechanism for ADK might vary (e.g., specific Google Cloud packages, internal tooling, etc.). Ensure your environment allows you to import google.adk.
2. LLM Libraries: Your agent requires a connection to a Large Language Model. The example uses litellm to connect to a local Ollama instance. Install necessary libraries for your chosen LLM connection method (e.g., litellm, google-generativeai, openai).

   pip install litellm # Example for Ollama/LiteLLM
   

3. Environment Variables: The library often relies on environment variables for configuration (like API keys). The example uses python-dotenv.

   pip install python-dotenv
   

Core Concept: Agent Sending Messages

A key feature of adkt is enabling the ADK agent itself to send messages back to the user. This is achieved by:

1. The TelegramBot wrapper dynamically creates a function (send_telegram_message_tool internally) for each chat that knows how to send a message to that specific chat ID via the Telegram Bot API.
2. This function is passed as an argument to your generate_agent_fn when a new agent/runner is created for a chat.
3. Your generate_agent_fn must accept this tool (e.g., as send_telegram_message_tool) and include it in the tools list when creating the LlmAgent.
4. You must instruct your LLM (via the agent's instruction prompt) to use this specific tool whenever it needs to communicate back to the user.

Basic Usage

Here's a typical structure for using adkt:

1. Set up your Environment (.env file):

TELEGRAM_TOKEN="YOUR_TELEGRAM_BOT_TOKEN_HERE"
# Optional: Restrict to specific chat IDs (comma-separated)
# TELEGRAM_CHAT_IDS="12345678,98765432"

• Get your TELEGRAM_TOKEN from BotFather on Telegram.

2. Create your main Python script (e.g., main.py):

import os
from dotenv import load_dotenv

from adkt import start_agent_bot # Import the main function from adkt
from google.adk.agents import LlmAgent
from google.adk.models.lite_llm import LiteLlm # Example using LiteLLM

# Load environment variables from .env file
load_dotenv()

# --- Define Your Agent's Tools ---
def lookup_order_status(order_id: str) -> dict:
  """Fetches the current status of a customer's order using its ID.

  Use this tool ONLY when a user explicitly asks for the status of
  a specific order and provides the order ID. Do not use it for
  general inquiries.

  Args:
      order_id: The unique identifier of the order to look up.

  Returns:
      A dictionary containing the order status.
      Possible statuses: 'shipped', 'processing', 'pending', 'error'.
      Example success: {'status': 'shipped', 'tracking_number': '1Z9...'}
      Example error: {'status': 'error', 'error_message': 'Order ID not found.'}
  """
  print(f"Tool: Looking up status for order_id: {order_id}")
  # Replace with your actual order lookup logic
  if order_id == "123":
      return {"status": 'processing', "tracking_number": '1ZABCDEF987'}
  else:
      return {"status": 'error', 'error_message': f'Order ID {order_id} not found.'}

# --- Define Your Agent Generation Function ---
# This function MUST accept 'user_id' and the messaging tool ('send_telegram_message_tool')
def generate_agent(user_id: str, send_telegram_message_tool: callable) -> LlmAgent:
    """
    Creates and configures the ADK Agent instance for a specific user.

    Args:
        user_id: The unique identifier for the user (derived from chat_id).
        send_telegram_message_tool: The tool function provided by adkt
                                     for sending messages back to this user's chat.
    Returns:
        An configured LlmAgent instance.
    """
    print(f"Generating agent for user_id: {user_id}")

    # Configure your LLM connection (Example uses LiteLLM for Ollama)
    # Ensure your Ollama server is running and has the model pulled.
    llm_model = LiteLlm(model="ollama/yasserrmd/Llama-4-Scout-17B-16E-Instruct", api_base="http://localhost:11434") # Adjust api_base if needed

    agent_instruction = """You are a helpful assistant interacting via Telegram.
Your goal is to answer user questions and fulfill requests.
You have access to the following tools:
1. `lookup_order_status`: Use this ONLY when the user asks for the status of a specific order and provides the ID.
2. `send_telegram_message`: This is the **ONLY** way to communicate back to the user. Use this tool for *all* responses, including asking for clarification, providing updates, confirming actions, and delivering the final answer. Do not output plain text, always use this tool.
"""

    agent = LlmAgent(
        model=llm_model,
        name=f"telegram_agent_{user_id}", # Unique name per user is good practice
        instruction=agent_instruction,
        tools=[
            lookup_order_status,
            send_telegram_message_tool # IMPORTANT: Include the provided messaging tool
            ]
    )
    return agent

# --- Start the Bot ---
if __name__ == "__main__":
    print("Starting ADK Telegram Bot...")
    telegram_token = os.environ.get("TELEGRAM_TOKEN")
    if not telegram_token:
        raise ValueError("TELEGRAM_TOKEN not found in environment variables.")

    # Optional: Load restricted chat IDs if set
    restricted_chats_str = os.environ.get("TELEGRAM_CHAT_IDS")
    restricted_chat_ids = restricted_chats_str.split(',') if restricted_chats_str else None

    start_agent_bot(
        telegram_token=telegram_token,
        generate_agent_fn=generate_agent, # Pass your agent generator function
        telegram_chat_ids=restricted_chat_ids, # Pass list of allowed chat IDs, or None
        debug=True # Enable verbose logging from adkt and ADK
    )

3. Run the script:

python main.py

4. Interact with your Bot: Find your bot on Telegram and start sending messages. The agent defined in generate_agent will process them, potentially using the lookup_order_status tool and sending responses back via the send_telegram_message_tool. Use the /get_chat_id command if you need to find your chat ID for the TELEGRAM_CHAT_IDS setting.

Contributing

Contributions are welcome! Please feel free to submit a pull request or open an issue.

License

This project is licensed under the MIT License.