DeepSeek AI Agent: Telegram Bot with Google Docs Long-Term Memory

Unlock Personalized & Context-Aware Interactions with this AI Agent

This n8n workflow deploys a sophisticated AI Agent on Telegram, powered by DeepSeek AI models (deepseek-chat or deepseek-reasoner). It's designed for rich, ongoing conversations by leveraging a unique long-term memory system built with Google Docs. The agent can recall past user interactions, preferences, and key information, delivering truly personalized and contextually relevant responses.

This AI Agent excels at:

• Intelligent Conversation: Engaging users naturally on Telegram, understanding intent, and providing helpful answers using DeepSeek's advanced language capabilities.
• Personalized Memory: Automatically identifying and saving noteworthy information from conversations into a designated Google Doc. This 'memory' is then retrieved in subsequent interactions to tailor responses.
• Contextual Understanding: Using both recent conversation history and long-term stored memories to maintain coherent and relevant dialogue.

Key Features & Benefits

• DeepSeek AI Integration: Utilizes powerful and cost-effective DeepSeek LLMs (OpenAI API compatible) for high-quality text generation and reasoning.
• Persistent Long-Term Memory: Connects to Google Docs to store and retrieve user-specific information, enabling conversations that evolve over time.
• Telegram Bot: Interacts with users directly within Telegram, a popular messaging platform.
• User-Specific Context: The AI Agent is prompted to remember the user's first name and use stored memories to personalize interactions.
• Memory Management Tool: Includes a 'Save Memory' tool (via Google Docs) that the AI Agent can invoke to store crucial details.
• Customizable System Prompt: Define the AI Agent's personality, rules, and how it should use memories and tools.
• Message Routing: Basic routing for different message types (text, audio, image), with text messages processed by the AI.
• Optional Short-Term Memory: Includes a 'Window Buffer Memory' node (can be connected for session-specific memory, if needed).
• User Validation (Placeholder): A basic mechanism to check user ID, which can be customized for specific user access or removed.

• An n8n instance (Cloud or self-hosted).
• DeepSeek API Key (or other OpenAI-compatible API key and endpoint if adapting).
• Telegram Bot Token (obtain from BotFather) and a publicly accessible n8n webhook URL for Telegram updates.
• Google Cloud Project with Google Docs API enabled, OAuth 2.0 credentials, and the ID of a Google Doc to be used for storing memories.
• (Optional) Customize the 'Validation' node values or logic if specific user restriction is needed.

1. Download the n8n workflow JSON file.
2. Import the workflow into your n8n instance.
3. Telegram Setup:
   • In the 'Listen for Telegram Events' (Webhook) node, copy the TEST webhook URL. Register this URL with your Telegram Bot using the method described in the workflow's 'How to set up a Telegram Bot WebHook' sticky note (requires your Bot Token).
   • Configure the 'Telegram Response', 'Error message', and 'Response Error message' nodes with your Telegram API credentials.
4. User Validation (Optional):
   • Modify the 'Validation' node: update the first_name, last_name, and id values to match the authorized Telegram user. Alternatively, remove or adapt this validation logic as needed.
5. DeepSeek AI Configuration:
   • In n8n, create new 'OpenAI API' credentials, naming it (e.g., 'DeepSeek API').
   • Enter your DeepSeek API Key.
   • Set the 'Base URL' to https://api.deepseek.com (or https://api.deepseek.com/v1).
   • Select these credentials in the 'DeepSeek-V3 Chat' (or 'DeepSeek-R1 Reasoning' if you switch) node. This node is connected to the 'AI Agent'.
6. Google Docs Long-Term Memory:
   • Create or identify a Google Doc to store memories. Note its Document ID (from the URL).
   • Configure the 'Retrieve Long Term Memories' (Google Docs node) with your Google Docs API credentials and the Document ID/URL.
   • Configure the 'Save Long Term Memories' (Google Docs Tool node) with the same credentials and Document ID/URL. Ensure the tool description is set (e.g., 'Save memories').
7. AI Agent Configuration:
   • Review and customize the systemMessage in the 'AI Agent' node to define its personality, rules, and how it interacts with memories. Pay attention to the {{ $('Retrieve Long Term Memories').item.json.content }} expression which injects memories.
8. (Optional) Short-Term Memory: If you wish to use session-specific short-term memory, connect the 'Window Buffer Memory' node's output to the 'AI Agent' node's 'Memory' input. Configure the 'Window Buffer Memory' node (e.g., session key like {{ $('Listen for Telegram Events').item.json.body.message.from.id }}).
9. Activate the workflow. After activation, use the PRODUCTION webhook URL from the 'Listen for Telegram Events' node for your Telegram bot.