Create a session-based Telegram chatbot with GPT-4o-mini and Google Sheets
Needs changesNeeds changes. This template uses 1 thing(s) that n8n 3.0 removes: AI Agent v1 (old agent modes). 1 breaking item(s) found. Static check, last run 2026-10-19.
What to replace before upgrading
| Removed in 3.0 | Replace with |
|---|---|
| AI Agent v1 (old agent modes) Telegram Chatbot (v1.9) | Latest AI Agent node (Tools Agent). For SQL Agent, use a Postgres or MySQL tool with a recent AI Agent. |
Nodes used (12)
- OpenAI Chat Model
@n8n/n8n-nodes-langchain.lmChatOpenAi - Simple Memory
@n8n/n8n-nodes-langchain.memoryBufferWindow - Telegram Trigger
n8n-nodes-base.telegramTrigger - Switch
n8n-nodes-base.switch - Google Sheets
n8n-nodes-base.googleSheets - Telegram
n8n-nodes-base.telegram - Summarization Chain
@n8n/n8n-nodes-langchain.chainSummarization - Basic LLM Chain
@n8n/n8n-nodes-langchain.chainLlm - Edit Fields (Set)
n8n-nodes-base.set - If
n8n-nodes-base.if - Code
n8n-nodes-base.code - AI Agent
@n8n/n8n-nodes-langchain.agent
What it does
How It Works This workflow creates an AI-powered Telegram chatbot with session management, allowing users to: - Start new conversations (/new). - Check current sessions (/current). - Resume past sessions (/resume). - Get summaries (/summary). - Ask questions (/question). Key Components: - Session Management: - Uses Google Sheets to track active/expired sessions (storing SESSION IDs and STATE). - /new creates a session; /resume reactivates past ones. - AI Processing: - OpenAI GPT-4 generates responses with contextual memory (via Simple Memory node). - Summarization: Condenses past conversations when requested. - Data Logging: - All interactions (prompts/responses) are saved to Google Sheets for audit and retrieval. - User Interaction: - Telegram commands trigger specific actions (e.g., /question query fetches answers from session history). --- Main Advantages 1. Multi-session Handling Each user can create, manage, and switch between multiple sessions independently, perfect for organizing different conversations without confusion. 2. Persistent Memory Conversations are stored in Google Sheets, ensuring that chat history and session states are preserved even if the server or n8n instance restarts. 3. Commands for Full Control With intuitive commands like /new, /current, /resume, /summary, and /question, users can manage sessions easily without needing a web interface. 4. Smart Summarization and Q&A Thanks to OpenAI models, the workflow can summarize entire conversations or answer specific questions about past discussions, saving time and improving the chatbot’s usability. 5. Easy Setup and Scalability By using Google Sheets instead of a database, the workflow is easy to clone, modify, and deploy — ideal for quick prototyping or lightweight production uses. 6. Modular and Extensible Each logical block (new session, get current session, resume, summarize, ask question) is modular, making it easy to extend the workflow with additional features like analytics, personalized greetings, or integrations with CRM systems. --- Setup Steps Prerequisites: - Telegram Bot Token: Create via BotFather. - Google Sheets: - Duplicate this template. - Two sheets: Session (active/inactive sessions) and Database (chat logs). - OpenAI API Key: For GPT-4 responses. Configuration: 1. Telegram Integration: - Add your bot token to the Telegram Trigger and Telegram Send nodes. 2. Google Sheets Setup: - Authenticate the Google Sheets nodes with OAuth. - Ensure sheet names (Session, Database) and column mappings match the template. 3. OpenAI & Memory: - Add your API key to the OpenAI Chat Model nodes. - Adjust contextWindowLength in the Simple Memory node for conversation history length. 4. Testing: - Use Telegram commands to test: - /new: Starts a session. - /question query: Tests AI responses. - /summary: Checks summarization. 5. Deployment: - Activate the workflow; the bot will respond to Telegram messages in real-time. --- Need help customizing? Contact me(mailto:info@n3w.it) for consulting and support or add me on Linkedin.