Telegram AI Agent with Message Buffering

This n8n workflow enables Telegram chatbots to handle users sending several short messages rapidly, treating them as a single coherent input rather than responding to each individually. Messages are queued in a Supabase PostgreSQL table, allowing the system to wait briefly for additional inputs before processing.

Upon a brief wait period (default 10 seconds), the workflow retrieves all pending messages for the user, concatenates them, and feeds the combined text to an OpenAI AI Agent for a thoughtful, context-aware response. The reply is sent back via Telegram, and the queue is cleared to prepare for the next interaction. This prevents fragmented conversations and improves response quality.

Ideal for customer support bots, virtual assistants, or any Telegram-based AI interface where users type in bursts. Benefits include reduced API calls, more natural dialogue flow, cost savings on LLM usage, and enhanced user experience. Easily customizable by adjusting wait time, adding system prompts, or swapping LLMs.

Setup requires creating a 'message_queue' table in Supabase with user_id, message, and message_id columns, then adding credentials for Telegram, Supabase, OpenAI, and PostgreSQL. Activate and test by sending rapid messages to the bot.