Telegram Memory AI Agent: Text & Voice Input → OpenAI Transcription → Persistent Memory Conversational Agent

Problem:

A team needed a Telegram-based AI assistant that could handle both text and voice messages, maintain persistent memory of past conversations beyond a single session, and respond intelligently in Telegram, standard chatbots had no voice input capability and lost all context between conversations.

Solution: Built a Telegram-triggered n8n AI agent that routes incoming messages through a Switch node based on message type, text messages pass directly to the field editor while audio messages are first retrieved from Telegram, transcribed to text via OpenAI's transcription model, then joined at the same field editor, the unified text input is passed to an OpenAI-powered AI Agent equipped with two persistent memory tools: Store_Memory (creates a new row for each conversation turn) and Get_Memory (retrieves all prior rows to reconstruct full conversation history), the agent's response is sent back to the user directly in Telegram.

Result:

Delivered a voice and text capable Telegram AI assistant with genuine persistent memory, the agent recalls previous conversations across sessions, understands spoken voice messages, and responds contextually in Telegram, functioning as a true long-term conversational assistant rather than a stateless chatbot.